diff --git a/.agents/ci-caching.md b/.agents/ci-caching.md
index 17cc6001d..26aba82e6 100644
--- a/.agents/ci-caching.md
+++ b/.agents/ci-caching.md
@@ -122,7 +122,7 @@ The per-backend prefix match only sees files under a backend's own directory, so
 
 | Changed path | Rebuilds |
 |---|---|
-| `backend/backend.proto` | everything (all languages compile or copy it) |
+| `backend/backend.proto` | nothing if the edit is additive-only, otherwise everything (see below) |
 | `backend/Dockerfile.<x>` | the Linux entries whose `dockerfile:` names it |
 | `backend/python/common/` | Python, Linux + Darwin |
 | `scripts/build/package-gpu-libs.sh` | Python, Linux only |
@@ -132,6 +132,17 @@ The per-backend prefix match only sees files under a backend's own directory, so
 
 Deliberately excluded: `backend/index.yaml` (gallery metadata, never enters an image), `.github/backend-matrix.yml` (adding a backend would rebuild all of them), `backend/Dockerfile.base-grpc-builder` (owned by `base-images.yml`), and the root `Makefile` (touched in ~11% of commits, and its backend-relevant edits arrive alongside the backend directory anyway). `make test-ci-scripts` pins all of this.
 
+#### `backend/backend.proto` is content-filtered, not path-filtered
+
+Every language consumes the proto, so a path rule for it can only ever say "rebuild all 473 images". It changes in ~1.3% of commits, and that was enough to make it the single largest CI cost driver in the repo: on 2026-07-29 four runs totalling 935 queued jobs traced to nothing but a proto edit, one of which (#11158) was a six-line diff adding `bool cache_prompt = 8;`.
+
+An additive proto edit cannot change how a backend that never references the new symbol behaves, so `filterMatrix()` suppresses the rule for one. `changed-backends.js` fetches `backend/backend.proto` at the base revision (same contents-API pattern as `.github/backend-matrix.yml`) and hands both texts to `protoChangeIsAdditive()`, which compares them structurally rather than textually:
+
+- **Additive, rebuilds nothing**: a new field with an unused number, a new message, a new enum value, a new RPC. Comment, whitespace and ordering changes also land here.
+- **Breaking, rebuilds everything**: a removed, renumbered, retyped or renamed field, a dropped RPC, a changed `option` or `package`. So does an unresolvable base revision, matching the run-all posture used for a truncated diff.
+
+Checked against every proto commit in the preceding six months, all nine resolvable ones classify as additive. Note the tradeoff this accepts: generated stubs do change for an additive edit, so image bytes would differ on a rebuild even though behavior does not. That is the same standard already applied when the filter declines to rebuild on unrelated `pkg/` changes, and the weekly cron remains the backstop.
+
 The Sunday 06:00 UTC cron on `backend.yml` exists specifically because path filtering can leave Python backends frozen on stale wheels. `DEPS_REFRESH` (below) only fires when the build actually runs, so an untouched Python backend would never re-resolve its unpinned deps. The weekly cron is the safety net.
 
 ## The `DEPS_REFRESH` cache-buster (Python backends)
diff --git a/.github/backend-matrix.yml b/.github/backend-matrix.yml
index 2ae8a26af..b549e4648 100644
--- a/.github/backend-matrix.yml
+++ b/.github/backend-matrix.yml
@@ -756,6 +756,19 @@ include:
     dockerfile: "./backend/Dockerfile.golang"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: 'cublas'
+    cuda-major-version: "12"
+    cuda-minor-version: "8"
+    platforms: 'linux/amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-nvidia-cuda-12-trellis2cpp'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "trellis2cpp"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'cublas'
     cuda-major-version: "12"
     cuda-minor-version: "8"
@@ -1716,6 +1729,19 @@ include:
     dockerfile: "./backend/Dockerfile.golang"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: 'cublas'
+    cuda-major-version: "13"
+    cuda-minor-version: "0"
+    platforms: 'linux/amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-nvidia-cuda-13-trellis2cpp'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "trellis2cpp"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'cublas'
     cuda-major-version: "13"
     cuda-minor-version: "0"
@@ -1729,6 +1755,19 @@ include:
     backend: "stablediffusion-ggml"
     dockerfile: "./backend/Dockerfile.golang"
     context: "./"
+  - build-type: 'cublas'
+    cuda-major-version: "13"
+    cuda-minor-version: "0"
+    platforms: 'linux/arm64'
+    skip-drivers: 'false'
+    tag-latest: 'auto'
+    tag-suffix: '-nvidia-l4t-cuda-13-arm64-trellis2cpp'
+    base-image: "ubuntu:24.04"
+    ubuntu-version: '2404'
+    runs-on: 'ubuntu-24.04-arm'
+    backend: "trellis2cpp"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
   - build-type: 'cublas'
     cuda-major-version: "13"
     cuda-minor-version: "0"
@@ -3267,6 +3306,35 @@ include:
     dockerfile: "./backend/Dockerfile.golang"
     context: "./"
     ubuntu-version: '2404'
+  # trellis2cpp
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    platform-tag: 'amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-trellis2cpp'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "trellis2cpp"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2404'
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/arm64'
+    platform-tag: 'arm64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-trellis2cpp'
+    runs-on: 'ubuntu-24.04-arm'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "trellis2cpp"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2404'
   # sam3-cpp
   - build-type: ''
     cuda-major-version: ""
@@ -3592,6 +3660,34 @@ include:
     dockerfile: "./backend/Dockerfile.golang"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: 'vulkan'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    platform-tag: 'amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-vulkan-trellis2cpp'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "trellis2cpp"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2404'
+  - build-type: 'vulkan'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/arm64'
+    platform-tag: 'arm64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-vulkan-trellis2cpp'
+    runs-on: 'ubuntu-24.04-arm'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "trellis2cpp"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'cublas'
     cuda-major-version: "12"
     cuda-minor-version: "0"
@@ -3605,6 +3701,19 @@ include:
     dockerfile: "./backend/Dockerfile.golang"
     context: "./"
     ubuntu-version: '2204'
+  - build-type: 'cublas'
+    cuda-major-version: "12"
+    cuda-minor-version: "0"
+    platforms: 'linux/arm64'
+    skip-drivers: 'false'
+    tag-latest: 'auto'
+    tag-suffix: '-nvidia-l4t-arm64-trellis2cpp'
+    base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
+    runs-on: 'ubuntu-24.04-arm'
+    backend: "trellis2cpp"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2204'
   - build-type: 'cublas'
     cuda-major-version: "12"
     cuda-minor-version: "0"
@@ -5436,6 +5545,35 @@ include:
     dockerfile: "./backend/Dockerfile.golang"
     context: "./"
     ubuntu-version: '2404'
+  # valkey-store
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    platform-tag: 'amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-valkey-store'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "valkey-store"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2404'
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/arm64'
+    platform-tag: 'arm64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-valkey-store'
+    runs-on: 'ubuntu-24.04-arm'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "valkey-store"
+    dockerfile: "./backend/Dockerfile.golang"
+    context: "./"
+    ubuntu-version: '2404'
   # rfdetr
   - build-type: ''
     cuda-major-version: ""
@@ -5977,6 +6115,10 @@ includeDarwin:
     tag-suffix: "-metal-darwin-arm64-stablediffusion-ggml"
     build-type: "metal"
     lang: "go"
+  - backend: "trellis2cpp"
+    tag-suffix: "-metal-darwin-arm64-trellis2cpp"
+    build-type: "metal"
+    lang: "go"
   - backend: "whisper"
     tag-suffix: "-metal-darwin-arm64-whisper"
     build-type: "metal"
@@ -6154,6 +6296,10 @@ includeDarwin:
     tag-suffix: "-metal-darwin-arm64-cloud-proxy"
     build-type: "metal"
     lang: "go"
+  - backend: "valkey-store"
+    tag-suffix: "-metal-darwin-arm64-valkey-store"
+    build-type: "metal"
+    lang: "go"
   - backend: "llama-cpp-quantization"
     tag-suffix: "-metal-darwin-arm64-llama-cpp-quantization"
     build-type: "mps"
diff --git a/.github/workflows/bump_deps.yaml b/.github/workflows/bump_deps.yaml
index 7cd472f4e..b6dbbb449 100644
--- a/.github/workflows/bump_deps.yaml
+++ b/.github/workflows/bump_deps.yaml
@@ -78,6 +78,10 @@ jobs:
             variable: "STABLEDIFFUSION_GGML_VERSION"
             branch: "master"
             file: "backend/go/stablediffusion-ggml/Makefile"
+          - repository: "localai-org/trellis2cpp"
+            variable: "TRELLIS2CPP_VERSION"
+            branch: "pbr-textures"
+            file: "backend/go/trellis2cpp/Makefile"
           - repository: "mudler/go-piper"
             variable: "PIPER_VERSION"
             branch: "master"
diff --git a/.github/workflows/test-extra.yml b/.github/workflows/test-extra.yml
index 899b98e60..96eb57155 100644
--- a/.github/workflows/test-extra.yml
+++ b/.github/workflows/test-extra.yml
@@ -38,6 +38,7 @@ jobs:
       acestep-cpp: ${{ steps.detect.outputs.acestep-cpp }}
       qwen3-tts-cpp: ${{ steps.detect.outputs.qwen3-tts-cpp }}
       magpie-tts-cpp: ${{ steps.detect.outputs.magpie-tts-cpp }}
+      trellis2cpp: ${{ steps.detect.outputs.trellis2cpp }}
       rfdetr-cpp: ${{ steps.detect.outputs.rfdetr-cpp }}
       locate-anything-cpp: ${{ steps.detect.outputs.locate-anything-cpp }}
       vibevoice-cpp: ${{ steps.detect.outputs.vibevoice-cpp }}
@@ -935,6 +936,41 @@ jobs:
       - name: Test rfdetr-cpp
         run: |
           make --jobs=5 --output-sync=target -C backend/go/rfdetr-cpp test
+  # Weight-free packaged-backend smoke for trellis2cpp. Starting run.sh loads
+  # libtrellis2 + ggml, resolves the complete C ABI (including remeshing), and
+  # answers gRPC Health without downloading or loading the multi-GB model set.
+  tests-trellis2cpp:
+    needs: detect-changes
+    if: needs.detect-changes.outputs.trellis2cpp == 'true' || needs.detect-changes.outputs.run-all == 'true'
+    runs-on: ubuntu-latest
+    timeout-minutes: 90
+    steps:
+      - name: Clone
+        uses: actions/checkout@v7
+        with:
+          submodules: true
+      - name: Dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y build-essential cmake curl unzip
+      - name: Setup Go
+        uses: actions/setup-go@v5
+      - name: Display Go version
+        run: go version
+      - name: Proto Dependencies
+        run: |
+          curl -L -s https://github.com/protocolbuffers/protobuf/releases/download/v26.1/protoc-26.1-linux-x86_64.zip -o protoc.zip && \
+          unzip -j -d /usr/local/bin protoc.zip bin/protoc && \
+          rm protoc.zip
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
+          PATH="$PATH:$HOME/go/bin" make protogen-go
+      - name: Build trellis2cpp
+        run: |
+          make --jobs=5 --output-sync=target -C backend/go/trellis2cpp
+      - name: Test trellis2cpp
+        run: |
+          make --jobs=5 --output-sync=target -C backend/go/trellis2cpp test
   # Per-backend e2e for locate-anything-cpp: builds the .so + Go binary and
   # runs `make -C backend/go/locate-anything-cpp test`. test.sh fetches the
   # locate-anything-q8_0 GGUF (~6.3 GB, NVIDIA LocateAnything-3B) from the
diff --git a/.gitignore b/.gitignore
index 5286461b3..4fb77f49e 100644
--- a/.gitignore
+++ b/.gitignore
@@ -30,6 +30,7 @@ LocalAI
 # Go backend packages whose main lives under backend/go/.
 /cloud-proxy
 /local-store
+/valkey-store
 # prevent above rules from omitting the helm chart
 !charts/*
 # prevent above rules from omitting the api/localai folder
diff --git a/Makefile b/Makefile
index 81b00d7c8..bd3b21d9d 100644
--- a/Makefile
+++ b/Makefile
@@ -1,5 +1,5 @@
 # Disable parallel execution for backend builds
-.NOTPARALLEL: backends/diffusers backends/llama-cpp backends/turboquant backends/bonsai backends/outetts backends/piper backends/stablediffusion-ggml backends/whisper backends/crispasr backends/parakeet-cpp backends/moss-transcribe-cpp backends/faster-whisper backends/silero-vad backends/local-store backends/cloud-proxy backends/huggingface backends/rfdetr backends/rfdetr-cpp backends/insightface backends/speaker-recognition backends/kitten-tts backends/kokoro backends/chatterbox backends/llama-cpp-darwin backends/neutts build-darwin-python-backend build-darwin-go-backend backends/mlx backends/diffuser-darwin backends/mlx-vlm backends/mlx-audio backends/mlx-distributed backends/stablediffusion-ggml-darwin backends/vllm backends/vllm-omni backends/longcat-video backends/sglang backends/moonshine backends/pocket-tts backends/qwen-tts backends/faster-qwen3-tts backends/qwen-asr backends/nemo backends/voxcpm backends/whisperx backends/ace-step backends/acestep-cpp backends/fish-speech backends/voxtral backends/opus backends/trl backends/llama-cpp-quantization backends/kokoros backends/sam3-cpp backends/qwen3-tts-cpp backends/moss-tts-cpp backends/magpie-tts-cpp backends/vllm-cpp backends/omnivoice-cpp backends/vibevoice-cpp backends/localvqe backends/tinygrad backends/sherpa-onnx backends/ds4 backends/ds4-darwin backends/liquid-audio backends/supertonic backends/depth-anything-cpp backends/privacy-filter backends/privacy-filter-darwin
+.NOTPARALLEL: backends/diffusers backends/llama-cpp backends/turboquant backends/bonsai backends/outetts backends/piper backends/stablediffusion-ggml backends/trellis2cpp backends/trellis2cpp-darwin backends/whisper backends/crispasr backends/parakeet-cpp backends/moss-transcribe-cpp backends/faster-whisper backends/silero-vad backends/local-store backends/valkey-store backends/cloud-proxy backends/huggingface backends/rfdetr backends/rfdetr-cpp backends/insightface backends/speaker-recognition backends/kitten-tts backends/kokoro backends/chatterbox backends/llama-cpp-darwin backends/neutts build-darwin-python-backend build-darwin-go-backend backends/mlx backends/diffuser-darwin backends/mlx-vlm backends/mlx-audio backends/mlx-distributed backends/stablediffusion-ggml-darwin backends/vllm backends/vllm-omni backends/longcat-video backends/sglang backends/moonshine backends/pocket-tts backends/qwen-tts backends/faster-qwen3-tts backends/qwen-asr backends/nemo backends/voxcpm backends/whisperx backends/ace-step backends/acestep-cpp backends/fish-speech backends/voxtral backends/opus backends/trl backends/llama-cpp-quantization backends/kokoros backends/sam3-cpp backends/qwen3-tts-cpp backends/moss-tts-cpp backends/magpie-tts-cpp backends/vllm-cpp backends/omnivoice-cpp backends/vibevoice-cpp backends/localvqe backends/tinygrad backends/sherpa-onnx backends/ds4 backends/ds4-darwin backends/liquid-audio backends/supertonic backends/depth-anything-cpp backends/privacy-filter backends/privacy-filter-darwin
 
 GOCMD=go
 GOTEST=$(GOCMD) test
@@ -69,7 +69,7 @@ else
 	GORELEASER=$(shell which goreleaser)
 endif
 
-TEST_PATHS?=./api/... ./pkg/... ./core/... ./backend/go/cloud-proxy/... ./backend/go/local-store/...
+TEST_PATHS?=./api/... ./pkg/... ./core/... ./backend/go/cloud-proxy/... ./backend/go/local-store/... ./backend/go/valkey-store/...
 
 ## Coverage output and the committed baseline that CI compares against.
 ## The gate is strict: total coverage must never decrease (no tolerance).
@@ -386,6 +386,15 @@ test-stores: backends/local-store
 	BACKENDS_PATH=$(abspath ./)/backends \
 	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --flake-attempts $(TEST_FLAKES) -v -r tests/integration
 
+## Valkey-backed vector-store integration. Requires a running Valkey Search
+## server (valkey/valkey-bundle:9.1.0) reachable at $$VALKEY_ADDR — the suite
+## skips itself when VALKEY_ADDR is unset. Builds the backend on demand and
+## points the model loader at it via BACKENDS_PATH. Label-filtered to the
+## valkey specs so it does not also run the in-memory local-store suite.
+test-valkey-store: backends/valkey-store
+	BACKENDS_PATH=$(abspath ./)/backends \
+	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --flake-attempts $(TEST_FLAKES) --label-filter='valkey' -v -r tests/integration
+
 test-opus:
 	@echo 'Running opus backend tests'
 	$(MAKE) -C backend/go/opus libopusshim.so
@@ -594,6 +603,8 @@ prepare-test-extra: protogen-python
 	$(MAKE) -C backend/rust/kokoros kokoros-grpc
 	$(MAKE) -C backend/go/rfdetr-cpp
 	$(MAKE) -C backend/go/locate-anything-cpp
+	$(MAKE) -C backend/go/trellis2cpp
+	$(MAKE) -C backend/go/valkey-store
 
 test-extra: prepare-test-extra
 	$(MAKE) -C backend/python/transformers test
@@ -626,6 +637,8 @@ test-extra: prepare-test-extra
 	$(MAKE) -C backend/go/depth-anything-cpp test
 	$(MAKE) -C backend/go/supertonic test
 	$(MAKE) -C backend/go/vllm-cpp test
+	$(MAKE) -C backend/go/trellis2cpp test
+	$(MAKE) -C backend/go/valkey-store test
 
 ##
 ## End-to-end gRPC tests that exercise a built backend container image.
@@ -1218,6 +1231,10 @@ backends/stablediffusion-ggml-darwin:
 	BACKEND=stablediffusion-ggml BUILD_TYPE=metal $(MAKE) build-darwin-go-backend
 	./local-ai backends install "ocifile://$(abspath ./backend-images/stablediffusion-ggml.tar)"
 
+backends/trellis2cpp-darwin:
+	BACKEND=trellis2cpp BUILD_TYPE=metal $(MAKE) build-darwin-go-backend
+	./local-ai backends install "ocifile://$(abspath ./backend-images/trellis2cpp.tar)"
+
 backend-images:
 	mkdir -p backend-images
 
@@ -1245,10 +1262,12 @@ BACKEND_PRIVACY_FILTER = privacy-filter|privacy-filter|.|false|false
 # Golang backends
 BACKEND_PIPER = piper|golang|.|false|true
 BACKEND_LOCAL_STORE = local-store|golang|.|false|true
+BACKEND_VALKEY_STORE = valkey-store|golang|.|false|true
 BACKEND_CLOUD_PROXY = cloud-proxy|golang|.|false|true
 BACKEND_HUGGINGFACE = huggingface|golang|.|false|true
 BACKEND_SILERO_VAD = silero-vad|golang|.|false|true
 BACKEND_STABLEDIFFUSION_GGML = stablediffusion-ggml|golang|.|--progress=plain|true
+BACKEND_TRELLIS2CPP = trellis2cpp|golang|.|--progress=plain|true
 BACKEND_WHISPER = whisper|golang|.|false|true
 BACKEND_CRISPASR = crispasr|golang|.|false|true
 BACKEND_PARAKEET_CPP = parakeet-cpp|golang|.|false|true
@@ -1344,10 +1363,12 @@ $(eval $(call generate-docker-build-target,$(BACKEND_DS4)))
 $(eval $(call generate-docker-build-target,$(BACKEND_PRIVACY_FILTER)))
 $(eval $(call generate-docker-build-target,$(BACKEND_PIPER)))
 $(eval $(call generate-docker-build-target,$(BACKEND_LOCAL_STORE)))
+$(eval $(call generate-docker-build-target,$(BACKEND_VALKEY_STORE)))
 $(eval $(call generate-docker-build-target,$(BACKEND_CLOUD_PROXY)))
 $(eval $(call generate-docker-build-target,$(BACKEND_HUGGINGFACE)))
 $(eval $(call generate-docker-build-target,$(BACKEND_SILERO_VAD)))
 $(eval $(call generate-docker-build-target,$(BACKEND_STABLEDIFFUSION_GGML)))
+$(eval $(call generate-docker-build-target,$(BACKEND_TRELLIS2CPP)))
 $(eval $(call generate-docker-build-target,$(BACKEND_WHISPER)))
 $(eval $(call generate-docker-build-target,$(BACKEND_CRISPASR)))
 $(eval $(call generate-docker-build-target,$(BACKEND_PARAKEET_CPP)))
@@ -1408,7 +1429,7 @@ $(eval $(call generate-docker-build-target,$(BACKEND_SUPERTONIC)))
 docker-save-%: backend-images
 	docker save local-ai-backend:$* -o backend-images/$*.tar
 
-docker-build-backends: docker-build-llama-cpp docker-build-ik-llama-cpp docker-build-turboquant docker-build-bonsai docker-build-ds4 docker-build-rerankers docker-build-vllm docker-build-vllm-omni docker-build-longcat-video docker-build-sglang docker-build-transformers docker-build-outetts docker-build-diffusers docker-build-kokoro docker-build-faster-whisper docker-build-crispasr docker-build-coqui docker-build-chatterbox docker-build-vibevoice docker-build-liquid-audio docker-build-moonshine docker-build-pocket-tts docker-build-qwen-tts docker-build-fish-speech docker-build-faster-qwen3-tts docker-build-qwen-asr docker-build-nemo docker-build-voxcpm docker-build-whisperx docker-build-ace-step docker-build-acestep-cpp docker-build-voxtral docker-build-mlx-distributed docker-build-trl docker-build-llama-cpp-quantization docker-build-tinygrad docker-build-kokoros docker-build-sam3-cpp docker-build-rfdetr-cpp docker-build-qwen3-tts-cpp docker-build-moss-tts-cpp docker-build-magpie-tts-cpp docker-build-vllm-cpp docker-build-omnivoice-cpp docker-build-vibevoice-cpp docker-build-localvqe docker-build-insightface docker-build-speaker-recognition docker-build-sherpa-onnx docker-build-cloud-proxy docker-build-supertonic docker-build-depth-anything-cpp docker-build-moss-transcribe-cpp docker-build-privacy-filter
+docker-build-backends: docker-build-llama-cpp docker-build-ik-llama-cpp docker-build-turboquant docker-build-bonsai docker-build-ds4 docker-build-rerankers docker-build-vllm docker-build-vllm-omni docker-build-longcat-video docker-build-sglang docker-build-transformers docker-build-outetts docker-build-diffusers docker-build-kokoro docker-build-faster-whisper docker-build-crispasr docker-build-coqui docker-build-chatterbox docker-build-vibevoice docker-build-liquid-audio docker-build-moonshine docker-build-pocket-tts docker-build-qwen-tts docker-build-fish-speech docker-build-faster-qwen3-tts docker-build-qwen-asr docker-build-nemo docker-build-voxcpm docker-build-whisperx docker-build-ace-step docker-build-acestep-cpp docker-build-voxtral docker-build-mlx-distributed docker-build-trl docker-build-llama-cpp-quantization docker-build-tinygrad docker-build-kokoros docker-build-sam3-cpp docker-build-rfdetr-cpp docker-build-qwen3-tts-cpp docker-build-moss-tts-cpp docker-build-magpie-tts-cpp docker-build-vllm-cpp docker-build-omnivoice-cpp docker-build-vibevoice-cpp docker-build-localvqe docker-build-insightface docker-build-speaker-recognition docker-build-sherpa-onnx docker-build-cloud-proxy docker-build-supertonic docker-build-depth-anything-cpp docker-build-moss-transcribe-cpp docker-build-privacy-filter docker-build-trellis2cpp docker-build-valkey-store
 
 ########################################################
 ### Mock Backend for E2E Tests
diff --git a/README.md b/README.md
index d2d0c1479..cbdc44b68 100644
--- a/README.md
+++ b/README.md
@@ -245,6 +245,7 @@ Most backends wrap a best-in-class upstream engine. A handful of them are native
 | [depth-anything.cpp](https://github.com/mudler/depth-anything.cpp) | Depth Anything 3 monocular metric depth + camera pose estimation |
 | [face-detect.cpp](https://github.com/mudler/face-detect.cpp) | Face detection, recognition, demographics and anti-spoofing (SCRFD/ArcFace, YuNet/SFace), replacing the Python insightface backend |
 | [free-splatter.cpp](https://github.com/localai-org/free-splatter.cpp) | Pose-free 3D reconstruction (FreeSplatter): turns a handful of plain photos into 3D Gaussians, no camera poses or GPU required |
+| [trellis2.cpp](https://github.com/localai-org/trellis2cpp) | C++/GGML port of Microsoft TRELLIS.2: single-image to textured 3D mesh (GLB with PBR materials) |
 | [privacy-filter.cpp](https://github.com/localai-org/privacy-filter.cpp) | Standalone GGML PII/NER token-classification engine powering LocalAI's PII redaction tier |
 | [LocalVQE](https://github.com/localai-org/LocalVQE) | Joint acoustic echo cancellation, noise suppression, and dereverberation |
 | [local-store](https://github.com/mudler/LocalAI) | Local-first vector database for embeddings (shipped in-tree) |
diff --git a/backend/README.md b/backend/README.md
index 0e92a0f03..a458da35f 100644
--- a/backend/README.md
+++ b/backend/README.md
@@ -56,6 +56,7 @@ The backend system provides language-specific Dockerfiles that handle the build
 - **stablediffusion-ggml**: Stable Diffusion in Go with GGML Cpp backend
 - **piper**: Text-to-speech synthesis Golang with C bindings using rhaspy/piper
 - **local-store**: Vector storage backend
+- **valkey-store**: Durable vector storage backend backed by Valkey Search (FT.*)
 
 #### C++ Backends (`cpp/`)
 - **llama-cpp**: Llama.cpp integration
diff --git a/backend/backend.proto b/backend/backend.proto
index 711e9465c..2935d1a62 100644
--- a/backend/backend.proto
+++ b/backend/backend.proto
@@ -16,6 +16,7 @@ service Backend {
   rpc Embedding(PredictOptions) returns (EmbeddingResult) {}
   rpc GenerateImage(GenerateImageRequest) returns (Result) {}
   rpc GenerateVideo(GenerateVideoRequest) returns (Result) {}
+  rpc Generate3D(Generate3DRequest) returns (Result) {}
   rpc AudioTranscription(TranscriptRequest) returns (TranscriptResult) {}
   rpc AudioTranscriptionStream(TranscriptRequest) returns (stream TranscriptStreamResponse) {}
   // AudioTranscriptionLive is the bidirectional live-microphone ASR RPC. The
@@ -658,6 +659,20 @@ message GenerateVideoRequest {
   string ModelIdentity = 15;
 }
 
+message Generate3DRequest {
+  string src = 1;          // Path to the staged conditioning image (3D generation is image-conditioned)
+  string dst = 2;          // Output path for the generated binary glTF (.glb) asset
+  int32 seed = 3;          // <=0 lets the backend pick a random seed
+  int32 step = 4;          // Flow sampling steps; <=0 uses the backend default
+  float cfg_scale = 5;     // Classifier-free guidance scale; <=0 uses the backend default
+  int32 texture_steps = 6; // Texture flow sampling steps; <=0 uses the backend default
+  string quality = 7;      // Mesh pipeline: ""|"auto"|"coarse"|"512"|"1024"
+  string background = 8;   // Conditioning-image background handling: ""|"auto"|"keep"|"black"|"white"
+  // Backend-specific per-request generation parameters. Values are strings
+  // and are validated/coerced by the selected backend.
+  map<string, string> params = 9;
+}
+
 message TTSRequest {
   string text = 1;
   string model = 2;
diff --git a/backend/go/trellis2cpp/.gitignore b/backend/go/trellis2cpp/.gitignore
new file mode 100644
index 000000000..b5cfa6aba
--- /dev/null
+++ b/backend/go/trellis2cpp/.gitignore
@@ -0,0 +1,6 @@
+package/
+sources/
+.cache/
+build-*/
+variants/
+trellis2cpp
diff --git a/backend/go/trellis2cpp/Makefile b/backend/go/trellis2cpp/Makefile
new file mode 100644
index 000000000..c20a991a5
--- /dev/null
+++ b/backend/go/trellis2cpp/Makefile
@@ -0,0 +1,132 @@
+CMAKE_ARGS?=
+BUILD_TYPE?=
+NATIVE?=false
+
+CURRENT_DIR=$(abspath ./)
+GOCMD?=go
+GO_TAGS?=
+JOBS?=$(shell nproc --ignore=1)
+
+# trellis2.cpp — C++/ggml port of Microsoft TRELLIS.2 (image -> 3D GLB).
+# The ggml submodule is pinned by trellis2cpp's .gitmodules and fetched via
+# --recursive. The commit pin lives here so bump_deps.yaml can update it.
+TRELLIS2CPP_REPO?=https://github.com/localai-org/trellis2cpp
+TRELLIS2CPP_VERSION?=73dfbe5dfc2cbefd0950853718086556c6d9b043
+
+# libtrellis2 + ggml as shared libraries; no example/test binaries.
+CMAKE_ARGS+=-DCMAKE_BUILD_TYPE=Release
+CMAKE_ARGS+=-DBUILD_SHARED_LIBS=ON
+CMAKE_ARGS+=-DTRELLIS2_BUILD_EXAMPLES=OFF
+CMAKE_ARGS+=-DTRELLIS2_BUILD_TESTS=OFF
+# Print remeshing is part of trellis2cpp's ABI, so upstream owns the tested
+# CGAL/Boost versions, checksums, fetch logic, and update automation. LocalAI
+# only opts into that dependency set and pins the trellis2cpp commit above.
+CMAKE_ARGS+=-DTRELLIS2_FETCH_PRINT_REMESH_DEPS=ON
+CMAKE_ARGS+=-DTRELLIS2_PRINT_REMESH_DEPS_DIR=$(CURRENT_DIR)/sources/print-remesh-deps
+
+ifeq ($(NATIVE),false)
+	CMAKE_ARGS+=-DGGML_NATIVE=OFF
+endif
+
+ifeq ($(BUILD_TYPE),cublas)
+	CMAKE_ARGS+=-DGGML_CUDA=ON
+else ifeq ($(BUILD_TYPE),vulkan)
+	CMAKE_ARGS+=-DGGML_VULKAN=ON
+else ifeq ($(BUILD_TYPE),hipblas)
+	ROCM_HOME ?= /opt/rocm
+	ROCM_PATH ?= /opt/rocm
+	export CXX=$(ROCM_HOME)/llvm/bin/clang++
+	export CC=$(ROCM_HOME)/llvm/bin/clang
+	AMDGPU_TARGETS?=gfx908,gfx90a,gfx942,gfx950,gfx1030,gfx1100,gfx1101,gfx1102,gfx1200,gfx1201
+	CMAKE_ARGS+=-DGGML_HIP=ON -DAMDGPU_TARGETS=$(AMDGPU_TARGETS)
+else ifeq ($(OS),Darwin)
+	ifneq ($(BUILD_TYPE),metal)
+		CMAKE_ARGS+=-DTRELLIS2_METAL=OFF -DGGML_METAL=OFF
+	else
+		# trellis2cpp turns on GGML_METAL(+EMBED_LIBRARY) itself when
+		# TRELLIS2_METAL is enabled on Apple platforms.
+		CMAKE_ARGS+=-DTRELLIS2_METAL=ON
+	endif
+	# Dependent libggml*.dylib resolve next to libtrellis2.dylib even
+	# without DYLD_LIBRARY_PATH being exported.
+	CMAKE_ARGS+=-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON -DCMAKE_INSTALL_RPATH=@loader_path
+endif
+
+ifeq ($(BUILD_TYPE),sycl_f16)
+	CMAKE_ARGS+=-DGGML_SYCL=ON \
+		-DCMAKE_C_COMPILER=icx \
+		-DCMAKE_CXX_COMPILER=icpx \
+		-DGGML_SYCL_F16=ON
+endif
+
+ifeq ($(BUILD_TYPE),sycl_f32)
+	CMAKE_ARGS+=-DGGML_SYCL=ON \
+		-DCMAKE_C_COMPILER=icx \
+		-DCMAKE_CXX_COMPILER=icpx
+endif
+
+sources/trellis2cpp:
+	git clone --recursive $(TRELLIS2CPP_REPO) sources/trellis2cpp && \
+	cd sources/trellis2cpp && \
+	git checkout $(TRELLIS2CPP_VERSION) && \
+	git submodule update --init --recursive --depth 1 --single-branch
+
+# Detect OS
+UNAME_S := $(shell uname -s)
+UNAME_M := $(shell uname -m)
+
+# The AVX variants are x86-only. ARM64 images use the portable fallback while
+# still enabling the selected GPU backend (Vulkan/CUDA) through CMAKE_ARGS.
+ifeq ($(UNAME_S),Linux)
+	ifneq (,$(filter x86_64 amd64,$(UNAME_M)))
+		VARIANTS = avx avx2 avx512 fallback
+	else
+		VARIANTS = fallback
+	endif
+else
+	# On non-Linux (e.g., Darwin), build only the fallback variant
+	VARIANTS = fallback
+endif
+VARIANT_TARGETS = $(foreach v,$(VARIANTS),variants/$(v)/.built)
+
+VARIANT_FLAGS_avx      = -DGGML_AVX=on  -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off
+VARIANT_FLAGS_avx2     = -DGGML_AVX=on  -DGGML_AVX2=on  -DGGML_AVX512=off -DGGML_FMA=on  -DGGML_F16C=on  -DGGML_BMI2=on
+VARIANT_FLAGS_avx512   = -DGGML_AVX=on  -DGGML_AVX2=on  -DGGML_AVX512=on  -DGGML_FMA=on  -DGGML_F16C=on  -DGGML_BMI2=on
+VARIANT_FLAGS_fallback = -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off
+
+# libtrellis2 links libggml/libggml-base/libggml-cpu (+ the GPU backend) by
+# soname, and those sonames collide across SIMD variants — so each variant
+# lives in its own directory and run.sh selects one via LD_LIBRARY_PATH,
+# unlike stablediffusion-ggml's flat renamed-.so scheme.
+variants/%/.built: sources/trellis2cpp
+	rm -rf build-$* variants/$*
+	mkdir -p build-$* variants/$*
+	cd build-$* && cmake ../sources/trellis2cpp $(CMAKE_ARGS) $(VARIANT_FLAGS_$*) && \
+		cmake --build . --config Release -j$(JOBS)
+	@for f in build-$*/libtrellis2.so build-$*/libtrellis2.dylib; do \
+		if [ -e $$f ]; then cp -a $$f variants/$*/; fi; done
+	find build-$*/ggml \( -name 'libggml*.so*' -o -name 'libggml*.dylib' \) -exec cp -a {} variants/$*/ \;
+	rm -rf build-$*
+	touch $@
+
+trellis2cpp: main.go trellis2.go $(VARIANT_TARGETS)
+	CGO_ENABLED=0 $(GOCMD) build -tags "$(GO_TAGS)" -o trellis2cpp ./
+
+package: trellis2cpp
+	bash package.sh
+
+build: package
+
+clean: purge
+	rm -rf variants trellis2cpp package sources
+
+purge:
+	rm -rf build-*
+
+# Weight-free by construction: pure-Go unit tests over model-path resolution,
+# validation, and request-parameter mapping. The multi-GB GGUF weights are
+# never downloaded in CI; end-to-end generation is exercised manually.
+test:
+	$(GOCMD) test -v ./...
+
+all: trellis2cpp package
diff --git a/backend/go/trellis2cpp/glb_mesh.go b/backend/go/trellis2cpp/glb_mesh.go
new file mode 100644
index 000000000..144c9243a
--- /dev/null
+++ b/backend/go/trellis2cpp/glb_mesh.go
@@ -0,0 +1,252 @@
+package main
+
+import (
+	"encoding/binary"
+	"encoding/json"
+	"fmt"
+	"math"
+)
+
+const (
+	glbMagic     = 0x46546c67
+	glbJSONChunk = 0x4e4f534a
+	glbBINChunk  = 0x004e4942
+)
+
+type glbAccessor struct {
+	BufferView    int    `json:"bufferView"`
+	ByteOffset    int    `json:"byteOffset"`
+	ComponentType int    `json:"componentType"`
+	Count         int    `json:"count"`
+	Type          string `json:"type"`
+	Normalized    bool   `json:"normalized"`
+}
+
+type glbBufferView struct {
+	Buffer     int `json:"buffer"`
+	ByteOffset int `json:"byteOffset"`
+	ByteLength int `json:"byteLength"`
+	ByteStride int `json:"byteStride"`
+}
+
+type glbPrimitive struct {
+	Attributes map[string]int `json:"attributes"`
+	Indices    *int           `json:"indices"`
+}
+
+type glbDocument struct {
+	Accessors   []glbAccessor   `json:"accessors"`
+	BufferViews []glbBufferView `json:"bufferViews"`
+	Meshes      []struct {
+		Primitives []glbPrimitive `json:"primitives"`
+	} `json:"meshes"`
+}
+
+type glbVertexMesh struct {
+	verts []float32
+	tris  []int32
+	pbr   []float32
+}
+
+func glbLayout(componentType int, accessorType string) (componentBytes, components int, err error) {
+	switch componentType {
+	case 5121:
+		componentBytes = 1
+	case 5123:
+		componentBytes = 2
+	case 5125, 5126:
+		componentBytes = 4
+	default:
+		return 0, 0, fmt.Errorf("unsupported GLB component type %d", componentType)
+	}
+	switch accessorType {
+	case "SCALAR":
+		components = 1
+	case "VEC2":
+		components = 2
+	case "VEC3":
+		components = 3
+	case "VEC4":
+		components = 4
+	default:
+		return 0, 0, fmt.Errorf("unsupported GLB accessor type %q", accessorType)
+	}
+	return componentBytes, components, nil
+}
+
+func glbAccessorData(doc *glbDocument, binChunk []byte, index int) (glbAccessor, []byte, error) {
+	if index < 0 || index >= len(doc.Accessors) {
+		return glbAccessor{}, nil, fmt.Errorf("missing GLB accessor %d", index)
+	}
+	a := doc.Accessors[index]
+	if a.BufferView < 0 || a.BufferView >= len(doc.BufferViews) {
+		return glbAccessor{}, nil, fmt.Errorf("missing GLB buffer view %d", a.BufferView)
+	}
+	v := doc.BufferViews[a.BufferView]
+	if v.Buffer != 0 || v.ByteStride != 0 {
+		return glbAccessor{}, nil, fmt.Errorf("interleaved or external GLB buffers are unsupported")
+	}
+	componentBytes, components, err := glbLayout(a.ComponentType, a.Type)
+	if err != nil {
+		return glbAccessor{}, nil, err
+	}
+	if a.Count <= 0 || a.Count > math.MaxInt/(componentBytes*components) {
+		return glbAccessor{}, nil, fmt.Errorf("invalid GLB accessor count %d", a.Count)
+	}
+	length := a.Count * componentBytes * components
+	if v.ByteOffset < 0 || v.ByteLength < 0 || a.ByteOffset < 0 ||
+		a.ByteOffset > v.ByteLength || length > v.ByteLength-a.ByteOffset ||
+		length > len(binChunk) || v.ByteOffset > len(binChunk)-length-a.ByteOffset {
+		return glbAccessor{}, nil, fmt.Errorf("GLB accessor %d is outside the BIN chunk", index)
+	}
+	start := v.ByteOffset + a.ByteOffset
+	return a, binChunk[start : start+length], nil
+}
+
+// parseVertexGLB reads the dense vertex-PBR form emitted by trellis2.cpp. GLB
+// coordinates and linear COLOR_0 values are converted back to the native
+// trellis coordinate/material convention before CGAL remeshing and rebaking.
+func parseVertexGLB(data []byte) (*glbVertexMesh, error) {
+	if len(data) < 20 || binary.LittleEndian.Uint32(data[0:4]) != glbMagic {
+		return nil, fmt.Errorf("input is not a GLB file")
+	}
+	if binary.LittleEndian.Uint32(data[4:8]) != 2 {
+		return nil, fmt.Errorf("unsupported GLB version")
+	}
+	total := int(binary.LittleEndian.Uint32(data[8:12]))
+	if total != len(data) {
+		return nil, fmt.Errorf("invalid GLB length")
+	}
+
+	var jsonChunk, binChunk []byte
+	for offset := 12; offset <= len(data)-8; {
+		length := int(binary.LittleEndian.Uint32(data[offset : offset+4]))
+		chunkType := binary.LittleEndian.Uint32(data[offset+4 : offset+8])
+		start := offset + 8
+		if length < 0 || start > len(data)-length {
+			return nil, fmt.Errorf("invalid GLB chunk length")
+		}
+		switch chunkType {
+		case glbJSONChunk:
+			if jsonChunk == nil {
+				jsonChunk = data[start : start+length]
+			}
+		case glbBINChunk:
+			if binChunk == nil {
+				binChunk = data[start : start+length]
+			}
+		}
+		offset = start + length
+	}
+	if jsonChunk == nil || binChunk == nil {
+		return nil, fmt.Errorf("GLB must contain JSON and BIN chunks")
+	}
+
+	var doc glbDocument
+	if err := json.Unmarshal(jsonChunk, &doc); err != nil {
+		return nil, fmt.Errorf("parsing GLB JSON: %w", err)
+	}
+	if len(doc.Meshes) != 1 || len(doc.Meshes[0].Primitives) != 1 {
+		return nil, fmt.Errorf("GLB must contain one mesh primitive")
+	}
+	primitive := doc.Meshes[0].Primitives[0]
+	positionIndex, ok := primitive.Attributes["POSITION"]
+	if !ok {
+		return nil, fmt.Errorf("GLB mesh has no POSITION attribute")
+	}
+	position, positionData, err := glbAccessorData(&doc, binChunk, positionIndex)
+	if err != nil {
+		return nil, err
+	}
+	if position.ComponentType != 5126 || position.Type != "VEC3" {
+		return nil, fmt.Errorf("GLB POSITION must be float32 VEC3")
+	}
+
+	mesh := &glbVertexMesh{verts: make([]float32, position.Count*3)}
+	for i := 0; i < position.Count; i++ {
+		x := math.Float32frombits(binary.LittleEndian.Uint32(positionData[(i*3)*4:]))
+		y := math.Float32frombits(binary.LittleEndian.Uint32(positionData[(i*3+1)*4:]))
+		z := math.Float32frombits(binary.LittleEndian.Uint32(positionData[(i*3+2)*4:]))
+		if math.IsNaN(float64(x)) || math.IsNaN(float64(y)) || math.IsNaN(float64(z)) ||
+			math.IsInf(float64(x), 0) || math.IsInf(float64(y), 0) || math.IsInf(float64(z), 0) {
+			return nil, fmt.Errorf("GLB POSITION contains a non-finite value")
+		}
+		mesh.verts[i*3] = x
+		mesh.verts[i*3+1] = -z
+		mesh.verts[i*3+2] = y
+	}
+
+	if primitive.Indices == nil {
+		if position.Count%3 != 0 {
+			return nil, fmt.Errorf("unindexed GLB vertex count is not divisible by three")
+		}
+		mesh.tris = make([]int32, position.Count)
+		for i := range mesh.tris {
+			mesh.tris[i] = int32(i)
+		}
+	} else {
+		indices, indexData, err := glbAccessorData(&doc, binChunk, *primitive.Indices)
+		if err != nil {
+			return nil, err
+		}
+		if indices.Type != "SCALAR" || indices.Count%3 != 0 || (indices.ComponentType != 5123 && indices.ComponentType != 5125) {
+			return nil, fmt.Errorf("GLB indices must be uint16/uint32 triangles")
+		}
+		mesh.tris = make([]int32, indices.Count)
+		for i := range mesh.tris {
+			var value uint32
+			if indices.ComponentType == 5123 {
+				value = uint32(binary.LittleEndian.Uint16(indexData[i*2:]))
+			} else {
+				value = binary.LittleEndian.Uint32(indexData[i*4:])
+			}
+			if value >= uint32(position.Count) || value > math.MaxInt32 {
+				return nil, fmt.Errorf("GLB index %d is outside the vertex buffer", value)
+			}
+			mesh.tris[i] = int32(value)
+		}
+	}
+
+	colorIndex, hasColor := primitive.Attributes["COLOR_0"]
+	if !hasColor {
+		return mesh, nil
+	}
+	color, colorData, err := glbAccessorData(&doc, binChunk, colorIndex)
+	if err != nil {
+		return nil, err
+	}
+	if color.ComponentType != 5123 || color.Type != "VEC4" || !color.Normalized || color.Count != position.Count {
+		return nil, fmt.Errorf("GLB COLOR_0 must be normalized uint16 VEC4 aligned with POSITION")
+	}
+	metalRoughIndex, hasMetalRough := primitive.Attributes["_METALLIC_ROUGHNESS"]
+	var metalRoughData []byte
+	if hasMetalRough {
+		metalRough, data, err := glbAccessorData(&doc, binChunk, metalRoughIndex)
+		if err != nil {
+			return nil, err
+		}
+		if metalRough.ComponentType != 5121 || metalRough.Type != "VEC2" || !metalRough.Normalized || metalRough.Count != position.Count {
+			return nil, fmt.Errorf("GLB _METALLIC_ROUGHNESS must be normalized uint8 VEC2 aligned with POSITION")
+		}
+		metalRoughData = data
+	}
+
+	mesh.pbr = make([]float32, position.Count*6)
+	for i := 0; i < position.Count; i++ {
+		for channel := 0; channel < 3; channel++ {
+			linear := float32(binary.LittleEndian.Uint16(colorData[(i*4+channel)*2:])) / 65535
+			if linear <= 0.0031308 {
+				mesh.pbr[i*6+channel] = linear * 12.92
+			} else {
+				mesh.pbr[i*6+channel] = 1.055*float32(math.Pow(float64(linear), 1.0/2.4)) - 0.055
+			}
+		}
+		mesh.pbr[i*6+5] = float32(binary.LittleEndian.Uint16(colorData[(i*4+3)*2:])) / 65535
+		mesh.pbr[i*6+4] = 0.6
+		if hasMetalRough {
+			mesh.pbr[i*6+3] = float32(metalRoughData[i*2]) / 255
+			mesh.pbr[i*6+4] = float32(metalRoughData[i*2+1]) / 255
+		}
+	}
+	return mesh, nil
+}
diff --git a/backend/go/trellis2cpp/glb_mesh_test.go b/backend/go/trellis2cpp/glb_mesh_test.go
new file mode 100644
index 000000000..8ba4f9264
--- /dev/null
+++ b/backend/go/trellis2cpp/glb_mesh_test.go
@@ -0,0 +1,74 @@
+package main
+
+import (
+	"encoding/binary"
+	"fmt"
+	"math"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+func tinyVertexGLB() []byte {
+	bin := make([]byte, 80)
+	positions := []float32{1, 2, 3, 4, 5, 6, 7, 8, 9}
+	for i, value := range positions {
+		binary.LittleEndian.PutUint32(bin[i*4:], math.Float32bits(value))
+	}
+	colors := []uint16{
+		65535, 0, 0, 65535,
+		0, 65535, 0, 32768,
+		0, 0, 65535, 65535,
+	}
+	for i, value := range colors {
+		binary.LittleEndian.PutUint16(bin[36+i*2:], value)
+	}
+	copy(bin[60:], []byte{0, 153, 64, 128, 255, 32})
+	for i, value := range []uint32{0, 1, 2} {
+		binary.LittleEndian.PutUint32(bin[68+i*4:], value)
+	}
+
+	jsonChunk := []byte(fmt.Sprintf(`{"asset":{"version":"2.0"},"meshes":[{"primitives":[{"attributes":{"POSITION":0,"COLOR_0":1,"_METALLIC_ROUGHNESS":2},"indices":3}]}],"accessors":[{"bufferView":0,"componentType":5126,"count":3,"type":"VEC3"},{"bufferView":1,"componentType":5123,"normalized":true,"count":3,"type":"VEC4"},{"bufferView":2,"componentType":5121,"normalized":true,"count":3,"type":"VEC2"},{"bufferView":3,"componentType":5125,"count":3,"type":"SCALAR"}],"bufferViews":[{"buffer":0,"byteOffset":0,"byteLength":36},{"buffer":0,"byteOffset":36,"byteLength":24},{"buffer":0,"byteOffset":60,"byteLength":6},{"buffer":0,"byteOffset":68,"byteLength":12}],"buffers":[{"byteLength":%d}]}`, len(bin)))
+	for len(jsonChunk)%4 != 0 {
+		jsonChunk = append(jsonChunk, ' ')
+	}
+	total := 12 + 8 + len(jsonChunk) + 8 + len(bin)
+	glb := make([]byte, total)
+	binary.LittleEndian.PutUint32(glb[0:], glbMagic)
+	binary.LittleEndian.PutUint32(glb[4:], 2)
+	binary.LittleEndian.PutUint32(glb[8:], uint32(total))
+	binary.LittleEndian.PutUint32(glb[12:], uint32(len(jsonChunk)))
+	binary.LittleEndian.PutUint32(glb[16:], glbJSONChunk)
+	copy(glb[20:], jsonChunk)
+	binHeader := 20 + len(jsonChunk)
+	binary.LittleEndian.PutUint32(glb[binHeader:], uint32(len(bin)))
+	binary.LittleEndian.PutUint32(glb[binHeader+4:], glbBINChunk)
+	copy(glb[binHeader+8:], bin)
+	return glb
+}
+
+var _ = Describe("vertex GLB parsing for print remeshing", func() {
+	It("restores trellis coordinates, topology, and PBR values", func() {
+		mesh, err := parseVertexGLB(tinyVertexGLB())
+		Expect(err).NotTo(HaveOccurred())
+		Expect(mesh.verts).To(Equal([]float32{1, -3, 2, 4, -6, 5, 7, -9, 8}))
+		Expect(mesh.tris).To(Equal([]int32{0, 1, 2}))
+		Expect(mesh.pbr).To(HaveLen(18))
+		Expect(mesh.pbr[0]).To(BeNumerically("~", 1, 1e-5))
+		Expect(mesh.pbr[3]).To(BeNumerically("~", 0, 1e-5))
+		Expect(mesh.pbr[4]).To(BeNumerically("~", 0.6, 0.01))
+		Expect(mesh.pbr[11]).To(BeNumerically("~", 32768.0/65535.0, 1e-5))
+	})
+
+	It("rejects indices outside the source vertex buffer", func() {
+		glb := tinyVertexGLB()
+		binary.LittleEndian.PutUint32(glb[len(glb)-12:], 3)
+		_, err := parseVertexGLB(glb)
+		Expect(err).To(MatchError(ContainSubstring("outside the vertex buffer")))
+	})
+
+	It("rejects non-GLB input", func() {
+		_, err := parseVertexGLB([]byte("not a mesh"))
+		Expect(err).To(MatchError("input is not a GLB file"))
+	})
+})
diff --git a/backend/go/trellis2cpp/main.go b/backend/go/trellis2cpp/main.go
new file mode 100644
index 000000000..14938bdd6
--- /dev/null
+++ b/backend/go/trellis2cpp/main.go
@@ -0,0 +1,50 @@
+package main
+
+import (
+	"flag"
+	"fmt"
+	"os"
+	"runtime"
+
+	"github.com/ebitengine/purego"
+	grpc "github.com/mudler/LocalAI/pkg/grpc"
+)
+
+var (
+	addr = flag.String("addr", "localhost:50051", "the address to connect to")
+)
+
+func registerLibFuncs(lib uintptr) {
+	registerLibFuncsWith(func(fptr any, name string) {
+		purego.RegisterLibFunc(fptr, lib, name)
+	})
+}
+
+func main() {
+	// run.sh selects the CPU-variant directory and points TRELLIS2_LIBRARY at it.
+	libName := os.Getenv("TRELLIS2_LIBRARY")
+	if libName == "" {
+		if runtime.GOOS == "darwin" {
+			libName = "./variants/fallback/libtrellis2.dylib"
+		} else {
+			libName = "./variants/fallback/libtrellis2.so"
+		}
+	}
+
+	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)
+	if err != nil {
+		panic(err)
+	}
+
+	registerLibFuncs(lib)
+
+	if got := t2AbiVersion(); got != abiVersion {
+		panic(fmt.Sprintf("trellis2 ABI mismatch: library reports %d, backend built for %d", got, abiVersion))
+	}
+
+	flag.Parse()
+
+	if err := grpc.StartServer(*addr, &Trellis2{}); err != nil {
+		panic(err)
+	}
+}
diff --git a/backend/go/trellis2cpp/package.sh b/backend/go/trellis2cpp/package.sh
new file mode 100755
index 000000000..20d7dedc2
--- /dev/null
+++ b/backend/go/trellis2cpp/package.sh
@@ -0,0 +1,63 @@
+#!/bin/bash
+
+# Script to copy the appropriate libraries based on architecture
+# This script is used in the final stage of the Dockerfile
+
+set -e
+
+CURDIR=$(dirname "$(realpath $0)")
+REPO_ROOT="${CURDIR}/../../.."
+
+# Create lib directory
+mkdir -p $CURDIR/package/lib
+
+# Each CPU variant keeps its libtrellis2 + libggml* set in its own directory
+# (their sonames collide across variants); run.sh selects one at startup.
+cp -a $CURDIR/variants $CURDIR/package/
+cp -avf $CURDIR/trellis2cpp $CURDIR/package/
+cp -fv $CURDIR/run.sh $CURDIR/package/
+
+# Detect architecture and copy appropriate libraries
+if [ -f "/lib64/ld-linux-x86-64.so.2" ]; then
+    # x86_64 architecture
+    echo "Detected x86_64 architecture, copying x86_64 libraries..."
+    cp -arfLv /lib64/ld-linux-x86-64.so.2 $CURDIR/package/lib/ld.so
+    cp -arfLv /lib/x86_64-linux-gnu/libc.so.6 $CURDIR/package/lib/libc.so.6
+    cp -arfLv /lib/x86_64-linux-gnu/libgcc_s.so.1 $CURDIR/package/lib/libgcc_s.so.1
+    cp -arfLv /lib/x86_64-linux-gnu/libstdc++.so.6 $CURDIR/package/lib/libstdc++.so.6
+    cp -arfLv /lib/x86_64-linux-gnu/libm.so.6 $CURDIR/package/lib/libm.so.6
+    cp -arfLv /lib/x86_64-linux-gnu/libgomp.so.1 $CURDIR/package/lib/libgomp.so.1
+    cp -arfLv /lib/x86_64-linux-gnu/libdl.so.2 $CURDIR/package/lib/libdl.so.2
+    cp -arfLv /lib/x86_64-linux-gnu/librt.so.1 $CURDIR/package/lib/librt.so.1
+    cp -arfLv /lib/x86_64-linux-gnu/libpthread.so.0 $CURDIR/package/lib/libpthread.so.0
+elif [ -f "/lib/ld-linux-aarch64.so.1" ]; then
+    # ARM64 architecture
+    echo "Detected ARM64 architecture, copying ARM64 libraries..."
+    cp -arfLv /lib/ld-linux-aarch64.so.1 $CURDIR/package/lib/ld.so
+    cp -arfLv /lib/aarch64-linux-gnu/libc.so.6 $CURDIR/package/lib/libc.so.6
+    cp -arfLv /lib/aarch64-linux-gnu/libgcc_s.so.1 $CURDIR/package/lib/libgcc_s.so.1
+    cp -arfLv /lib/aarch64-linux-gnu/libstdc++.so.6 $CURDIR/package/lib/libstdc++.so.6
+    cp -arfLv /lib/aarch64-linux-gnu/libm.so.6 $CURDIR/package/lib/libm.so.6
+    cp -arfLv /lib/aarch64-linux-gnu/libgomp.so.1 $CURDIR/package/lib/libgomp.so.1
+    cp -arfLv /lib/aarch64-linux-gnu/libdl.so.2 $CURDIR/package/lib/libdl.so.2
+    cp -arfLv /lib/aarch64-linux-gnu/librt.so.1 $CURDIR/package/lib/librt.so.1
+    cp -arfLv /lib/aarch64-linux-gnu/libpthread.so.0 $CURDIR/package/lib/libpthread.so.0
+elif [ $(uname -s) = "Darwin" ]; then
+    echo "Detected Darwin"
+else
+    echo "Error: Could not detect architecture"
+    exit 1
+fi
+
+# Package GPU libraries based on BUILD_TYPE
+# The GPU library packaging script will detect BUILD_TYPE and copy appropriate GPU libraries
+GPU_LIB_SCRIPT="${REPO_ROOT}/scripts/build/package-gpu-libs.sh"
+if [ -f "$GPU_LIB_SCRIPT" ]; then
+    echo "Packaging GPU libraries for BUILD_TYPE=${BUILD_TYPE:-cpu}..."
+    source "$GPU_LIB_SCRIPT" "$CURDIR/package/lib"
+    package_gpu_libs
+fi
+
+echo "Packaging completed successfully"
+ls -liah $CURDIR/package/
+ls -liah $CURDIR/package/lib/
diff --git a/backend/go/trellis2cpp/run.sh b/backend/go/trellis2cpp/run.sh
new file mode 100755
index 000000000..78f5ec362
--- /dev/null
+++ b/backend/go/trellis2cpp/run.sh
@@ -0,0 +1,61 @@
+#!/bin/bash
+set -ex
+
+# Get the absolute current dir where the script is located
+CURDIR=$(dirname "$(realpath "$0")")
+
+cd /
+
+echo "CPU info:"
+if [ "$(uname)" != "Darwin" ]; then
+	grep -e "model\sname" /proc/cpuinfo | head -1
+	grep -e "flags" /proc/cpuinfo | head -1
+fi
+
+# Each variant directory bundles libtrellis2 plus its libggml* set (the ggml
+# sonames collide across SIMD variants, so they can't share one directory).
+VARIANT=fallback
+
+if [ "$(uname)" = "Darwin" ]; then
+	LIBRARY="$CURDIR/variants/$VARIANT/libtrellis2.dylib"
+	if [ ! -e "$LIBRARY" ]; then
+		LIBRARY="$CURDIR/variants/$VARIANT/libtrellis2.so"
+	fi
+	export DYLD_LIBRARY_PATH="$CURDIR/variants/$VARIANT:$CURDIR/lib:$DYLD_LIBRARY_PATH"
+else
+	if grep -q -e "\savx\s" /proc/cpuinfo ; then
+		echo "CPU:    AVX    found OK"
+		if [ -d "$CURDIR/variants/avx" ]; then
+			VARIANT=avx
+		fi
+	fi
+
+	if grep -q -e "\savx2\s" /proc/cpuinfo ; then
+		echo "CPU:    AVX2   found OK"
+		if [ -d "$CURDIR/variants/avx2" ]; then
+			VARIANT=avx2
+		fi
+	fi
+
+	if grep -q -e "\savx512f\s" /proc/cpuinfo ; then
+		echo "CPU:    AVX512F found OK"
+		if [ -d "$CURDIR/variants/avx512" ]; then
+			VARIANT=avx512
+		fi
+	fi
+
+	LIBRARY="$CURDIR/variants/$VARIANT/libtrellis2.so"
+	export LD_LIBRARY_PATH="$CURDIR/variants/$VARIANT:$CURDIR/lib:$LD_LIBRARY_PATH"
+fi
+
+export TRELLIS2_LIBRARY=$LIBRARY
+
+# If there is a lib/ld.so, use it
+if [ -f "$CURDIR"/lib/ld.so ]; then
+	echo "Using lib/ld.so"
+	echo "Using library: $LIBRARY"
+	exec "$CURDIR"/lib/ld.so "$CURDIR"/trellis2cpp "$@"
+fi
+
+echo "Using library: $LIBRARY"
+exec "$CURDIR"/trellis2cpp "$@"
diff --git a/backend/go/trellis2cpp/trellis2.go b/backend/go/trellis2cpp/trellis2.go
new file mode 100644
index 000000000..addbf1bb0
--- /dev/null
+++ b/backend/go/trellis2cpp/trellis2.go
@@ -0,0 +1,511 @@
+package main
+
+// trellis2.go — purego bindings to libtrellis2's flat C ABI (trellis2_capi.h)
+// plus the LocalAI backend implementation. Adapted from the upstream demo
+// server's engine.go; the t2_abi_version binding guards against header/library
+// drift.
+
+import (
+	"fmt"
+	"math/rand"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+	"sync"
+	"unsafe"
+
+	"github.com/mudler/LocalAI/pkg/grpc/base"
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+	"github.com/mudler/LocalAI/pkg/utils"
+)
+
+const abiVersion = 11
+
+// Pipeline types (enum t2_pipeline_type) and background modes
+// (enum t2_background_mode).
+const (
+	pipeAuto   = 0
+	pipeCoarse = 1
+	pipe512    = 2
+	pipe1024   = 3
+
+	backgroundAuto  = 0
+	backgroundKeep  = 1
+	backgroundBlack = 2
+	backgroundWhite = 3
+)
+
+// The bindings use typed pointers (*float32/*int32/*byte) rather than uintptr
+// for C-owned buffers so no uintptr->unsafe.Pointer conversions are needed;
+// only the opaque t2_pipeline / t2_mesh_result handles stay uintptr.
+var (
+	t2AbiVersion   func() int32
+	t2PipelineLoad func(dino, ssFlow, ssDec, slatFlow, slatFlowHR, shapeDec,
+		shapeEnc, texDec, texFlow, texFlowHR string,
+		flags int32, err *byte, errLen int32) uintptr
+	t2PipelineFree    func(p uintptr)
+	t2PipelineBackend func(p uintptr) string
+	t2PipelineCaps    func(p uintptr) int32
+	t2Generate        func(p uintptr, img *byte, imgLen int32,
+		pipelineType, backgroundMode int32, seed uint64, steps int32,
+		guidance float32, textureSteps int32,
+		progress, user, preview, previewUser uintptr,
+		err *byte, errLen int32) uintptr
+	t2MeshNVerts func(r uintptr) int32
+	t2MeshNTris  func(r uintptr) int32
+	t2MeshVerts  func(r uintptr) *float32
+	t2MeshTris   func(r uintptr) *int32
+	t2MeshHasPBR func(r uintptr) int32
+	t2MeshPBR    func(r uintptr) *float32
+	t2MeshFree   func(r uintptr)
+	t2BakeGLB    func(verts *float32, nv int32, tris *int32, nt int32,
+		pbr *float32, texSize, componentFilter int32,
+		outLen *int32, err *byte, errLen int32) *byte
+	// CGAL Alpha Wrap print remeshing — availability is fixed at library build
+	// time, so gate every use on t2_print_remesh_available.
+	t2PrintRemeshAvailable func() int32
+	t2PreparePrintMesh     func(verts *float32, nv int32, tris *int32, nt int32,
+		pbr *float32, componentFilter int32, alphaRatio, offsetRatio float32,
+		err *byte, errLen int32) uintptr
+	t2BakeProjectedGLB func(targetVerts *float32, targetNV int32,
+		targetTris *int32, targetNT int32,
+		sourceVerts *float32, sourceNV int32,
+		sourceTris *int32, sourceNT int32,
+		sourcePBR *float32, texSize, sourceComponentFilter int32,
+		outLen *int32, err *byte, errLen int32) *byte
+	t2FreeBuffer func(buf *byte)
+)
+
+type libFunc struct {
+	funcPtr any
+	name    string
+}
+
+func registerLibFuncsWith(register func(fptr any, name string)) {
+	for _, lf := range []libFunc{
+		{&t2AbiVersion, "t2_abi_version"},
+		{&t2PipelineLoad, "t2_pipeline_load"},
+		{&t2PipelineFree, "t2_pipeline_free"},
+		{&t2PipelineBackend, "t2_pipeline_backend"},
+		{&t2PipelineCaps, "t2_pipeline_caps"},
+		{&t2Generate, "t2_generate"},
+		{&t2MeshNVerts, "t2_mesh_n_verts"},
+		{&t2MeshNTris, "t2_mesh_n_tris"},
+		{&t2MeshVerts, "t2_mesh_verts"},
+		{&t2MeshTris, "t2_mesh_tris"},
+		{&t2MeshHasPBR, "t2_mesh_has_pbr"},
+		{&t2MeshPBR, "t2_mesh_pbr"},
+		{&t2MeshFree, "t2_mesh_free"},
+		{&t2BakeGLB, "t2_bake_glb"},
+		{&t2PrintRemeshAvailable, "t2_print_remesh_available"},
+		{&t2PreparePrintMesh, "t2_prepare_print_mesh"},
+		{&t2BakeProjectedGLB, "t2_bake_projected_glb"},
+		{&t2FreeBuffer, "t2_free_buffer"},
+	} {
+		register(lf.funcPtr, lf.name)
+	}
+}
+
+// modelSet holds the resolved path for every pipeline role; optional roles are
+// "" when disabled (the C side treats NULL/"" as "omit").
+type modelSet struct {
+	dino, ssFlow, ssDec             string
+	slatFlow, slatFlow1024          string
+	shapeDec                        string
+	shapeEnc, texDec                string
+	texSlatFlow512, texSlatFlow1024 string
+}
+
+// role → (option key, default filename) in t2_pipeline_load argument order.
+// The option keys follow the sd-ggml `*_path` convention; the default
+// filenames are the ones the upstream converters emit and the demo server
+// looks up, so a gallery install needs no options at all.
+type modelRole struct {
+	key      string
+	filename string
+	required bool
+	assign   func(*modelSet, string)
+}
+
+var modelRoles = []modelRole{
+	{"dino_path", "dino_f16.gguf", true, func(s *modelSet, p string) { s.dino = p }},
+	{"ss_flow_path", "ss_flow_f16.gguf", true, func(s *modelSet, p string) { s.ssFlow = p }},
+	{"ss_dec_path", "ss_dec_f16.gguf", true, func(s *modelSet, p string) { s.ssDec = p }},
+	{"slat_flow_path", "slat_flow_f16.gguf", false, func(s *modelSet, p string) { s.slatFlow = p }},
+	{"slat_flow_1024_path", "slat_flow_1024_f16.gguf", false, func(s *modelSet, p string) { s.slatFlow1024 = p }},
+	{"shape_dec_path", "shape_dec_f16.gguf", false, func(s *modelSet, p string) { s.shapeDec = p }},
+	{"shape_enc_path", "shape_enc_f16.gguf", false, func(s *modelSet, p string) { s.shapeEnc = p }},
+	{"tex_dec_path", "tex_dec_f16.gguf", false, func(s *modelSet, p string) { s.texDec = p }},
+	{"tex_slat_flow_512_path", "tex_slat_flow_512_f16.gguf", false, func(s *modelSet, p string) { s.texSlatFlow512 = p }},
+	{"tex_slat_flow_1024_path", "tex_slat_flow_1024_f16.gguf", false, func(s *modelSet, p string) { s.texSlatFlow1024 = p }},
+}
+
+// resolveModels maps LocalAI's model file + options onto the ten pipeline
+// roles. The model file only anchors the GGUF directory; each role resolves
+// to an explicit `<role>_path` option when given, else to its default
+// filename in that directory. Missing required files refuse the load (a
+// backend must not capture arbitrary GGUFs — see issue #9287); missing
+// optional files degrade capabilities the same way the upstream demo does.
+func resolveModels(modelFile, modelPath string, options []string) (modelSet, error) {
+	base := modelFile
+	if !filepath.IsAbs(base) {
+		base = filepath.Join(modelPath, base)
+	}
+	ggufDir := filepath.Dir(base)
+
+	overrides := map[string]string{}
+	for _, op := range options {
+		key, value, found := strings.Cut(op, ":")
+		if !found || !strings.HasSuffix(key, "_path") {
+			continue
+		}
+		if !filepath.IsAbs(value) {
+			value = filepath.Join(modelPath, value)
+			if err := utils.VerifyPath(value, modelPath); err != nil {
+				return modelSet{}, fmt.Errorf("option %s: %w", key, err)
+			}
+		}
+		overrides[key] = value
+	}
+
+	var set modelSet
+	var missingRequired []string
+	for _, role := range modelRoles {
+		path, explicit := overrides[role.key]
+		if !explicit {
+			path = filepath.Join(ggufDir, role.filename)
+		}
+		if _, err := os.Stat(path); err != nil {
+			if explicit {
+				return modelSet{}, fmt.Errorf("option %s points at a missing file: %s", role.key, path)
+			}
+			if role.required {
+				missingRequired = append(missingRequired, role.filename)
+			}
+			path = ""
+		}
+		role.assign(&set, path)
+	}
+	if len(missingRequired) > 0 {
+		return modelSet{}, fmt.Errorf("not a trellis2 model set: missing required %s in %s", strings.Join(missingRequired, ", "), ggufDir)
+	}
+
+	// Degradation mirrors the upstream demo: the 512 pair enables everything
+	// finer than coarse; texturing needs its three-model set; a textured 1024
+	// cascade additionally needs the HR texture flow.
+	if set.slatFlow == "" || set.shapeDec == "" {
+		set.slatFlow, set.shapeDec = "", ""
+		set.slatFlow1024 = ""
+		set.shapeEnc, set.texDec, set.texSlatFlow512, set.texSlatFlow1024 = "", "", "", ""
+		return set, nil
+	}
+	if set.shapeEnc == "" || set.texDec == "" || set.texSlatFlow512 == "" {
+		set.shapeEnc, set.texDec, set.texSlatFlow512, set.texSlatFlow1024 = "", "", "", ""
+	} else if set.texSlatFlow1024 == "" {
+		set.slatFlow1024 = ""
+	}
+	return set, nil
+}
+
+func pipelineForQuality(quality string) int32 {
+	switch quality {
+	case "coarse":
+		return pipeCoarse
+	case "512":
+		return pipe512
+	case "1024":
+		return pipe1024
+	default:
+		return pipeAuto
+	}
+}
+
+func backgroundForMode(background string) int32 {
+	switch background {
+	case "keep":
+		return backgroundKeep
+	case "black":
+		return backgroundBlack
+	case "white":
+		return backgroundWhite
+	default:
+		return backgroundAuto
+	}
+}
+
+func componentFilterFor(components string) int32 {
+	switch components {
+	case "tiny":
+		return 0 // remove only tiny islands
+	case "largest":
+		return 1 // keep the largest connected component
+	default:
+		return 2 // preserve every connected component (demo default)
+	}
+}
+
+func atoiOr(s string, fallback int32) int32 {
+	if s == "" {
+		return fallback
+	}
+	n, err := strconv.Atoi(s)
+	if err != nil {
+		return fallback
+	}
+	return int32(n)
+}
+
+func boolParam(s string) bool {
+	return s == "1" || strings.EqualFold(s, "true")
+}
+
+// ratioOr parses a fraction-of-bounding-box-diagonal parameter. Out-of-range
+// or unparseable values fall back rather than error, mirroring atoiOr; the
+// accepted range matches what the upstream demo clamps to.
+func ratioOr(s string, fallback float32) float32 {
+	if s == "" {
+		return fallback
+	}
+	f, err := strconv.ParseFloat(s, 32)
+	if err != nil || f < 0.00001 || f > 0.5 {
+		return fallback
+	}
+	return float32(f)
+}
+
+type Trellis2 struct {
+	base.SingleThread
+	// t2_generate is not thread-safe per pipeline. The gRPC server already
+	// serializes calls via Locking(), but keep a local mutex too so the
+	// invariant doesn't depend on the transport.
+	mu       sync.Mutex
+	pipeline uintptr
+}
+
+func (t *Trellis2) Load(opts *pb.ModelOptions) error {
+	set, err := resolveModels(opts.ModelFile, opts.ModelPath, opts.Options)
+	if err != nil {
+		return err
+	}
+
+	errBuf := make([]byte, 512)
+	p := t2PipelineLoad(set.dino, set.ssFlow, set.ssDec,
+		set.slatFlow, set.slatFlow1024, set.shapeDec,
+		set.shapeEnc, set.texDec, set.texSlatFlow512, set.texSlatFlow1024,
+		0 /*flags*/, &errBuf[0], int32(len(errBuf)))
+	if p == 0 {
+		return fmt.Errorf("trellis2 pipeline load: %s", cstr(errBuf))
+	}
+
+	t.mu.Lock()
+	if t.pipeline != 0 {
+		t2PipelineFree(t.pipeline)
+	}
+	t.pipeline = p
+	t.mu.Unlock()
+
+	fmt.Fprintf(os.Stderr, "trellis2 pipeline loaded: backend=%s caps=%#x\n",
+		t2PipelineBackend(p), t2PipelineCaps(p))
+	return nil
+}
+
+func (t *Trellis2) Free() error {
+	t.mu.Lock()
+	defer t.mu.Unlock()
+	if t.pipeline != 0 {
+		t2PipelineFree(t.pipeline)
+		t.pipeline = 0
+	}
+	return nil
+}
+
+func (t *Trellis2) Generate3D(opts *pb.Generate3DRequest) error {
+	if opts.Dst == "" {
+		return fmt.Errorf("dst is empty")
+	}
+	if opts.GetParams()["operation"] == "print_remesh" {
+		t.mu.Lock()
+		defer t.mu.Unlock()
+		return remeshGLB(opts)
+	}
+	img, err := os.ReadFile(opts.Src)
+	if err != nil {
+		return fmt.Errorf("reading conditioning image: %w", err)
+	}
+	if len(img) == 0 {
+		return fmt.Errorf("conditioning image is empty")
+	}
+
+	seed := uint64(opts.Seed)
+	if opts.Seed <= 0 {
+		seed = rand.Uint64()
+	}
+	guidance := opts.CfgScale
+	if guidance <= 0 {
+		guidance = -1 // <0 selects the pipeline default (7.5)
+	}
+	texSize := atoiOr(opts.GetParams()["texture_size"], 0) // <=0 selects the bake default
+	componentFilter := componentFilterFor(opts.GetParams()["components"])
+
+	// Optional CGAL Alpha Wrap: wrap the generated mesh into a watertight,
+	// intersection-free 2-manifold for 3D printing. Ratios are fractions of
+	// the bounding-box diagonal; offset defaults to alpha/30 per the CGAL
+	// guideline the upstream demo uses. Offset is deliberately not an
+	// independent parameter: looser values produce puffy or degenerate wraps.
+	printRemesh := boolParam(opts.GetParams()["print_remesh"])
+	alphaRatio := ratioOr(opts.GetParams()["alpha_ratio"], 0.005)
+	offsetRatio := alphaRatio / 30
+	if printRemesh && t2PrintRemeshAvailable() == 0 {
+		return fmt.Errorf("print_remesh requested but libtrellis2 was built without CGAL Alpha Wrap")
+	}
+
+	t.mu.Lock()
+	defer t.mu.Unlock()
+	if t.pipeline == 0 {
+		return fmt.Errorf("model not loaded")
+	}
+
+	errBuf := make([]byte, 512)
+	r := t2Generate(t.pipeline, &img[0], int32(len(img)),
+		pipelineForQuality(opts.Quality), backgroundForMode(opts.Background),
+		seed, opts.Step, guidance, opts.TextureSteps,
+		0, 0, 0, 0, // no progress/preview callbacks
+		&errBuf[0], int32(len(errBuf)))
+	if r == 0 {
+		return fmt.Errorf("trellis2 generate: %s", cstr(errBuf))
+	}
+	defer t2MeshFree(r)
+
+	nv := t2MeshNVerts(r)
+	nt := t2MeshNTris(r)
+	if nv == 0 || nt == 0 {
+		return fmt.Errorf("empty mesh")
+	}
+	var pbr *float32
+	if t2MeshHasPBR(r) != 0 {
+		pbr = t2MeshPBR(r)
+	}
+
+	// Bake straight from the mesh accessor buffers — they stay valid until
+	// t2_mesh_free, so no copies are needed.
+	var outLen int32
+	var glb *byte
+	if printRemesh {
+		wrap := t2PreparePrintMesh(t2MeshVerts(r), nv, t2MeshTris(r), nt, pbr,
+			componentFilter, alphaRatio, offsetRatio,
+			&errBuf[0], int32(len(errBuf)))
+		if wrap == 0 {
+			return fmt.Errorf("trellis2 print remesh: %s", cstr(errBuf))
+		}
+		defer t2MeshFree(wrap)
+		wnv, wnt := t2MeshNVerts(wrap), t2MeshNTris(wrap)
+		if wnv == 0 || wnt == 0 {
+			return fmt.Errorf("empty print mesh")
+		}
+		if pbr != nil {
+			// Wrapping creates new vertices, so the source material is
+			// reprojected per texel onto the wrap's UV atlas (demo handleGLB).
+			glb = t2BakeProjectedGLB(t2MeshVerts(wrap), wnv, t2MeshTris(wrap), wnt,
+				t2MeshVerts(r), nv, t2MeshTris(r), nt, pbr,
+				texSize, componentFilter,
+				&outLen, &errBuf[0], int32(len(errBuf)))
+		} else {
+			glb = t2BakeGLB(t2MeshVerts(wrap), wnv, t2MeshTris(wrap), wnt, nil,
+				texSize, 2, // the wrap output is already component-filtered
+				&outLen, &errBuf[0], int32(len(errBuf)))
+		}
+	} else {
+		glb = t2BakeGLB(t2MeshVerts(r), nv, t2MeshTris(r), nt, pbr,
+			texSize, componentFilter,
+			&outLen, &errBuf[0], int32(len(errBuf)))
+	}
+	if glb == nil {
+		return fmt.Errorf("trellis2 GLB bake: %s", cstr(errBuf))
+	}
+	defer t2FreeBuffer(glb)
+
+	out := make([]byte, int(outLen))
+	copy(out, unsafe.Slice(glb, int(outLen)))
+	return os.WriteFile(opts.Dst, out, 0600)
+}
+
+// remeshGLB applies the demo's post-generation print workflow to an existing
+// dense vertex-PBR GLB. It does not touch the inference pipeline: CGAL wrapping,
+// UV unwrapping, and PBR projection are CPU-only post-processing operations.
+func remeshGLB(opts *pb.Generate3DRequest) error {
+	if opts.Src == "" {
+		return fmt.Errorf("src is empty")
+	}
+	if t2PrintRemeshAvailable() == 0 {
+		return fmt.Errorf("print remeshing is unavailable (libtrellis2 was built without CGAL Alpha Wrap)")
+	}
+	data, err := os.ReadFile(opts.Src)
+	if err != nil {
+		return fmt.Errorf("reading source GLB: %w", err)
+	}
+	mesh, err := parseVertexGLB(data)
+	if err != nil {
+		return fmt.Errorf("reading source GLB: %w", err)
+	}
+
+	params := opts.GetParams()
+	alphaRatio := ratioOr(params["alpha_ratio"], 0.005)
+	offsetRatio := alphaRatio / 30
+	componentFilter := componentFilterFor(params["components"])
+	textureSize := atoiOr(params["texture_size"], 2048)
+	var sourcePBR *float32
+	if len(mesh.pbr) != 0 {
+		sourcePBR = &mesh.pbr[0]
+	}
+	errBuf := make([]byte, 512)
+	wrap := t2PreparePrintMesh(
+		&mesh.verts[0], int32(len(mesh.verts)/3),
+		&mesh.tris[0], int32(len(mesh.tris)/3),
+		sourcePBR, componentFilter, alphaRatio, offsetRatio,
+		&errBuf[0], int32(len(errBuf)),
+	)
+	if wrap == 0 {
+		return fmt.Errorf("trellis2 print remesh: %s", cstr(errBuf))
+	}
+	defer t2MeshFree(wrap)
+
+	wrappedVerts, wrappedTris := t2MeshNVerts(wrap), t2MeshNTris(wrap)
+	if wrappedVerts == 0 || wrappedTris == 0 {
+		return fmt.Errorf("empty print mesh")
+	}
+	var outLen int32
+	var glb *byte
+	if sourcePBR != nil {
+		glb = t2BakeProjectedGLB(
+			t2MeshVerts(wrap), wrappedVerts, t2MeshTris(wrap), wrappedTris,
+			&mesh.verts[0], int32(len(mesh.verts)/3),
+			&mesh.tris[0], int32(len(mesh.tris)/3), sourcePBR,
+			int32(textureSize), componentFilter,
+			&outLen, &errBuf[0], int32(len(errBuf)),
+		)
+	} else {
+		glb = t2BakeGLB(
+			t2MeshVerts(wrap), wrappedVerts, t2MeshTris(wrap), wrappedTris,
+			nil, int32(textureSize), 2,
+			&outLen, &errBuf[0], int32(len(errBuf)),
+		)
+	}
+	if glb == nil || outLen <= 0 {
+		return fmt.Errorf("trellis2 GLB bake: %s", cstr(errBuf))
+	}
+	defer t2FreeBuffer(glb)
+
+	out := make([]byte, int(outLen))
+	copy(out, unsafe.Slice(glb, int(outLen)))
+	return os.WriteFile(opts.Dst, out, 0o600)
+}
+
+func cstr(b []byte) string {
+	for i, c := range b {
+		if c == 0 {
+			return string(b[:i])
+		}
+	}
+	return string(b)
+}
diff --git a/backend/go/trellis2cpp/trellis2_test.go b/backend/go/trellis2cpp/trellis2_test.go
new file mode 100644
index 000000000..d492756e4
--- /dev/null
+++ b/backend/go/trellis2cpp/trellis2_test.go
@@ -0,0 +1,247 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"net"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"testing"
+	"time"
+
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/credentials/insecure"
+)
+
+func TestTrellis2Cpp(t *testing.T) {
+	RegisterFailHandler(Fail)
+	RunSpecs(t, "trellis2cpp backend suite")
+}
+
+// touch creates empty files — resolveModels only checks existence, so the
+// tests never need real GGUF weights.
+func touch(dir string, names ...string) {
+	for _, name := range names {
+		Expect(os.WriteFile(filepath.Join(dir, name), nil, 0o600)).To(Succeed())
+	}
+}
+
+var requiredFiles = []string{"dino_f16.gguf", "ss_flow_f16.gguf", "ss_dec_f16.gguf"}
+
+var fullSet = append(append([]string{}, requiredFiles...),
+	"slat_flow_f16.gguf", "slat_flow_1024_f16.gguf", "shape_dec_f16.gguf",
+	"shape_enc_f16.gguf", "tex_dec_f16.gguf",
+	"tex_slat_flow_512_f16.gguf", "tex_slat_flow_1024_f16.gguf")
+
+var _ = Describe("resolveModels", func() {
+	var dir string
+
+	BeforeEach(func() {
+		dir = GinkgoT().TempDir()
+	})
+
+	It("refuses a directory without the trellis2 component files", func() {
+		touch(dir, "some-llm.gguf")
+
+		_, err := resolveModels("some-llm.gguf", dir, nil)
+		Expect(err).To(MatchError(ContainSubstring("not a trellis2 model set")))
+	})
+
+	It("resolves every role from the full default-named set", func() {
+		touch(dir, fullSet...)
+
+		set, err := resolveModels("ss_flow_f16.gguf", dir, nil)
+		Expect(err).NotTo(HaveOccurred())
+		for name, path := range map[string]string{
+			"dino":               set.dino,
+			"ss_flow":            set.ssFlow,
+			"ss_dec":             set.ssDec,
+			"slat_flow":          set.slatFlow,
+			"slat_flow_1024":     set.slatFlow1024,
+			"shape_dec":          set.shapeDec,
+			"shape_enc":          set.shapeEnc,
+			"tex_dec":            set.texDec,
+			"tex_slat_flow_512":  set.texSlatFlow512,
+			"tex_slat_flow_1024": set.texSlatFlow1024,
+		} {
+			Expect(path).NotTo(BeEmpty(), "role %s", name)
+		}
+	})
+
+	It("degrades to coarse-only without the 512 pair, even when texture files exist", func() {
+		touch(dir, requiredFiles...)
+		touch(dir, "shape_enc_f16.gguf", "tex_dec_f16.gguf", "tex_slat_flow_512_f16.gguf", "slat_flow_1024_f16.gguf")
+
+		set, err := resolveModels("ss_flow_f16.gguf", dir, nil)
+		Expect(err).NotTo(HaveOccurred())
+		Expect(set.slatFlow).To(BeEmpty())
+		Expect(set.shapeDec).To(BeEmpty())
+		Expect(set.slatFlow1024).To(BeEmpty())
+		Expect(set.shapeEnc).To(BeEmpty())
+		Expect(set.texDec).To(BeEmpty())
+		Expect(set.texSlatFlow512).To(BeEmpty())
+		Expect(set.texSlatFlow1024).To(BeEmpty())
+	})
+
+	It("disables texturing but keeps fine geometry when the texture set is incomplete", func() {
+		touch(dir, requiredFiles...)
+		touch(dir, "slat_flow_f16.gguf", "slat_flow_1024_f16.gguf", "shape_dec_f16.gguf", "tex_dec_f16.gguf")
+
+		set, err := resolveModels("ss_flow_f16.gguf", dir, nil)
+		Expect(err).NotTo(HaveOccurred())
+		Expect(set.slatFlow).NotTo(BeEmpty())
+		Expect(set.shapeDec).NotTo(BeEmpty())
+		Expect(set.slatFlow1024).NotTo(BeEmpty())
+		Expect(set.shapeEnc).To(BeEmpty())
+		Expect(set.texDec).To(BeEmpty())
+		Expect(set.texSlatFlow512).To(BeEmpty())
+		Expect(set.texSlatFlow1024).To(BeEmpty())
+	})
+
+	It("drops the 1024 cascade when texturing lacks the HR texture flow", func() {
+		touch(dir, fullSet...)
+		Expect(os.Remove(filepath.Join(dir, "tex_slat_flow_1024_f16.gguf"))).To(Succeed())
+
+		set, err := resolveModels("ss_flow_f16.gguf", dir, nil)
+		Expect(err).NotTo(HaveOccurred())
+		Expect(set.slatFlow1024).To(BeEmpty())
+		Expect(set.shapeEnc).NotTo(BeEmpty())
+		Expect(set.texDec).NotTo(BeEmpty())
+		Expect(set.texSlatFlow512).NotTo(BeEmpty())
+	})
+
+	It("honors explicit *_path option overrides", func() {
+		touch(dir, fullSet...)
+		custom := filepath.Join(dir, "custom")
+		Expect(os.Mkdir(custom, 0o750)).To(Succeed())
+		touch(custom, "my-dino.gguf")
+
+		set, err := resolveModels("ss_flow_f16.gguf", dir, []string{"dino_path:custom/my-dino.gguf"})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(set.dino).To(Equal(filepath.Join(custom, "my-dino.gguf")))
+	})
+
+	It("fails when an explicitly configured file is missing", func() {
+		touch(dir, fullSet...)
+
+		_, err := resolveModels("ss_flow_f16.gguf", dir, []string{"tex_dec_path:nope.gguf"})
+		Expect(err).To(MatchError(ContainSubstring("missing file")))
+	})
+
+	It("rejects option paths escaping the model directory", func() {
+		touch(dir, fullSet...)
+
+		_, err := resolveModels("ss_flow_f16.gguf", dir, []string{"dino_path:../outside.gguf"})
+		Expect(err).To(HaveOccurred())
+	})
+})
+
+var _ = DescribeTable("request parameter mapping",
+	func(got, want int32) {
+		Expect(got).To(Equal(want))
+	},
+	Entry("quality empty", pipelineForQuality(""), int32(pipeAuto)),
+	Entry("quality auto", pipelineForQuality("auto"), int32(pipeAuto)),
+	Entry("quality coarse", pipelineForQuality("coarse"), int32(pipeCoarse)),
+	Entry("quality 512", pipelineForQuality("512"), int32(pipe512)),
+	Entry("quality 1024", pipelineForQuality("1024"), int32(pipe1024)),
+	Entry("background empty", backgroundForMode(""), int32(backgroundAuto)),
+	Entry("background auto", backgroundForMode("auto"), int32(backgroundAuto)),
+	Entry("background keep", backgroundForMode("keep"), int32(backgroundKeep)),
+	Entry("background black", backgroundForMode("black"), int32(backgroundBlack)),
+	Entry("background white", backgroundForMode("white"), int32(backgroundWhite)),
+	Entry("components default", componentFilterFor(""), int32(2)),
+	Entry("components all", componentFilterFor("all"), int32(2)),
+	Entry("components largest", componentFilterFor("largest"), int32(1)),
+	Entry("components tiny", componentFilterFor("tiny"), int32(0)),
+	Entry("atoi empty", atoiOr("", 0), int32(0)),
+	Entry("atoi value", atoiOr("2048", 0), int32(2048)),
+	Entry("atoi junk", atoiOr("junk", 7), int32(7)),
+)
+
+var _ = Describe("print remesh parameters", func() {
+	It("parses the print_remesh toggle", func() {
+		Expect(boolParam("1")).To(BeTrue())
+		Expect(boolParam("true")).To(BeTrue())
+		Expect(boolParam("TRUE")).To(BeTrue())
+		Expect(boolParam("")).To(BeFalse())
+		Expect(boolParam("0")).To(BeFalse())
+		Expect(boolParam("no")).To(BeFalse())
+	})
+
+	It("parses ratios and clamps junk to the fallback", func() {
+		Expect(ratioOr("", 0.005)).To(BeNumerically("~", 0.005, 1e-6))
+		Expect(ratioOr("0.01", 0.005)).To(BeNumerically("~", 0.01, 1e-6))
+		Expect(ratioOr("junk", 0.005)).To(BeNumerically("~", 0.005, 1e-6))
+		Expect(ratioOr("-1", 0.005)).To(BeNumerically("~", 0.005, 1e-6))
+		Expect(ratioOr("0.9", 0.005)).To(BeNumerically("~", 0.005, 1e-6), "above the demo's 50% cap")
+	})
+})
+
+var _ = Describe("packaged backend", func() {
+	It("starts and answers Health without loading model weights", func() {
+		runScript := os.Getenv("TRELLIS2CPP_SMOKE_RUN")
+		if runScript == "" {
+			runScript = filepath.Join("package", "run.sh")
+		}
+		if _, err := os.Stat(runScript); os.IsNotExist(err) {
+			Skip("packaged backend is not present; run make before the smoke test")
+		}
+
+		listener, err := net.Listen("tcp", "127.0.0.1:0")
+		Expect(err).NotTo(HaveOccurred())
+		addr := listener.Addr().String()
+		Expect(listener.Close()).To(Succeed())
+
+		cmd := exec.Command("bash", runScript, "--addr="+addr)
+		cmd.Stdout = GinkgoWriter
+		cmd.Stderr = GinkgoWriter
+		Expect(cmd.Start()).To(Succeed())
+		processDone := make(chan error, 1)
+		go func() { processDone <- cmd.Wait() }()
+		processExited := false
+		DeferCleanup(func() {
+			if cmd.Process != nil && !processExited {
+				_ = cmd.Process.Kill()
+				<-processDone
+			}
+		})
+
+		Eventually(func() error {
+			select {
+			case err := <-processDone:
+				processExited = true
+				if err != nil {
+					return StopTrying("backend exited before Health succeeded").Wrap(err)
+				}
+				return StopTrying("backend exited before Health succeeded")
+			default:
+			}
+
+			ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+			defer cancel()
+
+			conn, err := grpc.DialContext(ctx, addr,
+				grpc.WithTransportCredentials(insecure.NewCredentials()),
+				grpc.WithBlock(),
+			)
+			if err != nil {
+				return err
+			}
+			defer func() { _ = conn.Close() }()
+
+			reply, err := pb.NewBackendClient(conn).Health(ctx, &pb.HealthMessage{})
+			if err != nil {
+				return err
+			}
+			if string(reply.GetMessage()) != "OK" {
+				return fmt.Errorf("unexpected Health reply %q", reply.GetMessage())
+			}
+			return nil
+		}, 30*time.Second, 200*time.Millisecond).Should(Succeed())
+	})
+})
diff --git a/backend/go/valkey-store/Makefile b/backend/go/valkey-store/Makefile
new file mode 100644
index 000000000..0d7ddb327
--- /dev/null
+++ b/backend/go/valkey-store/Makefile
@@ -0,0 +1,17 @@
+GOCMD=go
+
+valkey-store:
+	CGO_ENABLED=0 $(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o valkey-store ./
+
+package:
+	bash package.sh
+
+build: valkey-store package
+
+## Runs the backend's Ginkgo suite. The unit (mock) specs run without a
+## container; the integration specs skip automatically unless VALKEY_ADDR is set.
+test:
+	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --flake-attempts $(TEST_FLAKES) -v -r ./
+
+clean:
+	rm -f valkey-store
diff --git a/backend/go/valkey-store/config.go b/backend/go/valkey-store/config.go
new file mode 100644
index 000000000..02c5711b4
--- /dev/null
+++ b/backend/go/valkey-store/config.go
@@ -0,0 +1,261 @@
+package main
+
+// Connection + index configuration for the Valkey-backed vector store.
+//
+// Configuration is read from the model config `options:` list (a repeated
+// `key:value` string carried over gRPC in ModelOptions.Options) rather than
+// from process-wide environment variables. Driving it from the model config is
+// the LocalAI convention and, crucially, lets multiple stores each have their
+// own Valkey config (a face registry on one server, a router cache on another)
+// within a single LocalAI process — something a single VALKEY_* env surface
+// could never express. Every default lives as a named constant below — no
+// magic literals sprinkled through the store logic — so the defaults can be
+// audited in one place and referenced by the unit tests.
+//
+// Example model YAML:
+//
+//	name: my-vector-store
+//	backend: valkey-store
+//	options:
+//	  - addr:valkey.internal:6379
+//	  - index_algo:HNSW
+//	  - distance_metric:COSINE
+
+import (
+	"fmt"
+	"os"
+	"strconv"
+	"strings"
+	"time"
+
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+	"github.com/mudler/xlog"
+)
+
+const (
+	// _defaultAddr is the single-node Valkey address used when the `addr`
+	// option is unset. Matches the port Phase 0 reserved for integration tests.
+	_defaultAddr = "localhost:6379"
+
+	// _defaultClientName is mandatory: every connection identifies itself with
+	// this name so operators can spot LocalAI's traffic via CLIENT LIST. It is
+	// always set on the client, even if the operator clears the client_name option.
+	_defaultClientName = "localai-valkey-store"
+
+	// _defaultIndexAlgo is FLAT (exact brute-force KNN) to preserve parity with
+	// local-store's linear scan and keep the exact-cosine test expectations.
+	_defaultIndexAlgo = indexAlgoFlat
+
+	// _defaultDistanceMetric is COSINE so similarities match local-store
+	// (sim = 1 - cosine_distance). L2/IP are opt-in.
+	_defaultDistanceMetric = distanceCosine
+
+	// HNSW graph defaults (only used when index_algo=HNSW). Values follow
+	// the Valkey Search documented defaults.
+	_defaultHNSWM              = 16
+	_defaultHNSWEFConstruction = 200
+	_defaultHNSWEFRuntime      = 10
+
+	// _defaultRequestTimeoutMS bounds every command. We deliberately do NOT rely
+	// on the client's built-in write timeout: index back-fill or a slow KNN can
+	// exceed a short default, so we thread this explicit deadline into every
+	// command context.
+	_defaultRequestTimeoutMS = 5000
+
+	// Valkey Search index algorithms.
+	indexAlgoFlat = "FLAT"
+	indexAlgoHNSW = "HNSW"
+
+	// Supported distance metrics.
+	distanceCosine = "COSINE"
+	distanceL2     = "L2"
+	distanceIP     = "IP"
+
+	// Option keys recognised in the model config `options:` list. They mirror
+	// the previous VALKEY_* env var names without the prefix and lower-cased, so
+	// operators migrating a config have an obvious 1:1 mapping.
+	optAddr               = "addr"
+	optUsername           = "username"
+	optPassword           = "password"
+	optUsernameEnv        = "username_env"
+	optPasswordEnv        = "password_env"
+	optTLS                = "tls"
+	optTLSSkipVerify      = "tls_skip_verify"
+	optTLSCACert          = "tls_ca_cert"
+	optClientName         = "client_name"
+	optDB                 = "db"
+	optIndexAlgo          = "index_algo"
+	optDistanceMetric     = "distance_metric"
+	optHNSWM              = "hnsw_m"
+	optHNSWEFConstruction = "hnsw_ef_construction"
+	optHNSWEFRuntime      = "hnsw_ef_runtime"
+	optRequestTimeoutMS   = "request_timeout_ms"
+)
+
+// hnswParams holds the HNSW-only tuning knobs. They are ignored unless
+// IndexAlgo == indexAlgoHNSW.
+type hnswParams struct {
+	M              int
+	EFConstruction int
+	EFRuntime      int
+}
+
+// Config is the fully-resolved store configuration produced by loadConfig().
+type Config struct {
+	Addr           string
+	Username       string
+	Password       string
+	UseTLS         bool
+	TLSSkipVerify  bool
+	TLSCACert      string
+	ClientName     string
+	DB             int
+	IndexAlgo      string
+	DistanceMetric string
+	HNSW           hnswParams
+	RequestTimeout time.Duration
+}
+
+// parseOptions turns the repeated `key:value` ModelOptions.Options list into a
+// lookup map. The split is on the FIRST ':' via strings.Cut, so values that
+// themselves contain a colon (e.g. `addr:host:6379`) are preserved intact. A
+// malformed entry with no ':' is warned about and skipped rather than silently
+// dropped, so an operator typo is visible in the logs.
+func parseOptions(opts *pb.ModelOptions) map[string]string {
+	m := make(map[string]string)
+	if opts == nil {
+		return m
+	}
+	for _, o := range opts.GetOptions() {
+		k, v, ok := strings.Cut(o, ":")
+		if !ok {
+			xlog.Warn("valkey-store: ignoring malformed option (want key:value)", "option", o)
+			continue
+		}
+		m[strings.ToLower(strings.TrimSpace(k))] = strings.TrimSpace(v)
+	}
+	return m
+}
+
+// loadConfig resolves the store configuration from the model config options and
+// returns a validated Config. It fails fast on an unknown index algorithm or
+// distance metric (and on a malformed integer) so a misconfiguration surfaces
+// at Load() rather than silently degrading search.
+func loadConfig(opts *pb.ModelOptions) (Config, error) {
+	o := parseOptions(opts)
+
+	// intOr parses an integer option, failing fast on a malformed value the same
+	// way an invalid index algo or distance metric does. A typo like
+	// `hnsw_m:1x6` must surface at Load() rather than silently degrading to the
+	// default and producing subtly wrong (and hard-to-diagnose) index
+	// behaviour. The first parse error wins and is returned below.
+	var parseErr error
+	intOr := func(key string, fallback int) int {
+		v, ok := o[key]
+		if !ok || v == "" {
+			return fallback
+		}
+		n, err := strconv.Atoi(v)
+		if err != nil {
+			if parseErr == nil {
+				parseErr = fmt.Errorf("valkey-store: invalid option %s %q: %w", key, v, err)
+			}
+			return fallback
+		}
+		return n
+	}
+
+	cfg := Config{
+		Addr:           strOr(o, optAddr, _defaultAddr),
+		Username:       resolveCredential(o, optUsername, optUsernameEnv),
+		Password:       resolveCredential(o, optPassword, optPasswordEnv),
+		UseTLS:         boolOr(o, optTLS, false),
+		TLSSkipVerify:  boolOr(o, optTLSSkipVerify, false),
+		TLSCACert:      o[optTLSCACert],
+		ClientName:     strOr(o, optClientName, _defaultClientName),
+		DB:             intOr(optDB, 0),
+		IndexAlgo:      strings.ToUpper(strOr(o, optIndexAlgo, _defaultIndexAlgo)),
+		DistanceMetric: strings.ToUpper(strOr(o, optDistanceMetric, _defaultDistanceMetric)),
+		HNSW: hnswParams{
+			M:              intOr(optHNSWM, _defaultHNSWM),
+			EFConstruction: intOr(optHNSWEFConstruction, _defaultHNSWEFConstruction),
+			EFRuntime:      intOr(optHNSWEFRuntime, _defaultHNSWEFRuntime),
+		},
+		RequestTimeout: time.Duration(intOr(optRequestTimeoutMS, _defaultRequestTimeoutMS)) * time.Millisecond,
+	}
+	if parseErr != nil {
+		return Config{}, parseErr
+	}
+
+	// ClientName is mandatory. Restore the default if the operator blanked it,
+	// so the connection is always identifiable.
+	if cfg.ClientName == "" {
+		cfg.ClientName = _defaultClientName
+	}
+
+	if cfg.DB < 0 {
+		return Config{}, fmt.Errorf("valkey-store: invalid option %s %d (must be >= 0)", optDB, cfg.DB)
+	}
+
+	switch cfg.IndexAlgo {
+	case indexAlgoFlat, indexAlgoHNSW:
+	default:
+		return Config{}, fmt.Errorf("valkey-store: invalid option %s %q (want FLAT or HNSW)", optIndexAlgo, cfg.IndexAlgo)
+	}
+
+	switch cfg.DistanceMetric {
+	case distanceCosine, distanceL2, distanceIP:
+	default:
+		return Config{}, fmt.Errorf("valkey-store: invalid option %s %q (want COSINE, L2 or IP)", optDistanceMetric, cfg.DistanceMetric)
+	}
+
+	if cfg.RequestTimeout <= 0 {
+		cfg.RequestTimeout = time.Duration(_defaultRequestTimeoutMS) * time.Millisecond
+	}
+
+	return cfg, nil
+}
+
+// strOr returns the option value for key, or fallback when it is unset/empty.
+func strOr(o map[string]string, key, fallback string) string {
+	if v, ok := o[key]; ok && v != "" {
+		return v
+	}
+	return fallback
+}
+
+// boolOr parses a boolean option, falling back to the default on an unset or
+// unparseable value. A typo is surfaced via a warning (like the previous env
+// behaviour) rather than failing Load for a coarse on/off switch.
+func boolOr(o map[string]string, key string, fallback bool) bool {
+	v, ok := o[key]
+	if !ok || v == "" {
+		return fallback
+	}
+	b, err := strconv.ParseBool(v)
+	if err != nil {
+		xlog.Warn("valkey-store: ignoring unparseable option, using default", "key", key, "value", v, "default", fallback)
+		return fallback
+	}
+	return b
+}
+
+// resolveCredential resolves a credential value with the following priority:
+//  1. Direct value from the model config option (e.g. `username:admin`)
+//  2. Env-indirection: if `username_env` names an env var, read the credential
+//     from that variable (e.g. `username_env:MY_VALKEY_USER` → os.Getenv("MY_VALKEY_USER"))
+//
+// The env-indirection pattern (same as cloud-proxy's api_key_env) avoids putting
+// secrets directly in model YAML: distinct store configs can each reference a
+// different credential env var without any plaintext passwords in the config.
+func resolveCredential(o map[string]string, directKey, envKey string) string {
+	// Direct value takes precedence (backward compatible).
+	if v := o[directKey]; v != "" {
+		return v
+	}
+	// Env indirection: the option names an env var that holds the credential.
+	if envVar := o[envKey]; envVar != "" {
+		return os.Getenv(envVar)
+	}
+	return ""
+}
diff --git a/backend/go/valkey-store/encoding.go b/backend/go/valkey-store/encoding.go
new file mode 100644
index 000000000..cec086070
--- /dev/null
+++ b/backend/go/valkey-store/encoding.go
@@ -0,0 +1,80 @@
+package main
+
+// Vector⇄key encoding: the "vector IS the key" resolution.
+//
+// local-store keys entries *by* the vector itself (a []float32). Valkey hashes
+// are keyed by strings, so we synthesise a deterministic, lossless key:
+//
+//	key = prefix + hex(little-endian float32 bytes of the vector)
+//
+// The same vector always produces the same bytes, so HSET is an upsert and
+// HGET/DEL are exact matches — and the encoding is reversible, so we can hand
+// the original []float32 back on Get/Find.
+//
+// Divergence from local-store (documented and tested): local-store compares
+// keys with slices.Compare, which treats -0.0 == +0.0 and orders NaN, so those
+// collapse to the same logical key. Byte-encoding makes -0.0 and +0.0 (and any
+// distinct NaN bit-pattern) *distinct* keys. We accept this on purpose: a
+// lossless, deterministic, exact round-trip is more valuable for a persistent
+// store than reproducing local-store's float-equality quirk, and callers never
+// rely on -0.0/+0.0 aliasing.
+
+import (
+	"encoding/binary"
+	"encoding/hex"
+	"fmt"
+	"math"
+	"strings"
+)
+
+// _float32Bytes is the wire width of a single FLOAT32 component.
+const _float32Bytes = 4
+
+// vecToBytes encodes a vector as little-endian float32 bytes. This is byte-for
+// -byte identical to valkey.VectorString32, so the value we store in the hash
+// `vec` field and the bytes we hash into the key share one encoding.
+func vecToBytes(v []float32) []byte {
+	b := make([]byte, len(v)*_float32Bytes)
+	for i, e := range v {
+		off := i * _float32Bytes
+		binary.LittleEndian.PutUint32(b[off:off+_float32Bytes], math.Float32bits(e))
+	}
+	return b
+}
+
+// bytesToVec reverses vecToBytes. It rejects a payload whose length is not a
+// multiple of the float32 width, which would indicate a corrupted/foreign value.
+func bytesToVec(b []byte) ([]float32, error) {
+	if len(b)%_float32Bytes != 0 {
+		return nil, fmt.Errorf("valkey-store: vector byte length %d is not a multiple of %d", len(b), _float32Bytes)
+	}
+	v := make([]float32, len(b)/_float32Bytes)
+	for i := range v {
+		off := i * _float32Bytes
+		v[i] = math.Float32frombits(binary.LittleEndian.Uint32(b[off : off+_float32Bytes]))
+	}
+	return v, nil
+}
+
+// encodeKey builds the Valkey hash key for a vector: prefix + hex(bytes).
+// Hex keeps the key printable (so it is safe in FT.CREATE PREFIX and in logs)
+// while staying lossless.
+func encodeKey(prefix string, v []float32) string {
+	return prefix + hex.EncodeToString(vecToBytes(v))
+}
+
+// decodeKey reverses encodeKey. It is intentionally retained as the tested,
+// symmetric inverse of encodeKey — it is NOT on the hot Find path (StoresFind
+// decodes the returned `vec` bytes via bytesToVec directly), but keeping the
+// key↔vector mapping provably invertible guards the encoding contract and is
+// exercised by the round-trip unit tests.
+func decodeKey(prefix, key string) ([]float32, error) {
+	if !strings.HasPrefix(key, prefix) {
+		return nil, fmt.Errorf("valkey-store: key %q does not have expected prefix %q", key, prefix)
+	}
+	b, err := hex.DecodeString(strings.TrimPrefix(key, prefix))
+	if err != nil {
+		return nil, fmt.Errorf("valkey-store: decode key hex: %w", err)
+	}
+	return bytesToVec(b)
+}
diff --git a/backend/go/valkey-store/encoding_test.go b/backend/go/valkey-store/encoding_test.go
new file mode 100644
index 000000000..1c89f1ab1
--- /dev/null
+++ b/backend/go/valkey-store/encoding_test.go
@@ -0,0 +1,77 @@
+package main
+
+// Unit tests for the vector⇄key encoding. These need no Valkey server: they
+// exercise the pure lossless-encoding contract that the whole store relies on,
+// including the documented edge cases (-0.0/+0.0 and NaN) where this encoding
+// intentionally diverges from local-store's slices.Compare float equality.
+
+import (
+	"math"
+	"math/rand/v2"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+	valkey "github.com/valkey-io/valkey-go"
+)
+
+var _ = Describe("vector⇄bytes encoding", func() {
+	It("round-trips vectors of varying dimensions", func() {
+		r := rand.New(rand.NewPCG(1, 2))
+		for _, dim := range []int{1, 3, 4, 16, 128, 768} {
+			v := make([]float32, dim)
+			for i := range v {
+				v[i] = float32(r.NormFloat64())
+			}
+			got, err := bytesToVec(vecToBytes(v))
+			Expect(err).NotTo(HaveOccurred())
+			Expect(got).To(Equal(v))
+		}
+	})
+
+	It("matches valkey.VectorString32 byte-for-byte", func() {
+		// The stored `vec` field uses valkey.VectorString32; the key uses
+		// vecToBytes. They must be the same encoding or Get/Find break.
+		v := []float32{0.1, -0.2, 3.5, 0}
+		Expect(valkey.BinaryString(vecToBytes(v))).To(Equal(valkey.VectorString32(v)))
+	})
+
+	It("rejects a byte payload that is not a multiple of 4", func() {
+		_, err := bytesToVec([]byte{1, 2, 3})
+		Expect(err).To(HaveOccurred())
+	})
+})
+
+var _ = Describe("key encoding", func() {
+	const prefix = "vs:test:"
+
+	It("round-trips key encode/decode", func() {
+		v := []float32{0.5, 0.5, 0.5}
+		key := encodeKey(prefix, v)
+		Expect(key).To(HavePrefix(prefix))
+		got, err := decodeKey(prefix, key)
+		Expect(err).NotTo(HaveOccurred())
+		Expect(got).To(Equal(v))
+	})
+
+	It("produces distinct keys for -0.0 and +0.0 (documented divergence)", func() {
+		negZero := float32(math.Copysign(0, -1))
+		posZero := float32(0)
+		Expect(math.Signbit(float64(negZero))).To(BeTrue())
+		Expect(encodeKey(prefix, []float32{negZero})).NotTo(Equal(encodeKey(prefix, []float32{posZero})))
+	})
+
+	It("produces a stable, distinct key for a NaN component", func() {
+		nan := float32(math.NaN())
+		k1 := encodeKey(prefix, []float32{nan})
+		k2 := encodeKey(prefix, []float32{nan})
+		// Deterministic: same NaN bit-pattern → same key.
+		Expect(k1).To(Equal(k2))
+		// Distinct from a normal value.
+		Expect(k1).NotTo(Equal(encodeKey(prefix, []float32{0})))
+	})
+
+	It("rejects a key without the expected prefix", func() {
+		_, err := decodeKey(prefix, "wrong:deadbeef")
+		Expect(err).To(HaveOccurred())
+	})
+})
diff --git a/backend/go/valkey-store/main.go b/backend/go/valkey-store/main.go
new file mode 100644
index 000000000..ba030899a
--- /dev/null
+++ b/backend/go/valkey-store/main.go
@@ -0,0 +1,25 @@
+package main
+
+// Note: this is started internally by LocalAI and a server is allocated for each store
+
+import (
+	"flag"
+	"os"
+
+	grpc "github.com/mudler/LocalAI/pkg/grpc"
+	"github.com/mudler/xlog"
+)
+
+var (
+	addr = flag.String("addr", "localhost:50051", "the address to connect to")
+)
+
+func main() {
+	xlog.SetLogger(xlog.NewLogger(xlog.LogLevel(os.Getenv("LOCALAI_LOG_LEVEL")), os.Getenv("LOCALAI_LOG_FORMAT")))
+
+	flag.Parse()
+
+	if err := grpc.StartServer(*addr, NewValkeyStore()); err != nil {
+		panic(err)
+	}
+}
diff --git a/backend/go/valkey-store/package.sh b/backend/go/valkey-store/package.sh
new file mode 100644
index 000000000..c6b487836
--- /dev/null
+++ b/backend/go/valkey-store/package.sh
@@ -0,0 +1,12 @@
+#!/bin/bash
+
+# Script to copy the appropriate libraries based on architecture
+# This script is used in the final stage of the Dockerfile
+
+set -e
+
+CURDIR=$(dirname "$(realpath $0)")
+
+mkdir -p $CURDIR/package
+cp -avf $CURDIR/valkey-store $CURDIR/package/
+cp -rfv $CURDIR/run.sh $CURDIR/package/
diff --git a/backend/go/valkey-store/run.sh b/backend/go/valkey-store/run.sh
new file mode 100644
index 000000000..caf370e2e
--- /dev/null
+++ b/backend/go/valkey-store/run.sh
@@ -0,0 +1,6 @@
+#!/bin/bash
+set -ex
+
+CURDIR=$(dirname "$(realpath "$0")")
+
+exec "$CURDIR"/valkey-store "$@"
diff --git a/backend/go/valkey-store/store.go b/backend/go/valkey-store/store.go
new file mode 100644
index 000000000..b2b8814a5
--- /dev/null
+++ b/backend/go/valkey-store/store.go
@@ -0,0 +1,692 @@
+package main
+
+// Valkey-backed vector store, exposed as a gRPC backend. It mirrors the public
+// contract of backend/go/local-store (the four Stores* RPCs + Load) but swaps
+// the in-memory sorted slices for Valkey Search (FT.*) so the data persists
+// across restarts and can scale beyond an O(N) scan (opt-in HNSW).
+//
+// Data model — each entry is a Valkey HASH keyed by
+//
+//	prefix + hex(little-endian float32 bytes of the vector)
+//
+// with two fields: `vec` (the raw float32 bytes, indexed by a lazily-created
+// FT VECTOR index of the discovered dimension) and `val` (the opaque value
+// bytes). The vector-IS-the-key encoding (see encoding.go) makes Set an
+// HSET upsert, Get an HGET, Delete a DEL, and Find an FT.SEARCH KNN.
+//
+// Similarity — Valkey returns cosine *distance* (0 = identical, 2 = opposite),
+// while local-store returns cosine *similarity* (1 = identical, -1 = opposite).
+// We convert sim = 1 - distance for COSINE so the values match local-store's
+// integration expectations exactly. For L2/IP the raw score is passed through.
+//
+// Concurrency — base.SingleThread serialises gRPC calls, so the store's
+// scalar bookkeeping (keyLen, indexCreated) needs no extra locking. All Valkey
+// commands are synchronous via client.Do and bounded by an explicit
+// per-request deadline (cfg.RequestTimeout); there is no background event loop.
+
+import (
+	"context"
+	"crypto/sha256"
+	"crypto/tls"
+	"crypto/x509"
+	"encoding/hex"
+	"fmt"
+	"net"
+	"os"
+	"strconv"
+	"strings"
+
+	"github.com/mudler/LocalAI/pkg/grpc/base"
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+	"github.com/mudler/LocalAI/pkg/store"
+	"github.com/mudler/xlog"
+	valkey "github.com/valkey-io/valkey-go"
+)
+
+const (
+	// Hash field names. `vec` is the indexed vector; `val` is the opaque value.
+	_vecField = "vec"
+	_valField = "val"
+	// _scoreField is the KNN distance alias produced by the query and returned
+	// by FT.SEARCH. Double-underscore avoids colliding with a stored field.
+	_scoreField = "__score"
+
+	// _keyPrefixPrefix / _indexPrefix namespace the keys and index per model so
+	// two namespaces (e.g. a 512-d face store and a 192-d voice store) sharing
+	// one Valkey server never collide.
+	_keyPrefixPrefix = "vs:"
+	_indexPrefix     = "idx:"
+
+	// _maxTopK bounds a Find so an accidental or abusive huge TopK cannot force
+	// an unbounded server-side LIMIT / allocation. local-store has no cap, but
+	// it is in-memory; a networked backend wants a guard. Callers asking for
+	// more than this get the top _maxTopK results.
+	_maxTopK = 10000
+
+	// _maxNsTokenLen bounds the human-readable portion of a namespace token so
+	// a very long model name cannot produce an unbounded key prefix / index
+	// name. The appended short hash keeps distinct namespaces collision-free
+	// even when their sanitized prefixes are truncated to the same value.
+	_maxNsTokenLen = 64
+)
+
+// ValkeyStore implements the gRPC store Backend against Valkey Search.
+type ValkeyStore struct {
+	base.SingleThread
+
+	client valkey.Client
+	cfg    Config
+
+	// prefix is the per-namespace key prefix; indexName is the FT index name.
+	prefix    string
+	indexName string
+
+	// keyLen is the vector dimension, learned from the first Set. -1 means
+	// "no keys yet" — mirrors local-store so dimension-mismatch errors are
+	// identical. indexCreated tracks whether FT.CREATE has run (lazy creation).
+	keyLen       int
+	indexCreated bool
+}
+
+// NewValkeyStore returns a store with an open dimension and no index yet. The
+// Valkey client is established in Load once the connection config is known.
+func NewValkeyStore() *ValkeyStore {
+	return &ValkeyStore{keyLen: -1}
+}
+
+// newWithClient builds a store around an already-constructed client for a given
+// namespace. It exists so unit tests can inject a mock client without a real
+// Valkey server; Load is the production path.
+func newWithClient(client valkey.Client, cfg Config, namespace string) *ValkeyStore {
+	return &ValkeyStore{
+		client:    client,
+		cfg:       cfg,
+		prefix:    keyPrefix(namespace),
+		indexName: indexName(namespace),
+		keyLen:    -1,
+	}
+}
+
+// Load reads the store config from the model config options, connects, and
+// verifies the connection. The mandatory ClientName is always set so the
+// connection is identifiable via CLIENT LIST. opts.Model is the namespace
+// identifier (one process per (backend, model) tuple upstream), so we derive
+// an isolated key prefix and index name from it, and opts.Options carries the
+// per-store connection/index configuration.
+//
+// The NamespacePrefix gate mirrors local-store: core's StoreBackend always
+// sends the model name with store.NamespacePrefix; anything else is the model
+// loader's greedy autoload probing with a real model name, which must be
+// refused or the LLM binds to the vector store (the #9287 failure mode).
+func (s *ValkeyStore) Load(opts *pb.ModelOptions) error {
+	if opts == nil {
+		return fmt.Errorf("valkey-store: refusing to load: nil model options (expected %q prefix)", store.NamespacePrefix)
+	}
+	if !strings.HasPrefix(opts.GetModel(), store.NamespacePrefix) {
+		return fmt.Errorf("valkey-store: refusing to load %q: not a store namespace (expected %q prefix)", opts.GetModel(), store.NamespacePrefix)
+	}
+
+	cfg, err := loadConfig(opts)
+	if err != nil {
+		return err
+	}
+	s.cfg = cfg
+
+	namespace := opts.Model
+	s.prefix = keyPrefix(namespace)
+	s.indexName = indexName(namespace)
+
+	clientOpt := valkey.ClientOption{
+		InitAddress: []string{cfg.Addr},
+		Username:    cfg.Username,
+		Password:    cfg.Password,
+		ClientName:  cfg.ClientName,
+		// SelectDB picks a logical Valkey DB (SELECT n) for deployments that use
+		// numbered DBs for isolation. Defaults to 0; namespace prefixing already
+		// isolates keyspaces on a shared DB.
+		SelectDB: cfg.DB,
+		// Disable client-side caching: values are opaque blobs written once and
+		// read rarely, so tracking invalidations would only add overhead.
+		DisableCache: true,
+	}
+	if cfg.UseTLS {
+		tlsCfg, err := buildTLSConfig(cfg)
+		if err != nil {
+			return err
+		}
+		clientOpt.TLSConfig = tlsCfg
+	}
+
+	// Close any client from a previous Load so a re-entrant Load does not leak
+	// the old connection. Not reachable in the one-process-per-namespace model
+	// today, but keeps Load idempotent.
+	if s.client != nil {
+		s.client.Close()
+		s.client = nil
+	}
+
+	client, err := valkey.NewClient(clientOpt)
+	if err != nil {
+		return fmt.Errorf("valkey-store: connect to %s: %w", cfg.Addr, err)
+	}
+	s.client = client
+
+	// Fail fast if the server is unreachable, mirroring how a real vector DB
+	// backend would refuse to load against a dead endpoint.
+	ctx, cancel := s.ctx()
+	defer cancel()
+	if err := s.client.Do(ctx, s.client.B().Ping().Build()).Error(); err != nil {
+		s.client.Close()
+		s.client = nil
+		return fmt.Errorf("valkey-store: ping %s: %w", cfg.Addr, err)
+	}
+
+	// A durable Valkey may already hold this namespace's index from a previous
+	// run (this is the persistence capability local-store lacks). Recover both
+	// its existence AND its vector dimension so Find works before this fresh
+	// process issues its first Set, and — critically — so a post-restart Set
+	// validates the incoming dimension against the real persisted DIM instead
+	// of silently re-learning a wrong one and dropping mismatched vectors from
+	// the index (which would return success while making the entry unsearchable).
+	s.loadIndexState(ctx)
+
+	// Log the sanitized index name (which identifies the namespace) rather than
+	// the raw model-derived namespace, which could carry control characters.
+	xlog.Info("valkey-store loaded", "addr", cfg.Addr, "index", s.indexName, "algo", cfg.IndexAlgo, "metric", cfg.DistanceMetric, "indexExists", s.indexCreated, "keyLen", s.keyLen)
+	return nil
+}
+
+// loadIndexState issues one FT.INFO at Load to recover the persisted index
+// state. FT.INFO returns an error for an unknown index, so a successful reply
+// means the index exists (indexCreated=true). We then recover the vector
+// dimension from the reply and seed keyLen with it: without this, keyLen would
+// stay -1 after a restart and the next Set would blindly re-learn whatever
+// dimension the caller happened to send, accepting a mismatched vector that
+// FT never indexes (silent search-side data loss). If the dimension can't be
+// parsed (e.g. an unexpected FT.INFO layout on some server version), keyLen
+// is left at -1 and validation degrades to the pre-restart lazy behaviour.
+func (s *ValkeyStore) loadIndexState(ctx context.Context) {
+	msg, err := s.client.Do(ctx, s.client.B().FtInfo().Index(s.indexName).Build()).ToMessage()
+	if err != nil {
+		return
+	}
+	s.indexCreated = true
+	if dim, ok := findDimensions(msg); ok && dim > 0 {
+		s.keyLen = dim
+	}
+}
+
+// findDimensions walks an FT.INFO reply for the vector field's dimension. In
+// Valkey Search the VECTOR attribute nests its parameters under an `index`
+// array whose `dimensions` key holds the DIM the index was created with. The
+// reply is a nested array (RESP2) or map (RESP3), so we search recursively for
+// a `dimensions` key/token and read the value that follows it, tolerating both
+// integer and string-encoded values.
+func findDimensions(m valkey.ValkeyMessage) (int, bool) {
+	if m.IsMap() {
+		mp, err := m.AsMap()
+		if err != nil {
+			return 0, false
+		}
+		for k, v := range mp {
+			if strings.EqualFold(k, "dimensions") {
+				if n, ok := msgToInt(v); ok {
+					return n, true
+				}
+			}
+			if d, ok := findDimensions(v); ok {
+				return d, true
+			}
+		}
+		return 0, false
+	}
+	if m.IsArray() {
+		arr, err := m.ToArray()
+		if err != nil {
+			return 0, false
+		}
+		for i := range arr {
+			if s, err := arr[i].ToString(); err == nil && strings.EqualFold(s, "dimensions") && i+1 < len(arr) {
+				if n, ok := msgToInt(arr[i+1]); ok {
+					return n, true
+				}
+			}
+			if d, ok := findDimensions(arr[i]); ok {
+				return d, true
+			}
+		}
+	}
+	return 0, false
+}
+
+// msgToInt reads an integer from a ValkeyMessage that may be an integer reply
+// or a string-encoded integer (FT.INFO mixes both across fields/versions).
+func msgToInt(m valkey.ValkeyMessage) (int, bool) {
+	if n, err := m.ToInt64(); err == nil {
+		return int(n), true
+	}
+	if s, err := m.ToString(); err == nil {
+		if n, err := strconv.Atoi(s); err == nil {
+			return n, true
+		}
+	}
+	return 0, false
+}
+
+// Free closes the Valkey client. Called by the gRPC server on shutdown.
+func (s *ValkeyStore) Free() error {
+	if s.client != nil {
+		s.client.Close()
+		s.client = nil
+	}
+	return nil
+}
+
+// buildTLSConfig assembles the tls.Config for a tls=true connection.
+// Go only auto-derives ServerName (SNI) from the dial address for hostnames;
+// for an IP-addressed endpoint (e.g. 10.0.0.5:6379) SNI is left empty and the
+// certificate's SANs won't match the raw IP, so verification fails. We set it
+// explicitly from the configured host so both hostname and IP endpoints verify.
+// A custom CA bundle (tls_ca_cert) and an explicit insecure-skip escape hatch
+// (tls_skip_verify) are supported for enterprise/self-signed setups.
+func buildTLSConfig(cfg Config) (*tls.Config, error) {
+	tlsCfg := &tls.Config{}
+	if host, _, err := net.SplitHostPort(cfg.Addr); err == nil && host != "" {
+		tlsCfg.ServerName = host
+	}
+	if cfg.TLSSkipVerify {
+		tlsCfg.InsecureSkipVerify = true
+	}
+	if cfg.TLSCACert != "" {
+		pem, err := os.ReadFile(cfg.TLSCACert)
+		if err != nil {
+			return nil, fmt.Errorf("valkey-store: read tls_ca_cert %q: %w", cfg.TLSCACert, err)
+		}
+		pool := x509.NewCertPool()
+		if !pool.AppendCertsFromPEM(pem) {
+			return nil, fmt.Errorf("valkey-store: tls_ca_cert %q: no valid certificate found", cfg.TLSCACert)
+		}
+		tlsCfg.RootCAs = pool
+	}
+	return tlsCfg, nil
+}
+
+// ctx returns a request-scoped context bounded by the configured timeout. We
+// never rely on the client's built-in write timeout because index back-fill
+// and large KNN queries can legitimately exceed a short default.
+func (s *ValkeyStore) ctx() (context.Context, context.CancelFunc) {
+	return context.WithTimeout(context.Background(), s.cfg.RequestTimeout)
+}
+
+func (s *ValkeyStore) StoresSet(opts *pb.StoresSetOptions) error {
+	keys := store.UnwrapKeys(opts.Keys)
+	values := store.UnwrapValues(opts.Values)
+	if len(keys) == 0 {
+		return fmt.Errorf("valkey-store: Set: no keys to add")
+	}
+	if len(keys) != len(values) {
+		return fmt.Errorf("valkey-store: Set: len(keys) = %d, len(values) = %d", len(keys), len(values))
+	}
+
+	// Learn the dimension from the first key ever set (mirrors local-store's
+	// keyLen == -1 sentinel), then reject anything that disagrees. checkDims is
+	// the single source of truth for the per-key length check (shared with
+	// Get/Delete/Find) so the four RPCs cannot drift apart.
+	if s.keyLen == -1 {
+		s.keyLen = len(keys[0])
+	}
+	if err := s.checkDims("Set", keys); err != nil {
+		return err
+	}
+
+	// The index needs the dimension up front, but local-store learns it from
+	// the first Set — so we create it lazily here, once, before writing.
+	if err := s.ensureIndex(s.keyLen); err != nil {
+		return err
+	}
+
+	// Write each entry with an individual round-trip rather than pipelining the
+	// whole batch via DoMulti. Valkey Search indexes every HSET into the
+	// FLAT/HNSW index synchronously on the server's main thread; a large
+	// pipeline of indexed writes can fill the socket buffers while that
+	// indexing keeps the server from draining them, deadlocking the connection
+	// (observed as an i/o timeout on high-dimension batches — a single 768-d
+	// DoMulti of ~20 vectors hangs, while the same writes issued sequentially
+	// complete in milliseconds). Sequential writes keep each command fully
+	// round-tripped and stay fast (hundreds of 768-d vectors in a few hundred
+	// ms). A single failure fails the whole Set — partial writes are surfaced,
+	// not swallowed.
+	//
+	// The request timeout is applied PER command, not once across the whole
+	// loop: an unbounded SetCols against a remote Valkey would otherwise exhaust
+	// a single aggregate deadline mid-batch and leave a partial, non-atomic
+	// write.
+	for i, k := range keys {
+		cmd := s.client.B().Hset().Key(encodeKey(s.prefix, k)).
+			FieldValue().
+			FieldValue(_vecField, valkey.BinaryString(vecToBytes(k))).
+			FieldValue(_valField, valkey.BinaryString(values[i])).
+			Build()
+		ctx, cancel := s.ctx()
+		err := s.client.Do(ctx, cmd).Error()
+		cancel()
+		if err != nil {
+			return fmt.Errorf("valkey-store: Set: HSET key %d: %w", i, err)
+		}
+	}
+	return nil
+}
+
+// StoresGet fetches values for the given keys. Missing keys are omitted from
+// the result (not errored), matching local-store; returned slices are aligned.
+func (s *ValkeyStore) StoresGet(opts *pb.StoresGetOptions) (pb.StoresGetResult, error) {
+	keys := store.UnwrapKeys(opts.Keys)
+	if len(keys) == 0 {
+		return pb.StoresGetResult{}, nil
+	}
+	if err := s.checkDims("Get", keys); err != nil {
+		return pb.StoresGetResult{}, err
+	}
+
+	// Reads pipeline the whole batch via DoMulti under ONE aggregate deadline,
+	// unlike Set/Delete which use a per-command timeout. That asymmetry is
+	// deliberate: HGET is non-mutating, so exhausting the deadline mid-batch
+	// only truncates the result (surfaced as an error) — it can never leave a
+	// partial write behind, which is the specific hazard the per-command timeout
+	// guards against for Set/Delete. Pipelining is also safe here because these
+	// are non-indexed reads (the indexed-DoMulti deadlock only affects writes).
+	ctx, cancel := s.ctx()
+	defer cancel()
+
+	cmds := make([]valkey.Completed, len(keys))
+	for i, k := range keys {
+		cmds[i] = s.client.B().Hget().Key(encodeKey(s.prefix, k)).Field(_valField).Build()
+	}
+
+	var foundKeys [][]float32
+	var foundValues [][]byte
+	for i, res := range s.client.DoMulti(ctx, cmds...) {
+		v, err := res.ToString()
+		if err != nil {
+			// A nil reply means the key/field is absent — omit it, don't error.
+			if valkey.IsValkeyNil(err) {
+				continue
+			}
+			return pb.StoresGetResult{}, fmt.Errorf("valkey-store: Get: HGET key %d: %w", i, err)
+		}
+		// The request vector is exact, so we return it verbatim as the key.
+		foundKeys = append(foundKeys, keys[i])
+		foundValues = append(foundValues, []byte(v))
+	}
+
+	return pb.StoresGetResult{
+		Keys:   store.WrapKeys(foundKeys),
+		Values: store.WrapValues(foundValues),
+	}, nil
+}
+
+// StoresDelete removes entries by exact vector. Missing keys are tolerated
+// (DEL returns 0), matching local-store.
+func (s *ValkeyStore) StoresDelete(opts *pb.StoresDeleteOptions) error {
+	keys := store.UnwrapKeys(opts.Keys)
+	if len(keys) == 0 {
+		return fmt.Errorf("valkey-store: Delete: no keys to delete")
+	}
+	if err := s.checkDims("Delete", keys); err != nil {
+		return err
+	}
+
+	// Sequential DELs for the same reason StoresSet avoids DoMulti: a DEL of an
+	// indexed key mutates the search index on the server's main thread, and a
+	// large pipeline of such mutations can deadlock the connection. Missing
+	// keys (DEL returns 0) are tolerated, matching local-store. As in Set, the
+	// timeout is per command so a large DeleteCols cannot exhaust one aggregate
+	// deadline mid-batch.
+	for i, k := range keys {
+		ctx, cancel := s.ctx()
+		err := s.client.Do(ctx, s.client.B().Del().Key(encodeKey(s.prefix, k)).Build()).Error()
+		cancel()
+		if err != nil {
+			return fmt.Errorf("valkey-store: Delete: DEL key %d: %w", i, err)
+		}
+	}
+	return nil
+}
+
+// StoresFind returns the topK nearest entries by the configured distance
+// metric, ordered most-similar first. An empty/uncreated index returns empty
+// slices and no error, matching local-store's empty-store behaviour.
+func (s *ValkeyStore) StoresFind(opts *pb.StoresFindOptions) (pb.StoresFindResult, error) {
+	// Guard against a malformed gRPC request with a nil/empty Key before
+	// dereferencing it — a nil opts.Key would otherwise panic the backend.
+	if opts.Key == nil || len(opts.Key.Floats) == 0 {
+		return pb.StoresFindResult{}, fmt.Errorf("valkey-store: Find: query key is empty")
+	}
+	query := opts.Key.Floats
+	topK := int(opts.TopK)
+	if topK < 1 {
+		return pb.StoresFindResult{}, fmt.Errorf("valkey-store: Find: topK = %d, must be >= 1", topK)
+	}
+	if topK > _maxTopK {
+		xlog.Warn("valkey-store: Find topK clamped", "requested", topK, "max", _maxTopK)
+		topK = _maxTopK
+	}
+	// No index yet means nothing has been Set (and none was found at Load) —
+	// an empty result, not an error.
+	if !s.indexCreated {
+		return pb.StoresFindResult{}, nil
+	}
+	// Enforce the query dimension against the known keyLen — recovered from
+	// FT.INFO at Load after a restart, or learned from the first Set — so a
+	// wrong-dimension query gets the clean local-store-style error. keyLen is
+	// only -1 in the degraded case where FT.INFO gave no parseable dimension;
+	// then we let Valkey's own FT.SEARCH validate the query vector.
+	if s.keyLen != -1 && len(query) != s.keyLen {
+		return pb.StoresFindResult{}, fmt.Errorf("valkey-store: Find: query length %d does not match existing %d", len(query), s.keyLen)
+	}
+
+	ctx, cancel := s.ctx()
+	defer cancel()
+
+	// KNN pre-filter query: match everything, rank by vector distance into the
+	// __score alias. A pure KNN query already returns its topK results ordered
+	// by distance ascending (nearest-first), so we do NOT add SORTBY __score:
+	// Valkey Search rejects sorting on the KNN score alias ("Index field
+	// `__score` does not exist" — it is a query-time computed field, not a
+	// SORTABLE schema attribute). LIMIT 0 topK caps the result and DIALECT 2 is
+	// required for the =>[KNN ...] vector syntax. The __score field is still
+	// returned in each document and read back for the similarity conversion.
+	//
+	// Injection-safety: the only caller-controlled value interpolated here is
+	// topK (an int, already bounded above). _vecField and _scoreField are
+	// compile-time constants, so this Sprintf cannot be used to inject query
+	// syntax. Do NOT make those fields operator-configurable without sanitizing
+	// them first — the KNN query string is otherwise built only from constants.
+	q := fmt.Sprintf("*=>[KNN %d @%s $q AS %s]", topK, _vecField, _scoreField)
+	cmd := s.client.B().FtSearch().Index(s.indexName).Query(q).
+		Return("3").Identifier(_vecField).Identifier(_valField).Identifier(_scoreField).
+		Limit().OffsetNum(0, int64(topK)).
+		Params().Nargs(2).NameValue().NameValue("q", valkey.VectorString32(query)).
+		Dialect(2).
+		Build()
+
+	_, docs, err := s.client.Do(ctx, cmd).AsFtSearch()
+	if err != nil {
+		// The cached indexCreated flag can go stale: an operator runs
+		// FT.DROPINDEX out of band, or two processes race on a fresh namespace.
+		// If the index is gone, mirror local-store's empty-store behaviour
+		// (empty result, no error) and clear the flag so a later Set recreates
+		// it, rather than surfacing a hard error for what looks like an empty
+		// store to the caller.
+		if isNoSuchIndexErr(err) {
+			s.indexCreated = false
+			return pb.StoresFindResult{}, nil
+		}
+		return pb.StoresFindResult{}, fmt.Errorf("valkey-store: Find: FT.SEARCH: %w", err)
+	}
+
+	keys := make([][]float32, 0, len(docs))
+	values := make([][]byte, 0, len(docs))
+	sims := make([]float32, 0, len(docs))
+	for _, doc := range docs {
+		// Decode the key from the returned `vec` bytes rather than the Valkey
+		// key string: this guarantees the exact original float ordering/values
+		// without a hex round-trip.
+		vecBytes := []byte(doc.Doc[_vecField])
+		k, err := bytesToVec(vecBytes)
+		if err != nil {
+			return pb.StoresFindResult{}, fmt.Errorf("valkey-store: Find: decode vec: %w", err)
+		}
+		dist, err := strconv.ParseFloat(doc.Doc[_scoreField], 64)
+		if err != nil {
+			return pb.StoresFindResult{}, fmt.Errorf("valkey-store: Find: parse score %q: %w", doc.Doc[_scoreField], err)
+		}
+		keys = append(keys, k)
+		values = append(values, []byte(doc.Doc[_valField]))
+		sims = append(sims, distanceToSimilarity(s.cfg.DistanceMetric, dist))
+	}
+
+	return pb.StoresFindResult{
+		Keys:         store.WrapKeys(keys),
+		Values:       store.WrapValues(values),
+		Similarities: sims,
+	}, nil
+}
+
+// ensureIndex creates the FT vector index once, lazily, on the first Set. The
+// dimension is fixed at creation (a second guard on top of the Go-side keyLen
+// check). An "already exists" error is treated as success so a restart against
+// a persisted index is a no-op.
+func (s *ValkeyStore) ensureIndex(dim int) error {
+	if s.indexCreated {
+		return nil
+	}
+
+	// VECTOR attribute tokens. The count that follows the algorithm name is the
+	// number of these tokens, so we build the slice and derive the count from
+	// it — no hand-maintained magic number that drifts when HNSW knobs change.
+	attrs := []string{"TYPE", "FLOAT32", "DIM", strconv.Itoa(dim), "DISTANCE_METRIC", s.cfg.DistanceMetric}
+	if s.cfg.IndexAlgo == indexAlgoHNSW {
+		attrs = append(attrs,
+			"M", strconv.Itoa(s.cfg.HNSW.M),
+			"EF_CONSTRUCTION", strconv.Itoa(s.cfg.HNSW.EFConstruction),
+			"EF_RUNTIME", strconv.Itoa(s.cfg.HNSW.EFRuntime),
+		)
+	}
+
+	args := []string{
+		s.indexName,
+		"ON", "HASH",
+		"PREFIX", "1", s.prefix,
+		"SCHEMA", _vecField, "VECTOR", s.cfg.IndexAlgo, strconv.Itoa(len(attrs)),
+	}
+	args = append(args, attrs...)
+
+	// FT.CREATE has no typed builder entry point, so we use the Arbitrary escape
+	// hatch. All tokens are non-key args in standalone mode.
+	ctx, cancel := s.ctx()
+	defer cancel()
+	err := s.client.Do(ctx, s.client.B().Arbitrary("FT.CREATE").Args(args...).Build()).Error()
+	if err != nil && !isIndexExistsErr(err) {
+		return fmt.Errorf("valkey-store: FT.CREATE %s: %w", s.indexName, err)
+	}
+
+	s.indexCreated = true
+	return nil
+}
+
+// checkDims rejects any key whose dimension disagrees with the learned keyLen.
+// When keyLen is still open (-1, nothing set yet) there is nothing to check.
+func (s *ValkeyStore) checkDims(op string, keys [][]float32) error {
+	if s.keyLen == -1 {
+		return nil
+	}
+	for i, k := range keys {
+		if len(k) != s.keyLen {
+			return fmt.Errorf("valkey-store: %s: key %d length %d does not match existing %d", op, i, len(k), s.keyLen)
+		}
+	}
+	return nil
+}
+
+// distanceToSimilarity converts a Valkey distance into local-store's similarity
+// convention. Only COSINE has a defined [-1, 1] similarity (sim = 1 - dist);
+// for L2/IP the raw score is returned as the "similarity" with a documented
+// meaning (smaller L2 = closer; larger IP = closer).
+func distanceToSimilarity(metric string, dist float64) float32 {
+	if metric == distanceCosine {
+		return float32(1 - dist)
+	}
+	return float32(dist)
+}
+
+// isIndexExistsErr reports whether an FT.CREATE error is the benign
+// "index already exists" case (e.g. after a restart against a persisted index).
+func isIndexExistsErr(err error) bool {
+	return strings.Contains(strings.ToLower(err.Error()), "already exists")
+}
+
+// isNoSuchIndexErr reports whether an FT.SEARCH error means the index no longer
+// exists (dropped out of band, or never really created despite a stale cached
+// flag). Valkey Search phrases this differently across versions, so we match
+// the common variants rather than one exact string.
+func isNoSuchIndexErr(err error) bool {
+	msg := strings.ToLower(err.Error())
+	if !strings.Contains(msg, "index") {
+		return false
+	}
+	return strings.Contains(msg, "no such index") ||
+		strings.Contains(msg, "not exist") ||
+		strings.Contains(msg, "not found") ||
+		strings.Contains(msg, "unknown index")
+}
+
+// keyPrefix / indexName derive per-namespace identifiers from the model name so
+// entries and indexes never collide across namespaces on a shared server.
+func keyPrefix(namespace string) string {
+	return _keyPrefixPrefix + nsToken(namespace) + ":"
+}
+
+func indexName(namespace string) string {
+	return _indexPrefix + nsToken(namespace)
+}
+
+// nsToken maps a namespace to a collision-resistant, printable token. sanitize()
+// alone is lossy (many distinct characters all fold to '_'), so namespaces like
+// "a b", "a/b" and "a:b" would otherwise share one keyspace and FT index — a
+// silent data-isolation bug (one store reading/clobbering another). We append a
+// short hash of the ORIGINAL namespace so distinct names never collide, while
+// the sanitized part keeps the token human-readable. It is deterministic, so a
+// persisted index is found again after a restart.
+func nsToken(namespace string) string {
+	sum := sha256.Sum256([]byte(namespace))
+	// Cap the human-readable part so a pathologically long model name can't
+	// produce an unbounded key prefix / index name (which would degrade Valkey
+	// performance). The 8-char hash suffix below already guarantees collision
+	// resistance regardless of truncation, so trimming the readable part is safe.
+	readable := sanitize(namespace)
+	if len(readable) > _maxNsTokenLen {
+		readable = readable[:_maxNsTokenLen]
+	}
+	return readable + "-" + hex.EncodeToString(sum[:])[:8]
+}
+
+// sanitize maps a namespace to a safe token for keys/index names: alphanumeric,
+// '_', '-' and '.' pass through; everything else becomes '_'. An empty
+// namespace becomes "default" so the key/index names stay well-formed.
+func sanitize(namespace string) string {
+	if namespace == "" {
+		return "default"
+	}
+	var b strings.Builder
+	b.Grow(len(namespace))
+	for _, r := range namespace {
+		switch {
+		case r >= 'a' && r <= 'z', r >= 'A' && r <= 'Z', r >= '0' && r <= '9', r == '_', r == '-', r == '.':
+			b.WriteRune(r)
+		default:
+			b.WriteRune('_')
+		}
+	}
+	return b.String()
+}
diff --git a/backend/go/valkey-store/store_suite_test.go b/backend/go/valkey-store/store_suite_test.go
new file mode 100644
index 000000000..d15f3fe2a
--- /dev/null
+++ b/backend/go/valkey-store/store_suite_test.go
@@ -0,0 +1,13 @@
+package main
+
+import (
+	"testing"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+func TestValkeyStore(t *testing.T) {
+	RegisterFailHandler(Fail)
+	RunSpecs(t, "valkey-store test suite")
+}
diff --git a/backend/go/valkey-store/store_test.go b/backend/go/valkey-store/store_test.go
new file mode 100644
index 000000000..c6d563d2b
--- /dev/null
+++ b/backend/go/valkey-store/store_test.go
@@ -0,0 +1,598 @@
+package main
+
+// Unit tests for the Valkey store, using the valkey-go gomock client so they
+// run with no container. They assert the exact commands built for each RPC
+// (the wire contract) plus the local-store parity semantics: empty/len/dim
+// rejects, omit-missing Get, tolerate-missing Delete, topK<1 reject, the
+// sim = 1 - distance conversion, lazy FT.CREATE, and the HNSW arg-shape.
+
+import (
+	"context"
+	"fmt"
+	"strconv"
+
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+	"github.com/mudler/LocalAI/pkg/store"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+	valkey "github.com/valkey-io/valkey-go"
+	"github.com/valkey-io/valkey-go/mock"
+	"go.uber.org/mock/gomock"
+)
+
+const testNamespace = "test"
+
+func testCfg() Config {
+	cfg, err := loadConfig(nil) // reads defaults when no options are set
+	Expect(err).NotTo(HaveOccurred())
+	return cfg
+}
+
+// opts builds a *pb.ModelOptions carrying the given key:value option strings,
+// mirroring how core threads a store's model-config `options:` list to the
+// backend's LoadModel.
+func opts(kv ...string) *pb.ModelOptions {
+	return &pb.ModelOptions{Options: kv}
+}
+
+func newMockStore(cfg Config) (*ValkeyStore, *mock.Client) {
+	ctrl := gomock.NewController(GinkgoT())
+	DeferCleanup(ctrl.Finish)
+	c := mock.NewClient(ctrl)
+	return newWithClient(c, cfg, testNamespace), c
+}
+
+func wrapSet(keys [][]float32, values [][]byte) *pb.StoresSetOptions {
+	return &pb.StoresSetOptions{Keys: store.WrapKeys(keys), Values: store.WrapValues(values)}
+}
+
+var _ = Describe("loadConfig", func() {
+	It("uses documented defaults", func() {
+		cfg, err := loadConfig(nil)
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.Addr).To(Equal("localhost:6379"))
+		Expect(cfg.ClientName).To(Equal("localai-valkey-store"))
+		Expect(cfg.IndexAlgo).To(Equal("FLAT"))
+		Expect(cfg.DistanceMetric).To(Equal("COSINE"))
+		Expect(cfg.RequestTimeout.Milliseconds()).To(Equal(int64(5000)))
+	})
+
+	It("honours option overrides", func() {
+		cfg, err := loadConfig(opts(
+			"addr:valkey.example:6380",
+			"index_algo:hnsw",
+			"distance_metric:l2",
+			"request_timeout_ms:1234",
+		))
+		Expect(err).NotTo(HaveOccurred())
+		// addr keeps its embedded colon: strings.Cut splits on the first ':'.
+		Expect(cfg.Addr).To(Equal("valkey.example:6380"))
+		Expect(cfg.IndexAlgo).To(Equal("HNSW"))
+		Expect(cfg.DistanceMetric).To(Equal("L2"))
+		Expect(cfg.RequestTimeout.Milliseconds()).To(Equal(int64(1234)))
+	})
+
+	It("keeps the mandatory client name when blanked", func() {
+		cfg, err := loadConfig(opts("client_name:"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.ClientName).To(Equal("localai-valkey-store"))
+	})
+
+	It("ignores a malformed option without a colon", func() {
+		cfg, err := loadConfig(opts("addr:valkey.example:6380", "not-a-kv-pair"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.Addr).To(Equal("valkey.example:6380"))
+	})
+
+	It("rejects an invalid index algo", func() {
+		_, err := loadConfig(opts("index_algo:bogus"))
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("rejects an invalid distance metric", func() {
+		_, err := loadConfig(opts("distance_metric:bogus"))
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("fails fast on a malformed HNSW integer instead of silently defaulting", func() {
+		_, err := loadConfig(opts("hnsw_m:1x6"))
+		Expect(err).To(HaveOccurred())
+		Expect(err.Error()).To(ContainSubstring("hnsw_m"))
+	})
+
+	It("honours a valid db override", func() {
+		cfg, err := loadConfig(opts("db:3"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.DB).To(Equal(3))
+	})
+
+	It("rejects a negative db", func() {
+		_, err := loadConfig(opts("db:-1"))
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("resolves username from direct option", func() {
+		cfg, err := loadConfig(opts("username:admin"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.Username).To(Equal("admin"))
+	})
+
+	It("resolves password from env indirection via password_env", func() {
+		GinkgoT().Setenv("TEST_VALKEY_PW_INDIRECT", "s3cret")
+		cfg, err := loadConfig(opts("password_env:TEST_VALKEY_PW_INDIRECT"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.Password).To(Equal("s3cret"))
+	})
+
+	It("resolves username from env indirection via username_env", func() {
+		GinkgoT().Setenv("TEST_VALKEY_USER_INDIRECT", "myuser")
+		cfg, err := loadConfig(opts("username_env:TEST_VALKEY_USER_INDIRECT"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.Username).To(Equal("myuser"))
+	})
+
+	It("prefers the direct option over env indirection", func() {
+		GinkgoT().Setenv("TEST_VALKEY_PW_CLASH", "from-env")
+		cfg, err := loadConfig(opts("password:direct-value", "password_env:TEST_VALKEY_PW_CLASH"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.Password).To(Equal("direct-value"))
+	})
+
+	It("returns empty when neither direct nor env indirection is set", func() {
+		cfg, err := loadConfig(opts("addr:localhost:6379"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(cfg.Username).To(BeEmpty())
+		Expect(cfg.Password).To(BeEmpty())
+	})
+})
+
+var _ = Describe("Load namespace gate", func() {
+	It("accepts prefixed store namespaces", func() {
+		s := NewValkeyStore()
+		// Load will fail at the Valkey connect step (no server), but only after
+		// passing the namespace gate. A network error is acceptable here — it
+		// means the prefix check passed.
+		err := s.Load(&pb.ModelOptions{Model: store.NamespacePrefix + "any-namespace", Options: []string{"addr:localhost:1"}})
+		Expect(err).NotTo(MatchError(ContainSubstring("not a store namespace")))
+	})
+
+	It("accepts the prefix alone (default store)", func() {
+		s := NewValkeyStore()
+		err := s.Load(&pb.ModelOptions{Model: store.NamespacePrefix, Options: []string{"addr:localhost:1"}})
+		Expect(err).NotTo(MatchError(ContainSubstring("not a store namespace")))
+	})
+
+	It("refuses model names without the namespace prefix", func() {
+		s := NewValkeyStore()
+		err := s.Load(&pb.ModelOptions{Model: "some-llm.gguf"})
+		Expect(err).To(MatchError(ContainSubstring("not a store namespace")))
+	})
+
+	It("refuses an empty model name", func() {
+		s := NewValkeyStore()
+		err := s.Load(&pb.ModelOptions{})
+		Expect(err).To(MatchError(ContainSubstring("not a store namespace")))
+	})
+
+	It("refuses nil opts", func() {
+		s := NewValkeyStore()
+		err := s.Load(nil)
+		Expect(err).To(HaveOccurred())
+	})
+})
+
+var _ = Describe("StoresSet", func() {
+	It("rejects empty input", func() {
+		s, _ := newMockStore(testCfg())
+		Expect(s.StoresSet(&pb.StoresSetOptions{})).NotTo(Succeed())
+	})
+
+	It("rejects key/value length mismatch", func() {
+		s, _ := newMockStore(testCfg())
+		err := s.StoresSet(wrapSet([][]float32{{1, 0, 0}}, [][]byte{[]byte("a"), []byte("b")}))
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("rejects dimension mismatch on a later add", func() {
+		s, c := newMockStore(testCfg())
+		// First Set issues FT.CREATE then a sequential HSET (both via Do).
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).DoAndReturn(
+			func(ctx context.Context, cmd valkey.Completed) valkey.ValkeyResult {
+				if cmd.Commands()[0] == "FT.CREATE" {
+					return assertFTCreate(3, "FLAT")(ctx, cmd)
+				}
+				return mock.Result(mock.ValkeyInt64(1)) // HSET
+			}).AnyTimes()
+		Expect(s.StoresSet(wrapSet([][]float32{{1, 0, 0}}, [][]byte{[]byte("3d")}))).To(Succeed())
+
+		err := s.StoresSet(wrapSet([][]float32{{1, 0}}, [][]byte{[]byte("2d")}))
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("rejects dimension mismatch within a batch", func() {
+		s, _ := newMockStore(testCfg())
+		err := s.StoresSet(wrapSet([][]float32{{1, 0, 0}, {1, 0}}, [][]byte{[]byte("3d"), []byte("2d")}))
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("creates the FLAT index once and HSETs each entry", func() {
+		s, c := newMockStore(testCfg())
+		// FT.CREATE must run exactly once (on the first Set); each entry is then
+		// written with an individual sequential HSET (Do, not DoMulti — see the
+		// pipeline-deadlock note in StoresSet).
+		var ftCreateCount, hsetCount int
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).DoAndReturn(
+			func(ctx context.Context, cmd valkey.Completed) valkey.ValkeyResult {
+				toks := cmd.Commands()
+				switch toks[0] {
+				case "FT.CREATE":
+					ftCreateCount++
+					return assertFTCreate(3, "FLAT")(ctx, cmd)
+				case "HSET":
+					hsetCount++
+					Expect(toks[1]).To(HavePrefix(s.prefix))
+					Expect(toks).To(ContainElements("vec", "val"))
+					return mock.Result(mock.ValkeyInt64(1))
+				default:
+					Fail("unexpected command: " + toks[0])
+					return valkey.ValkeyResult{}
+				}
+			}).AnyTimes()
+		Expect(s.StoresSet(wrapSet([][]float32{{1, 0, 0}}, [][]byte{[]byte("a")}))).To(Succeed())
+		Expect(s.StoresSet(wrapSet([][]float32{{2, 0, 0}}, [][]byte{[]byte("b")}))).To(Succeed())
+
+		Expect(ftCreateCount).To(Equal(1))
+		Expect(hsetCount).To(Equal(2))
+	})
+})
+
+var _ = Describe("StoresGet", func() {
+	It("round-trips values and omits missing keys", func() {
+		s, c := newMockStore(testCfg())
+		s.keyLen = 3
+		s.indexCreated = true
+		// First key present, second missing (nil reply).
+		c.EXPECT().DoMulti(gomock.Any(), gomock.Any(), gomock.Any()).Return([]valkey.ValkeyResult{
+			mock.Result(mock.ValkeyString("hello")),
+			mock.Result(mock.ValkeyNil()),
+		}).Times(1)
+
+		res, err := s.StoresGet(&pb.StoresGetOptions{
+			Keys: store.WrapKeys([][]float32{{1, 0, 0}, {9, 0, 0}}),
+		})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(res.Keys).To(HaveLen(1))
+		Expect(res.Values).To(HaveLen(1))
+		Expect(res.Values[0].Bytes).To(Equal([]byte("hello")))
+	})
+
+	It("rejects dimension mismatch", func() {
+		s, _ := newMockStore(testCfg())
+		s.keyLen = 3
+		_, err := s.StoresGet(&pb.StoresGetOptions{Keys: store.WrapKeys([][]float32{{1, 0}})})
+		Expect(err).To(HaveOccurred())
+	})
+})
+
+var _ = Describe("StoresDelete", func() {
+	It("issues DEL per key and tolerates missing", func() {
+		s, c := newMockStore(testCfg())
+		s.keyLen = 3
+		// DEL of a missing key returns 0 — still a success. DELs are issued
+		// sequentially (Do, not DoMulti — see the deadlock note in StoresSet).
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).DoAndReturn(
+			func(_ context.Context, cmd valkey.Completed) valkey.ValkeyResult {
+				Expect(cmd.Commands()[0]).To(Equal("DEL"))
+				return mock.Result(mock.ValkeyInt64(0))
+			}).Times(1)
+		Expect(s.StoresDelete(&pb.StoresDeleteOptions{
+			Keys: store.WrapKeys([][]float32{{9, 0, 0}}),
+		})).To(Succeed())
+	})
+
+	It("rejects dimension mismatch", func() {
+		s, _ := newMockStore(testCfg())
+		s.keyLen = 3
+		err := s.StoresDelete(&pb.StoresDeleteOptions{Keys: store.WrapKeys([][]float32{{1, 0}})})
+		Expect(err).To(HaveOccurred())
+	})
+})
+
+var _ = Describe("StoresFind", func() {
+	It("builds the KNN query and converts distance to similarity nearest-first", func() {
+		s, c := newMockStore(testCfg())
+		s.keyLen = 3
+		s.indexCreated = true
+
+		// Distances 0, 1, 2 must map to similarities 1, 0, -1 (COSINE).
+		docs := []ftDoc{
+			{vec: []float32{1, 0, 0}, val: "a", dist: 0},
+			{vec: []float32{0, 1, 0}, val: "b", dist: 1},
+			{vec: []float32{-1, 0, 0}, val: "c", dist: 2},
+		}
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).DoAndReturn(
+			func(_ context.Context, cmd valkey.Completed) valkey.ValkeyResult {
+				toks := cmd.Commands()
+				Expect(toks[0]).To(Equal("FT.SEARCH"))
+				Expect(toks[1]).To(Equal(s.indexName))
+				Expect(toks[2]).To(ContainSubstring("KNN 3 @vec $q AS __score"))
+				Expect(toks).To(ContainElements("PARAMS", "2", "q", "DIALECT", "2"))
+				return mock.Result(ftSearchReply(s.prefix, docs))
+			}).Times(1)
+
+		keys, values, sims, err := findViaRPC(s, []float32{1, 0, 0}, 3)
+		Expect(err).NotTo(HaveOccurred())
+		Expect(sims).To(Equal([]float32{1, 0, -1}))
+		Expect(values[0]).To(Equal([]byte("a")))
+		// Key decoded from returned vec bytes equals the original vector.
+		Expect(keys[0]).To(Equal([]float32{1, 0, 0}))
+	})
+
+	It("rejects topK < 1", func() {
+		s, _ := newMockStore(testCfg())
+		s.keyLen = 3
+		s.indexCreated = true
+		_, err := s.StoresFind(&pb.StoresFindOptions{Key: &pb.StoresKey{Floats: []float32{1, 0, 0}}, TopK: 0})
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("rejects a nil Key without panicking", func() {
+		s, _ := newMockStore(testCfg())
+		s.keyLen = 3
+		s.indexCreated = true
+		_, err := s.StoresFind(&pb.StoresFindOptions{TopK: 5})
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("rejects an empty query vector", func() {
+		s, _ := newMockStore(testCfg())
+		s.keyLen = 3
+		s.indexCreated = true
+		_, err := s.StoresFind(&pb.StoresFindOptions{Key: &pb.StoresKey{Floats: []float32{}}, TopK: 5})
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("rejects query dimension mismatch", func() {
+		s, _ := newMockStore(testCfg())
+		s.keyLen = 3
+		s.indexCreated = true
+		_, err := s.StoresFind(&pb.StoresFindOptions{Key: &pb.StoresKey{Floats: []float32{1, 0}}, TopK: 1})
+		Expect(err).To(HaveOccurred())
+	})
+
+	It("returns empty (no error) when the index was never created", func() {
+		s, _ := newMockStore(testCfg())
+		res, err := s.StoresFind(&pb.StoresFindOptions{Key: &pb.StoresKey{Floats: []float32{1, 0, 0}}, TopK: 5})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(res.Keys).To(BeEmpty())
+	})
+})
+
+var _ = Describe("ensureIndex arg-shape", func() {
+	It("emits HNSW tuning tokens when the algo is HNSW", func() {
+		cfg := testCfg()
+		cfg.IndexAlgo = indexAlgoHNSW
+		cfg.HNSW = hnswParams{M: 16, EFConstruction: 200, EFRuntime: 10}
+		s, c := newMockStore(cfg)
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).DoAndReturn(
+			func(_ context.Context, cmd valkey.Completed) valkey.ValkeyResult {
+				toks := cmd.Commands()
+				Expect(toks).To(ContainElements("HNSW", "M", "16", "EF_CONSTRUCTION", "200", "EF_RUNTIME", "10"))
+				return mock.Result(mock.ValkeyString("OK"))
+			}).Times(1)
+		Expect(s.ensureIndex(4)).To(Succeed())
+	})
+
+	It("treats an already-exists error as success", func() {
+		s, c := newMockStore(testCfg())
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).Return(
+			mock.ErrorResult(fmt.Errorf("Index already exists"))).Times(1)
+		Expect(s.ensureIndex(4)).To(Succeed())
+		Expect(s.indexCreated).To(BeTrue())
+	})
+})
+
+var _ = Describe("distanceToSimilarity", func() {
+	It("converts cosine distance to similarity", func() {
+		Expect(distanceToSimilarity(distanceCosine, 0)).To(Equal(float32(1)))
+		Expect(distanceToSimilarity(distanceCosine, 1)).To(Equal(float32(0)))
+		Expect(distanceToSimilarity(distanceCosine, 2)).To(Equal(float32(-1)))
+	})
+
+	It("passes the raw score through for non-cosine metrics", func() {
+		Expect(distanceToSimilarity(distanceL2, 0.42)).To(Equal(float32(0.42)))
+	})
+})
+
+var _ = Describe("findDimensions", func() {
+	It("recovers an integer dimension from a nested FT.INFO reply", func() {
+		dim, ok := findDimensions(ftInfoReply(mock.ValkeyInt64(768)))
+		Expect(ok).To(BeTrue())
+		Expect(dim).To(Equal(768))
+	})
+
+	It("recovers a string-encoded dimension", func() {
+		dim, ok := findDimensions(ftInfoReply(mock.ValkeyString("384")))
+		Expect(ok).To(BeTrue())
+		Expect(dim).To(Equal(384))
+	})
+
+	It("reports not-found when no dimension token is present", func() {
+		reply := mock.ValkeyArray(
+			mock.ValkeyString("index_name"), mock.ValkeyString("idx:test"),
+			mock.ValkeyString("num_docs"), mock.ValkeyInt64(0),
+		)
+		_, ok := findDimensions(reply)
+		Expect(ok).To(BeFalse())
+	})
+})
+
+var _ = Describe("loadIndexState", func() {
+	It("recovers indexCreated and keyLen from a persisted index", func() {
+		s, c := newMockStore(testCfg())
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).DoAndReturn(
+			func(_ context.Context, cmd valkey.Completed) valkey.ValkeyResult {
+				Expect(cmd.Commands()[0]).To(Equal("FT.INFO"))
+				return mock.Result(ftInfoReply(mock.ValkeyInt64(768)))
+			}).Times(1)
+
+		ctx, cancel := s.ctx()
+		defer cancel()
+		s.loadIndexState(ctx)
+		Expect(s.indexCreated).To(BeTrue())
+		Expect(s.keyLen).To(Equal(768))
+	})
+
+	It("leaves state untouched when the index does not exist", func() {
+		s, c := newMockStore(testCfg())
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).Return(
+			mock.ErrorResult(fmt.Errorf("Index with name 'idx:test' not found"))).Times(1)
+
+		ctx, cancel := s.ctx()
+		defer cancel()
+		s.loadIndexState(ctx)
+		Expect(s.indexCreated).To(BeFalse())
+		Expect(s.keyLen).To(Equal(-1))
+	})
+
+	It("marks the index created but leaves keyLen open when the dim is unparseable", func() {
+		s, c := newMockStore(testCfg())
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).Return(
+			mock.Result(mock.ValkeyArray(mock.ValkeyString("index_name"), mock.ValkeyString("idx:test")))).Times(1)
+
+		ctx, cancel := s.ctx()
+		defer cancel()
+		s.loadIndexState(ctx)
+		Expect(s.indexCreated).To(BeTrue())
+		Expect(s.keyLen).To(Equal(-1))
+	})
+})
+
+var _ = Describe("StoresFind on a dropped index", func() {
+	It("returns empty (no error) and clears the stale flag when the index is gone", func() {
+		s, c := newMockStore(testCfg())
+		s.keyLen = 3
+		s.indexCreated = true
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).Return(
+			mock.ErrorResult(fmt.Errorf("Index with name 'idx:test' not found"))).Times(1)
+
+		res, err := s.StoresFind(&pb.StoresFindOptions{Key: &pb.StoresKey{Floats: []float32{1, 0, 0}}, TopK: 5})
+		Expect(err).NotTo(HaveOccurred())
+		Expect(res.Keys).To(BeEmpty())
+		Expect(s.indexCreated).To(BeFalse())
+	})
+
+	It("still surfaces a genuine FT.SEARCH error", func() {
+		s, c := newMockStore(testCfg())
+		s.keyLen = 3
+		s.indexCreated = true
+		c.EXPECT().Do(gomock.Any(), gomock.Any()).Return(
+			mock.ErrorResult(fmt.Errorf("timeout"))).Times(1)
+
+		_, err := s.StoresFind(&pb.StoresFindOptions{Key: &pb.StoresKey{Floats: []float32{1, 0, 0}}, TopK: 5})
+		Expect(err).To(HaveOccurred())
+		Expect(s.indexCreated).To(BeTrue())
+	})
+})
+
+// --- test helpers ---
+
+type ftDoc struct {
+	vec  []float32
+	val  string
+	dist float64
+}
+
+func findViaRPC(s *ValkeyStore, query []float32, topK int) ([][]float32, [][]byte, []float32, error) {
+	res, err := s.StoresFind(&pb.StoresFindOptions{Key: &pb.StoresKey{Floats: query}, TopK: int32(topK)})
+	if err != nil {
+		return nil, nil, nil, err
+	}
+	return store.UnwrapKeys(res.Keys), store.UnwrapValues(res.Values), res.Similarities, nil
+}
+
+// assertFTCreate returns a DoAndReturn func that verifies the FT.CREATE command
+// carries the expected dimension and algorithm, then replies OK.
+func assertFTCreate(dim int, algo string) func(context.Context, valkey.Completed) valkey.ValkeyResult {
+	return func(_ context.Context, cmd valkey.Completed) valkey.ValkeyResult {
+		toks := cmd.Commands()
+		Expect(toks[0]).To(Equal("FT.CREATE"))
+		Expect(toks).To(ContainElements("VECTOR", algo, "TYPE", "FLOAT32", "DIM", strconv.Itoa(dim), "DISTANCE_METRIC", "COSINE"))
+		return mock.Result(mock.ValkeyString("OK"))
+	}
+}
+
+// ftInfoReply builds a RESP2-shaped FT.INFO reply that mirrors Valkey Search's
+// nesting: the VECTOR attribute carries its params under an `index` array whose
+// `dimensions` key holds the DIM. dimValue is the message the parser must read
+// back (integer or string-encoded), so both wire shapes can be exercised.
+func ftInfoReply(dimValue valkey.ValkeyMessage) valkey.ValkeyMessage {
+	vectorAttr := mock.ValkeyArray(
+		mock.ValkeyString("identifier"), mock.ValkeyString(_vecField),
+		mock.ValkeyString("attribute"), mock.ValkeyString(_vecField),
+		mock.ValkeyString("type"), mock.ValkeyString("VECTOR"),
+		mock.ValkeyString("index"), mock.ValkeyArray(
+			mock.ValkeyString("capacity"), mock.ValkeyInt64(1000),
+			mock.ValkeyString("dimensions"), dimValue,
+			mock.ValkeyString("distance_metric"), mock.ValkeyString("COSINE"),
+			mock.ValkeyString("data_type"), mock.ValkeyString("FLOAT32"),
+		),
+	)
+	return mock.ValkeyArray(
+		mock.ValkeyString("index_name"), mock.ValkeyString("idx:test"),
+		mock.ValkeyString("attributes"), mock.ValkeyArray(vectorAttr),
+		mock.ValkeyString("num_docs"), mock.ValkeyInt64(0),
+	)
+}
+
+// ftSearchReply builds a RESP2-shaped FT.SEARCH reply: [total, key, attrs, ...]
+// where attrs carries the returned vec/val/__score fields.
+func ftSearchReply(prefix string, docs []ftDoc) valkey.ValkeyMessage {
+	arr := []valkey.ValkeyMessage{mock.ValkeyInt64(int64(len(docs)))}
+	for _, d := range docs {
+		arr = append(arr, mock.ValkeyString(encodeKey(prefix, d.vec)))
+		attrs := mock.ValkeyArray(
+			mock.ValkeyString(_vecField), mock.ValkeyString(valkey.BinaryString(vecToBytes(d.vec))),
+			mock.ValkeyString(_valField), mock.ValkeyString(d.val),
+			mock.ValkeyString(_scoreField), mock.ValkeyString(strconv.FormatFloat(d.dist, 'f', -1, 64)),
+		)
+		arr = append(arr, attrs)
+	}
+	return mock.ValkeyArray(arr...)
+}
+
+var _ = Describe("namespace token", func() {
+	It("is stable for the same namespace (so a persisted index is found again)", func() {
+		Expect(nsToken("faces")).To(Equal(nsToken("faces")))
+	})
+
+	It("does not collide for namespaces that sanitize to the same token", func() {
+		// "a b", "a/b" and "a:b" all sanitize to "a_b"; the hash suffix must
+		// keep them distinct so two logically-distinct stores never share one
+		// keyspace/index (the data-isolation guarantee).
+		Expect(sanitize("a b")).To(Equal(sanitize("a/b")))
+		Expect(nsToken("a b")).NotTo(Equal(nsToken("a/b")))
+		Expect(nsToken("a/b")).NotTo(Equal(nsToken("a:b")))
+	})
+
+	It("keeps the sanitized part human-readable", func() {
+		Expect(nsToken("faces")).To(HavePrefix("faces-"))
+	})
+
+	It("maps an empty namespace to a stable default token", func() {
+		Expect(nsToken("")).To(HavePrefix("default-"))
+		Expect(nsToken("")).To(Equal(nsToken("")))
+	})
+})
+
+var _ = Describe("sanitize", func() {
+	It("passes through allowed runes and folds the rest to '_'", func() {
+		Expect(sanitize("Ok_9.-")).To(Equal("Ok_9.-"))
+		Expect(sanitize("a b/c:d")).To(Equal("a_b_c_d"))
+	})
+
+	It("maps empty to 'default'", func() {
+		Expect(sanitize("")).To(Equal("default"))
+	})
+})
diff --git a/backend/index.yaml b/backend/index.yaml
index 2d002d690..b447d8682 100644
--- a/backend/index.yaml
+++ b/backend/index.yaml
@@ -425,6 +425,32 @@
     nvidia-cuda-12: "cuda12-stablediffusion-ggml"
     nvidia-l4t-cuda-12: "nvidia-l4t-arm64-stablediffusion-ggml"
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-stablediffusion-ggml"
+- &trellis2cpp
+  name: "trellis2cpp"
+  alias: "trellis2cpp"
+  license: mit
+  description: |
+     TRELLIS.2 image-to-3D generation (GLB meshes with PBR textures) in C++/ggml
+  urls:
+    - https://github.com/localai-org/trellis2cpp
+    - https://github.com/microsoft/TRELLIS.2
+  tags:
+    - image-to-3d
+    - 3d-generation
+    - CPU
+    - GPU
+    - CUDA
+    - Metal
+  capabilities:
+    default: "cpu-trellis2cpp"
+    nvidia: "cuda12-trellis2cpp"
+    vulkan: "vulkan-trellis2cpp"
+    nvidia-l4t: "nvidia-l4t-arm64-trellis2cpp"
+    metal: "metal-trellis2cpp"
+    nvidia-cuda-13: "cuda13-trellis2cpp"
+    nvidia-cuda-12: "cuda12-trellis2cpp"
+    nvidia-l4t-cuda-12: "nvidia-l4t-arm64-trellis2cpp"
+    nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-trellis2cpp"
 - &rfdetr
   name: "rfdetr"
   alias: "rfdetr"
@@ -1837,6 +1863,23 @@
   capabilities:
     default: "cpu-cloud-proxy"
     metal: "metal-cloud-proxy"
+- &valkey-store
+  name: "valkey-store"
+  urls:
+    - https://github.com/mudler/LocalAI
+  description: |
+    Valkey Store is a Valkey Search (FT.*) backed vector store for LocalAI. It
+    persists vectors across restarts and supports opt-in HNSW indexing. Requires
+    a reachable Valkey Search server (valkey/valkey-bundle).
+  tags:
+    - vector-database
+    - valkey
+    - open-source
+    - CPU
+  license: MIT
+  capabilities:
+    default: "cpu-valkey-store"
+    metal: "metal-valkey-store"
 - &kitten-tts
   name: "kitten-tts"
   urls:
@@ -1972,6 +2015,18 @@
     nvidia-cuda-12: "cuda12-stablediffusion-ggml-development"
     nvidia-l4t-cuda-12: "nvidia-l4t-arm64-stablediffusion-ggml-development"
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-stablediffusion-ggml-development"
+- !!merge <<: *trellis2cpp
+  name: "trellis2cpp-development"
+  capabilities:
+    default: "cpu-trellis2cpp-development"
+    nvidia: "cuda12-trellis2cpp-development"
+    vulkan: "vulkan-trellis2cpp-development"
+    nvidia-l4t: "nvidia-l4t-arm64-trellis2cpp-development"
+    metal: "metal-trellis2cpp-development"
+    nvidia-cuda-13: "cuda13-trellis2cpp-development"
+    nvidia-cuda-12: "cuda12-trellis2cpp-development"
+    nvidia-l4t-cuda-12: "nvidia-l4t-arm64-trellis2cpp-development"
+    nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-trellis2cpp-development"
 - !!merge <<: *neutts
   name: "cpu-neutts"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-neutts"
@@ -2369,6 +2424,35 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-cloud-proxy"
   mirrors:
     - localai/localai-backends:master-metal-darwin-arm64-cloud-proxy
+- !!merge <<: *valkey-store
+  name: "cpu-valkey-store"
+  alias: "valkey-store"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-valkey-store"
+  mirrors:
+    - localai/localai-backends:latest-cpu-valkey-store
+- !!merge <<: *valkey-store
+  name: "cpu-valkey-store-development"
+  alias: "valkey-store"
+  uri: "quay.io/go-skynet/local-ai-backends:master-cpu-valkey-store"
+  mirrors:
+    - localai/localai-backends:master-cpu-valkey-store
+- !!merge <<: *valkey-store
+  name: "valkey-store-development"
+  alias: "valkey-store"
+  capabilities:
+    default: "cpu-valkey-store-development"
+    metal: "metal-valkey-store-development"
+- !!merge <<: *valkey-store
+  name: "metal-valkey-store"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-valkey-store"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-valkey-store
+- !!merge <<: *valkey-store
+  name: "metal-valkey-store-development"
+  alias: "valkey-store"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-valkey-store"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-valkey-store
 - !!merge <<: *opus
   name: "cpu-opus"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-opus"
@@ -3688,6 +3772,77 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-13-stablediffusion-ggml"
   mirrors:
     - localai/localai-backends:master-gpu-nvidia-cuda-13-stablediffusion-ggml
+## trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "cpu-trellis2cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:latest-cpu-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "cpu-trellis2cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-cpu-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:master-cpu-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "metal-trellis2cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "metal-trellis2cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "vulkan-trellis2cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-vulkan-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:latest-gpu-vulkan-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "vulkan-trellis2cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-vulkan-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:master-gpu-vulkan-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "cuda12-trellis2cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:latest-gpu-nvidia-cuda-12-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "cuda12-trellis2cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:master-gpu-nvidia-cuda-12-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "cuda13-trellis2cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-13-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:latest-gpu-nvidia-cuda-13-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "cuda13-trellis2cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-13-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:master-gpu-nvidia-cuda-13-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "nvidia-l4t-arm64-trellis2cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-arm64-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:latest-nvidia-l4t-arm64-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "nvidia-l4t-arm64-trellis2cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-arm64-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:master-nvidia-l4t-arm64-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "cuda13-nvidia-l4t-arm64-trellis2cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-cuda-13-arm64-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:latest-nvidia-l4t-cuda-13-arm64-trellis2cpp
+- !!merge <<: *trellis2cpp
+  name: "cuda13-nvidia-l4t-arm64-trellis2cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-cuda-13-arm64-trellis2cpp"
+  mirrors:
+    - localai/localai-backends:master-nvidia-l4t-cuda-13-arm64-trellis2cpp
 ## privacy-filter
 - !!merge <<: *privacyfilter
   name: "cpu-privacy-filter"
diff --git a/backend/rust/kokoros/src/service.rs b/backend/rust/kokoros/src/service.rs
index 495b0a423..80fd3ceb8 100644
--- a/backend/rust/kokoros/src/service.rs
+++ b/backend/rust/kokoros/src/service.rs
@@ -334,6 +334,13 @@ impl Backend for KokorosService {
         Err(Status::unimplemented("Not supported"))
     }
 
+    async fn generate3_d(
+        &self,
+        _: Request<backend::Generate3DRequest>,
+    ) -> Result<Response<backend::Result>, Status> {
+        Err(Status::unimplemented("Not supported"))
+    }
+
     async fn audio_transcription(
         &self,
         _: Request<backend::TranscriptRequest>,
diff --git a/core/application/application.go b/core/application/application.go
index df904c02d..c5853a925 100644
--- a/core/application/application.go
+++ b/core/application/application.go
@@ -165,7 +165,7 @@ func newApplication(appConfig *config.ApplicationConfig) *Application {
 		voiceStoreName = "localai-voice-biometrics"
 	)
 	faceStoreResolver := func(_ context.Context, storeName string) (pkggrpc.Backend, error) {
-		return corebackend.StoreBackend(ml, appConfig, storeName, "")
+		return corebackend.StoreBackend(ml, appConfig, app.backendLoader, storeName, "")
 	}
 	app.faceRegistry = facerecognition.NewStoreRegistry(faceStoreResolver, faceStoreName, faceEmbeddingDim)
 
@@ -173,7 +173,7 @@ func newApplication(appConfig *config.ApplicationConfig) *Application {
 	// namespace so embedding spaces stay isolated (a face vector and a
 	// speaker vector are not comparable and differ in dimensionality).
 	voiceStoreResolver := func(_ context.Context, storeName string) (pkggrpc.Backend, error) {
-		return corebackend.StoreBackend(ml, appConfig, storeName, "")
+		return corebackend.StoreBackend(ml, appConfig, app.backendLoader, storeName, "")
 	}
 	app.voiceRegistry = voicerecognition.NewStoreRegistry(voiceStoreResolver, voiceStoreName, voiceEmbeddingDim)
 
diff --git a/core/application/router_factories.go b/core/application/router_factories.go
index 2bee523ea..523c4465e 100644
--- a/core/application/router_factories.go
+++ b/core/application/router_factories.go
@@ -116,5 +116,5 @@ func (l *lazyEmbedder) Embed(ctx context.Context, text string) ([]float32, error
 // VectorStore takes a store name, not a model name — no adapterConfig, no
 // staleness to avoid.
 func (a *Application) VectorStore(storeName string) backend.VectorStore {
-	return backend.NewVectorStore(a.modelLoader, a.applicationConfig, storeName)
+	return backend.NewVectorStore(a.modelLoader, a.applicationConfig, a.backendLoader, storeName)
 }
diff --git a/core/backend/model3d.go b/core/backend/model3d.go
new file mode 100644
index 000000000..1952501d9
--- /dev/null
+++ b/core/backend/model3d.go
@@ -0,0 +1,104 @@
+package backend
+
+import (
+	"maps"
+	"time"
+
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/trace"
+	"github.com/mudler/LocalAI/pkg/grpc/proto"
+	model "github.com/mudler/LocalAI/pkg/model"
+)
+
+// Model3DGenerationOptions is the backend-neutral request passed to 3D
+// generators. Image contains a staged local path by the time it reaches
+// this layer.
+type Model3DGenerationOptions struct {
+	Image        string
+	Destination  string
+	Seed         int32
+	Step         int32
+	CFGScale     float32
+	TextureSteps int32
+	Quality      string
+	Background   string
+	Params       map[string]string
+}
+
+func Model3DGeneration(options Model3DGenerationOptions, loader *model.ModelLoader, modelConfig config.ModelConfig, appConfig *config.ApplicationConfig) (func() error, error) {
+	opts := ModelOptions(modelConfig, appConfig)
+	inferenceModel, err := loader.Load(opts...)
+	if err != nil {
+		recordModelLoadFailure(appConfig, modelConfig.Name, modelConfig.Backend, err, nil)
+		return nil, err
+	}
+
+	fn := func() error {
+		_, err := inferenceModel.Generate3D(
+			appConfig.Context,
+			&proto.Generate3DRequest{
+				Src:          options.Image,
+				Dst:          options.Destination,
+				Seed:         options.Seed,
+				Step:         options.Step,
+				CfgScale:     options.CFGScale,
+				TextureSteps: options.TextureSteps,
+				Quality:      options.Quality,
+				Background:   options.Background,
+				Params:       maps.Clone(options.Params),
+			},
+		)
+		return err
+	}
+
+	if appConfig.EnableTracing {
+		trace.InitBackendTracingIfEnabled(appConfig.TracingMaxItems, appConfig.TracingMaxBodyBytes)
+
+		traceType := trace.BackendTrace3DGeneration
+		traceSummary := "3d: " + options.Quality
+		traceData := map[string]any{}
+		if options.Params["operation"] == "print_remesh" {
+			traceType = trace.BackendTrace3DRemesh
+			traceSummary = "3d: remesh"
+			traceData["detail_percent"] = options.Params["detail_percent"]
+			traceData["has_mesh"] = options.Image != ""
+		} else {
+			traceData = map[string]any{
+				"seed":          options.Seed,
+				"step":          options.Step,
+				"cfg_scale":     options.CFGScale,
+				"texture_steps": options.TextureSteps,
+				"quality":       options.Quality,
+				"background":    options.Background,
+				"has_image":     options.Image != "",
+			}
+		}
+
+		startTime := time.Now()
+		originalFn := fn
+		fn = func() error {
+			err := originalFn()
+			duration := time.Since(startTime)
+
+			errStr := ""
+			if err != nil {
+				errStr = err.Error()
+			}
+
+			trace.RecordBackendTrace(trace.BackendTrace{
+				Timestamp: startTime,
+				Duration:  duration,
+				Type:      traceType,
+				ModelName: modelConfig.Name,
+				Backend:   modelConfig.Backend,
+				Summary:   trace.TruncateString(traceSummary, 200),
+				Error:     errStr,
+				Data:      traceData,
+			})
+
+			return err
+		}
+	}
+
+	return fn, nil
+}
diff --git a/core/backend/stores.go b/core/backend/stores.go
index 480400f42..1c324ab1a 100644
--- a/core/backend/stores.go
+++ b/core/backend/stores.go
@@ -9,6 +9,7 @@ import (
 	"github.com/mudler/LocalAI/core/trace"
 
 	"github.com/mudler/LocalAI/pkg/grpc"
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
 	"github.com/mudler/LocalAI/pkg/model"
 	"github.com/mudler/LocalAI/pkg/store"
 )
@@ -23,21 +24,25 @@ type VectorStore interface {
 
 // NewVectorStore returns a VectorStore backed by the local-store
 // gRPC backend, namespaced by storeName so two routers don't collide.
-func NewVectorStore(loader *model.ModelLoader, appConfig *config.ApplicationConfig, storeName string) VectorStore {
+// cl resolves the per-store model config (backend + options); it may be nil,
+// in which case the store falls back to the default backend and its built-in
+// defaults.
+func NewVectorStore(loader *model.ModelLoader, appConfig *config.ApplicationConfig, cl *config.ModelConfigLoader, storeName string) VectorStore {
 	if storeName == "" {
 		return nil
 	}
-	return &localVectorStore{loader: loader, appConfig: appConfig, storeName: storeName}
+	return &localVectorStore{loader: loader, appConfig: appConfig, cl: cl, storeName: storeName}
 }
 
 type localVectorStore struct {
 	loader    *model.ModelLoader
 	appConfig *config.ApplicationConfig
+	cl        *config.ModelConfigLoader
 	storeName string
 }
 
 func (s *localVectorStore) backend(_ context.Context) (grpc.Backend, error) {
-	return StoreBackend(s.loader, s.appConfig, s.storeName, "")
+	return StoreBackend(s.loader, s.appConfig, s.cl, s.storeName, "")
 }
 
 func (s *localVectorStore) Search(ctx context.Context, vec []float32) (sim float64, payload []byte, ok bool, err error) {
@@ -121,7 +126,24 @@ func (s *localVectorStore) recordTrace(start time.Time, op string, vecDim int, s
 	})
 }
 
-func StoreBackend(sl *model.ModelLoader, appConfig *config.ApplicationConfig, storeName string, backend string) (grpc.Backend, error) {
+func StoreBackend(sl *model.ModelLoader, appConfig *config.ApplicationConfig, cl *config.ModelConfigLoader, storeName string, backend string) (grpc.Backend, error) {
+	// Resolve the per-store model config (keyed by the store namespace, which
+	// is the model ID for a store). This is the LocalAI-native config surface:
+	// a store's backend selection and its backend-specific settings live in a
+	// model YAML's `backend:` and `options:` fields, so different stores can
+	// point at different servers/indexes. When no config exists for the store,
+	// we fall back to the default backend and let the backend apply its own
+	// built-in defaults — preserving the zero-config experience.
+	var loadOpts []string
+	if cl != nil {
+		if cfg, ok := cl.GetModelConfig(storeName); ok {
+			if backend == "" {
+				backend = cfg.Backend
+			}
+			loadOpts = cfg.Options
+		}
+	}
+
 	if backend == "" {
 		backend = model.LocalStoreBackend
 	}
@@ -145,5 +167,12 @@ func StoreBackend(sl *model.ModelLoader, appConfig *config.ApplicationConfig, st
 		model.WithModel(store.NamespacePrefix + storeName),
 	}
 
+	// Thread the store's configured options through to the backend's LoadModel
+	// via ModelOptions.Options (field 62). The loader clones these opts and
+	// overrides only Model/ModelFile, so the namespace set above is preserved.
+	if len(loadOpts) > 0 {
+		sc = append(sc, model.WithLoadGRPCLoadModelOpts(&pb.ModelOptions{Options: loadOpts}))
+	}
+
 	return sl.Load(sc...)
 }
diff --git a/core/config/backend_capabilities.go b/core/config/backend_capabilities.go
index ad8b4d622..3de119bbb 100644
--- a/core/config/backend_capabilities.go
+++ b/core/config/backend_capabilities.go
@@ -16,6 +16,7 @@ const (
 	UsecaseTokenize            = "tokenize"
 	UsecaseImage               = "image"
 	UsecaseVideo               = "video"
+	Usecase3D                  = "3d"
 	UsecaseTranscript          = "transcript"
 	UsecaseTTS                 = "tts"
 	UsecaseSoundGeneration     = "sound_generation"
@@ -42,6 +43,7 @@ const (
 	MethodEmbedding          GRPCMethod = "Embedding"
 	MethodGenerateImage      GRPCMethod = "GenerateImage"
 	MethodGenerateVideo      GRPCMethod = "GenerateVideo"
+	MethodGenerate3D         GRPCMethod = "Generate3D"
 	MethodAudioTranscription GRPCMethod = "AudioTranscription"
 	MethodTTS                GRPCMethod = "TTS"
 	MethodTTSStream          GRPCMethod = "TTSStream"
@@ -124,6 +126,11 @@ var UsecaseInfoMap = map[string]UsecaseInfo{
 		GRPCMethod:  MethodGenerateVideo,
 		Description: "Video generation via the GenerateVideo RPC, with optional image or audio conditioning when supported by the backend.",
 	},
+	Usecase3D: {
+		Flag:        FLAG_3D,
+		GRPCMethod:  MethodGenerate3D,
+		Description: "Image-conditioned 3D asset generation via the Generate3D RPC — a binary glTF (GLB) mesh with optional PBR material (TRELLIS.2).",
+	},
 	UsecaseTranscript: {
 		Flag:        FLAG_TRANSCRIPT,
 		GRPCMethod:  MethodAudioTranscription,
@@ -351,6 +358,14 @@ var BackendCapabilities = map[string]BackendCapability{
 		Description:      "Stable Diffusion via GGML quantized models",
 	},
 
+	// --- 3D generation backends ---
+	"trellis2cpp": {
+		GRPCMethods:      []GRPCMethod{MethodGenerate3D},
+		PossibleUsecases: []string{Usecase3D},
+		DefaultUsecases:  []string{Usecase3D},
+		Description:      "trellis2.cpp — C++/GGML port of Microsoft TRELLIS.2: single-image to textured 3D mesh (GLB)",
+	},
+
 	// --- Speech-to-text backends ---
 	"whisper": {
 		GRPCMethods:      []GRPCMethod{MethodAudioTranscription, MethodVAD},
diff --git a/core/config/model_capabilities.go b/core/config/model_capabilities.go
index 79a117b4a..6b432bddc 100644
--- a/core/config/model_capabilities.go
+++ b/core/config/model_capabilities.go
@@ -15,9 +15,10 @@ const (
 	ModalityImage = "image"
 	ModalityAudio = "audio"
 	ModalityVideo = "video"
+	Modality3D    = "3d"
 )
 
-var modalityOrder = []string{ModalityText, ModalityImage, ModalityAudio, ModalityVideo}
+var modalityOrder = []string{ModalityText, ModalityImage, ModalityAudio, ModalityVideo, Modality3D}
 
 func declaredModalities(modalities []string) map[string]bool {
 	declared := make(map[string]bool, len(modalities))
@@ -154,6 +155,7 @@ func (c *ModelConfig) Capabilities() []string {
 	add(c.HasUsecases(FLAG_SOUND_GENERATION), UsecaseSoundGeneration)
 	add(c.HasUsecases(FLAG_IMAGE), UsecaseImage)
 	add(c.HasUsecases(FLAG_VIDEO), UsecaseVideo)
+	add(c.HasUsecases(FLAG_3D), Usecase3D)
 	add(c.HasUsecases(FLAG_VAD), UsecaseVAD)
 	add(c.HasUsecases(FLAG_DETECTION), UsecaseDetection)
 	add(c.HasUsecases(FLAG_DEPTH), UsecaseDepth)
@@ -181,9 +183,10 @@ func (c *ModelConfig) InputModalities() []string {
 		c.HasUsecases(FLAG_TTS) || c.HasUsecases(FLAG_SOUND_GENERATION) || imageGen || videoGen
 
 	// Image input via a chat model requires vision (gated on chat, like the
-	// Ollama surface); detection/depth/face models consume images directly.
+	// Ollama surface); detection/depth/face/3D models consume images directly.
 	imageIn := (chatish && c.VisionSupported()) || c.LimitMMPerPrompt.LimitImagePerPrompt > 0 ||
-		c.HasUsecases(FLAG_DETECTION) || c.HasUsecases(FLAG_DEPTH) || c.HasUsecases(FLAG_FACE_RECOGNITION)
+		c.HasUsecases(FLAG_DETECTION) || c.HasUsecases(FLAG_DEPTH) || c.HasUsecases(FLAG_FACE_RECOGNITION) ||
+		c.HasUsecases(FLAG_3D)
 
 	audioIn := c.AudioInputSupported() || c.HasUsecases(FLAG_TRANSCRIPT) || c.HasUsecases(FLAG_AUDIO_TRANSFORM) ||
 		c.HasUsecases(FLAG_REALTIME_AUDIO) || c.HasUsecases(FLAG_VAD) || c.HasUsecases(FLAG_DIARIZATION) ||
@@ -208,10 +211,12 @@ func (c *ModelConfig) OutputModalities() []string {
 	audioOut := c.HasUsecases(FLAG_TTS) || c.HasUsecases(FLAG_SOUND_GENERATION) ||
 		c.HasUsecases(FLAG_AUDIO_TRANSFORM) || c.HasUsecases(FLAG_REALTIME_AUDIO)
 	videoOut := c.HasUsecases(FLAG_VIDEO)
+	threeDOut := c.HasUsecases(FLAG_3D)
 
 	modalities[ModalityText] = modalities[ModalityText] || textOut
 	modalities[ModalityImage] = modalities[ModalityImage] || imageOut
 	modalities[ModalityAudio] = modalities[ModalityAudio] || audioOut
 	modalities[ModalityVideo] = modalities[ModalityVideo] || videoOut
+	modalities[Modality3D] = modalities[Modality3D] || threeDOut
 	return orderedModalities(modalities)
 }
diff --git a/core/config/model_capabilities_test.go b/core/config/model_capabilities_test.go
index 545bad50b..84f01af7b 100644
--- a/core/config/model_capabilities_test.go
+++ b/core/config/model_capabilities_test.go
@@ -109,6 +109,25 @@ var _ = Describe("Model capabilities derivation", func() {
 			Expect(cfg.OutputModalities()).To(Equal([]string{"image"}))
 		})
 
+		It("guesses the 3d usecase from the trellis2cpp backend and only that backend", func() {
+			cfg := &ModelConfig{Backend: "trellis2cpp"}
+			Expect(cfg.HasUsecases(FLAG_3D)).To(BeTrue())
+			Expect(cfg.Capabilities()).To(ContainElement(Usecase3D))
+
+			other := &ModelConfig{Backend: "llama-cpp"}
+			Expect(other.HasUsecases(FLAG_3D)).To(BeFalse())
+		})
+
+		It("a 3D-generation model reads an image and writes a 3D asset", func() {
+			// Pins the wire strings the UI depends on: capability "3d",
+			// input modality "image" (no text prompt — TRELLIS.2 is
+			// image-conditioned only), output modality "3d".
+			cfg := &ModelConfig{KnownUsecases: usecaseBits(FLAG_3D), Backend: "trellis2cpp"}
+			Expect(cfg.Capabilities()).To(Equal([]string{Usecase3D}))
+			Expect(cfg.InputModalities()).To(Equal([]string{ModalityImage}))
+			Expect(cfg.OutputModalities()).To(Equal([]string{Modality3D}))
+		})
+
 		It("conditioned video uses declared modalities without backend-specific inference", func() {
 			cfg := &ModelConfig{
 				KnownUsecases:         usecaseBits(FLAG_VIDEO),
diff --git a/core/config/model_config.go b/core/config/model_config.go
index 3a32ac2c8..21bc42fac 100644
--- a/core/config/model_config.go
+++ b/core/config/model_config.go
@@ -1673,6 +1673,11 @@ const (
 	// labels via the SoundDetection RPC, e.g. ced).
 	FLAG_SOUND_CLASSIFICATION ModelConfigUsecase = 0b10000000000000000000000
 
+	// Marks a model as wired for the Generate3D gRPC primitive
+	// (image-conditioned 3D asset generation — a binary glTF mesh with
+	// optional PBR material, e.g. trellis2cpp).
+	FLAG_3D ModelConfigUsecase = 0b100000000000000000000000
+
 	// Common Subsets
 	FLAG_LLM ModelConfigUsecase = FLAG_CHAT | FLAG_COMPLETION | FLAG_EDIT
 )
@@ -1686,7 +1691,7 @@ var ModalityGroups = []ModelConfigUsecase{
 	FLAG_TRANSCRIPT | FLAG_REALTIME_AUDIO | FLAG_SOUND_CLASSIFICATION, // audio input — realtime_audio is any-to-any, so it counts here too
 	FLAG_TTS | FLAG_SOUND_GENERATION | FLAG_REALTIME_AUDIO,            // audio output — and here, so a lone realtime_audio flag still reads as multimodal
 	FLAG_AUDIO_TRANSFORM,                                              // audio in/out transforms
-	FLAG_IMAGE | FLAG_VIDEO,                                           // visual generation
+	FLAG_IMAGE | FLAG_VIDEO | FLAG_3D,                                 // visual generation
 }
 
 // IsMultimodal returns true if the given usecases span two or more orthogonal
@@ -1733,6 +1738,7 @@ func GetAllModelConfigUsecases() map[string]ModelConfigUsecase {
 		"FLAG_SCORE":                FLAG_SCORE,
 		"FLAG_DEPTH":                FLAG_DEPTH,
 		"FLAG_TOKEN_CLASSIFY":       FLAG_TOKEN_CLASSIFY,
+		"FLAG_3D":                   FLAG_3D,
 	}
 }
 
@@ -1884,6 +1890,13 @@ func (c *ModelConfig) GuessUsecases(u ModelConfigUsecase) bool {
 		}
 	}
 
+	if (u & FLAG_3D) == FLAG_3D {
+		threeDBackends := []string{"trellis2cpp"}
+		if !slices.Contains(threeDBackends, c.Backend) {
+			return false
+		}
+	}
+
 	if (u & FLAG_FACE_RECOGNITION) == FLAG_FACE_RECOGNITION {
 		faceBackends := []string{"insightface"}
 		if !slices.Contains(faceBackends, c.Backend) {
diff --git a/core/gallery/importers/importers.go b/core/gallery/importers/importers.go
index 8c156144d..a86e86530 100644
--- a/core/gallery/importers/importers.go
+++ b/core/gallery/importers/importers.go
@@ -143,6 +143,11 @@ var defaultImporters = []Importer{
 	&CoquiImporter{},
 	// Image/Video (Batch 3)
 	&StableDiffusionGGMLImporter{},
+	// Trellis2CppImporter (TRELLIS.2 image-to-3D, native C++/ggml port) must
+	// run before LlamaCPPImporter so its GGUF sets aren't claimed by the
+	// generic .gguf importer; matches only trellis-named URIs/repos or the
+	// distinctive component filenames, so arbitrary GGUFs are never claimed.
+	&Trellis2CppImporter{},
 	&ACEStepImporter{},
 	// LongCat repositories carry generic Diffusers metadata, so this exact
 	// owner/repo matcher must run before DiffuserImporter.
diff --git a/core/gallery/importers/trellis2cpp.go b/core/gallery/importers/trellis2cpp.go
new file mode 100644
index 000000000..b94bb904d
--- /dev/null
+++ b/core/gallery/importers/trellis2cpp.go
@@ -0,0 +1,170 @@
+package importers
+
+import (
+	"encoding/json"
+	"path/filepath"
+	"strings"
+
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/gallery"
+	"github.com/mudler/LocalAI/core/schema"
+	"go.yaml.in/yaml/v2"
+)
+
+var _ Importer = &Trellis2CppImporter{}
+
+// trellis2File describes one component of the TRELLIS.2 GGUF set hosted on
+// the LocalAI-io HuggingFace org. The pipeline spans three source repos
+// (TRELLIS.2-4B, TRELLIS-image-large for the SS decoder, and a DINOv3
+// mirror), so a single import URI always expands to this full set — no one
+// repo can describe it alone. Filenames follow the trellis2cpp converter
+// defaults, which the backend resolves without any options.
+type trellis2File struct {
+	filename string
+	uri      string
+	sha256   string
+}
+
+var trellis2Files = []trellis2File{
+	{"dino_f16.gguf", "https://huggingface.co/LocalAI-io/dinov3-vitl16-pretrain-lvd1689m-GGUF/resolve/main/dino_f16.gguf", "385d8186a38a2328ec740fb2ac1f33f9194d8774efc7ccafd4aa2e51cf5f6450"},
+	{"ss_flow_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF/resolve/main/ss_flow_f16.gguf", "1dded5b74237d24e6876a642a26f90b43742e3554418573860f810e3bbe61e8c"},
+	{"ss_dec_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS-image-large-GGUF/resolve/main/ss_dec_f16.gguf", "9c2210b7ed830fdc8286961a8189878ff5bcfd3bfc83ab4eacee005d293d2185"},
+	{"slat_flow_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF/resolve/main/slat_flow_f16.gguf", "2f94bad7b1c524ad8c01943bc38fcc0c314e7d482ce896f3c6e96eb6e7cec15c"},
+	{"slat_flow_1024_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF/resolve/main/slat_flow_1024_f16.gguf", "b6a2270131e2e9235e9b6cb525193eb85ae132fa5af3274322aacd39e40a6bc5"},
+	{"shape_dec_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF/resolve/main/shape_dec_f16.gguf", "6fe53f1d7763dabf7c8d72bc38f4053d87fde6f65bf17a9d378d27edb39d3530"},
+	{"shape_enc_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF/resolve/main/shape_enc_f16.gguf", "3ec80ff580987fcdb9bc594fc8b6fda890d63101ca442eb2b26f5dc315e8696c"},
+	{"tex_dec_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF/resolve/main/tex_dec_f16.gguf", "afd304f4dfcb8c94df851b85519b415b99f04070f7d29de1320c50631b1be4e0"},
+	{"tex_slat_flow_512_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF/resolve/main/tex_slat_flow_512_f16.gguf", "89a081b7f5487a5b31f03d240e4d959a56db0cc2c46c327230097a2554da52ae"},
+	{"tex_slat_flow_1024_f16.gguf", "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF/resolve/main/tex_slat_flow_1024_f16.gguf", "bbb55b0910c7929aac5e0612a9bb15113837a2c674cafb9f0f170eda8b5558a8"},
+}
+
+// trellis2ComponentNames are the distinctive default component filenames. A
+// raw .gguf URL with one of these basenames is a strong trellis2 signal.
+// dino_f16.gguf is deliberately absent — DINO checkpoints are common enough
+// that the bare name would over-claim.
+var trellis2ComponentNames = map[string]struct{}{
+	"ss_flow_f16.gguf":            {},
+	"ss_dec_f16.gguf":             {},
+	"slat_flow_f16.gguf":          {},
+	"slat_flow_1024_f16.gguf":     {},
+	"shape_dec_f16.gguf":          {},
+	"shape_enc_f16.gguf":          {},
+	"tex_dec_f16.gguf":            {},
+	"tex_slat_flow_512_f16.gguf":  {},
+	"tex_slat_flow_1024_f16.gguf": {},
+}
+
+// Trellis2CppImporter recognises Microsoft TRELLIS.2 image-to-3D GGUF sets
+// (the trellis2.cpp converter outputs hosted under LocalAI-io). It must be
+// registered BEFORE LlamaCPPImporter so llama-cpp does not steal the .gguf
+// match. preferences.backend="trellis2cpp" overrides detection.
+type Trellis2CppImporter struct{}
+
+func (i *Trellis2CppImporter) Name() string      { return "trellis2cpp" }
+func (i *Trellis2CppImporter) Modality() string  { return "3d" }
+func (i *Trellis2CppImporter) AutoDetects() bool { return true }
+
+// containsTrellisToken reports whether s (compared case-insensitively)
+// carries a TRELLIS marker ("trellis" covers TRELLIS.2 / trellis2 too).
+func containsTrellisToken(s string) bool {
+	return strings.Contains(strings.ToLower(s), "trellis")
+}
+
+func (i *Trellis2CppImporter) Match(details Details) bool {
+	preferences, err := details.Preferences.MarshalJSON()
+	if err != nil {
+		return false
+	}
+	preferencesMap := make(map[string]any)
+	if len(preferences) > 0 {
+		if err := json.Unmarshal(preferences, &preferencesMap); err != nil {
+			return false
+		}
+	}
+
+	if b, ok := preferencesMap["backend"].(string); ok && b != "" {
+		return b == "trellis2cpp"
+	}
+
+	// Raw .gguf URL named after a distinctive pipeline component.
+	if strings.HasSuffix(strings.ToLower(details.URI), ".gguf") {
+		base := strings.ToLower(filepath.Base(details.URI))
+		if _, ok := trellis2ComponentNames[base]; ok {
+			return true
+		}
+	}
+
+	// A trellis-named URI or HF repo carrying GGUFs.
+	if containsTrellisToken(details.URI) {
+		if strings.HasSuffix(strings.ToLower(details.URI), ".gguf") {
+			return true
+		}
+		if details.HuggingFace != nil && hasGGUF(details.HuggingFace.Files) {
+			return true
+		}
+		// HF details may be nil (tree-listing quirk) — decide from the
+		// owner/repo alone.
+		if _, repo, ok := HFOwnerRepoFromURI(details.URI); ok && containsTrellisToken(repo) {
+			return true
+		}
+	}
+
+	return false
+}
+
+func (i *Trellis2CppImporter) Import(details Details) (gallery.ModelConfig, error) {
+	preferences, err := details.Preferences.MarshalJSON()
+	if err != nil {
+		return gallery.ModelConfig{}, err
+	}
+	preferencesMap := make(map[string]any)
+	if len(preferences) > 0 {
+		if err := json.Unmarshal(preferences, &preferencesMap); err != nil {
+			return gallery.ModelConfig{}, err
+		}
+	}
+
+	name, ok := preferencesMap["name"].(string)
+	if !ok {
+		name = "trellis2-4b"
+	}
+
+	description, ok := preferencesMap["description"].(string)
+	if !ok {
+		description = "TRELLIS.2 image-to-3D (GLB with PBR textures) — imported from " + details.URI
+	}
+
+	cfg := gallery.ModelConfig{
+		Name:        name,
+		Description: description,
+	}
+	// The full pipeline spans three HF repos, so any trellis URI imports the
+	// complete known-good set rather than whatever single repo was pasted.
+	for _, f := range trellis2Files {
+		cfg.Files = append(cfg.Files, gallery.File{
+			URI:      f.uri,
+			Filename: f.filename,
+			SHA256:   f.sha256,
+		})
+	}
+
+	modelConfig := config.ModelConfig{
+		Name:                name,
+		Description:         description,
+		Backend:             "trellis2cpp",
+		KnownUsecaseStrings: []string{"FLAG_3D"},
+		PredictionOptions: schema.PredictionOptions{
+			// ss_flow anchors the GGUF directory; the backend resolves the
+			// other components from their default filenames next to it.
+			BasicModelRequest: schema.BasicModelRequest{Model: "ss_flow_f16.gguf"},
+		},
+	}
+
+	data, err := yaml.Marshal(modelConfig)
+	if err != nil {
+		return gallery.ModelConfig{}, err
+	}
+
+	cfg.ConfigFile = string(data)
+	return cfg, nil
+}
diff --git a/core/gallery/importers/trellis2cpp_test.go b/core/gallery/importers/trellis2cpp_test.go
new file mode 100644
index 000000000..00251526f
--- /dev/null
+++ b/core/gallery/importers/trellis2cpp_test.go
@@ -0,0 +1,105 @@
+package importers_test
+
+import (
+	"encoding/json"
+	"fmt"
+
+	"github.com/mudler/LocalAI/core/gallery/importers"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("Trellis2CppImporter", func() {
+	Context("detection from HuggingFace", func() {
+		// LocalAI-io/TRELLIS.2-4B-GGUF is the canonical GGUF conversion of
+		// microsoft/TRELLIS.2-4B produced by the trellis2cpp converters.
+		// Detection must route it to trellis2cpp (and NOT to llama-cpp,
+		// which otherwise steals every .gguf repo).
+		It("matches the TRELLIS.2 GGUF repo and imports the full component set", func() {
+			uri := "https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF"
+			preferences := json.RawMessage(`{}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("backend: trellis2cpp"), fmt.Sprintf("Model config: %+v", modelConfig))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("known_usecases"))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("FLAG_3D"))
+			// The pipeline spans three repos; the import must carry the whole
+			// set, anchored on ss_flow.
+			Expect(modelConfig.Files).To(HaveLen(10))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("model: ss_flow_f16.gguf"))
+		})
+
+		It("matches a raw .gguf URL named after a distinctive pipeline component", func() {
+			uri := "https://example.com/models/tex_slat_flow_512_f16.gguf"
+			preferences := json.RawMessage(`{}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("backend: trellis2cpp"), fmt.Sprintf("Model config: %+v", modelConfig))
+		})
+	})
+
+	Context("preference override", func() {
+		It("honours preferences.backend=trellis2cpp for arbitrary URIs", func() {
+			uri := "https://example.com/some-unrelated-model"
+			preferences := json.RawMessage(`{"backend": "trellis2cpp"}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("backend: trellis2cpp"), fmt.Sprintf("Model config: %+v", modelConfig))
+		})
+
+		It("does not override a different explicit backend", func() {
+			imp := &importers.Trellis2CppImporter{}
+			match := imp.Match(importers.Details{
+				URI:         "https://example.com/models/tex_slat_flow_512_f16.gguf",
+				Preferences: json.RawMessage(`{"backend": "llama-cpp"}`),
+			})
+
+			Expect(match).To(BeFalse())
+		})
+
+		It("still auto-detects when the backend preference is empty", func() {
+			imp := &importers.Trellis2CppImporter{}
+			match := imp.Match(importers.Details{
+				URI:         "https://example.com/models/tex_slat_flow_512_f16.gguf",
+				Preferences: json.RawMessage(`{"backend": ""}`),
+			})
+
+			Expect(match).To(BeTrue())
+		})
+	})
+
+	Context("negative detection", func() {
+		It("does not claim an unrelated raw .gguf URL", func() {
+			imp := &importers.Trellis2CppImporter{}
+			match := imp.Match(importers.Details{
+				URI:         "https://example.com/models/llama-3-8b-Q4_K.gguf",
+				Preferences: json.RawMessage(`{}`),
+			})
+			Expect(match).To(BeFalse())
+		})
+
+		It("does not claim a bare dino_f16.gguf (too generic a name)", func() {
+			imp := &importers.Trellis2CppImporter{}
+			match := imp.Match(importers.Details{
+				URI:         "https://example.com/models/dino_f16.gguf",
+				Preferences: json.RawMessage(`{}`),
+			})
+			Expect(match).To(BeFalse())
+		})
+	})
+
+	Context("Importer interface metadata", func() {
+		It("exposes name/modality/autodetect", func() {
+			imp := &importers.Trellis2CppImporter{}
+			Expect(imp.Name()).To(Equal("trellis2cpp"))
+			Expect(imp.Modality()).To(Equal("3d"))
+			Expect(imp.AutoDetects()).To(BeTrue())
+		})
+	})
+})
diff --git a/core/http/app.go b/core/http/app.go
index bee47013b..8b08af4ad 100644
--- a/core/http/app.go
+++ b/core/http/app.go
@@ -55,6 +55,12 @@ var quietPaths = []string{"/api/operations", "/api/resources", "/healthz", "/rea
 // conditional revalidation round-trip.
 const immutableAssetCacheControl = "public, max-age=31536000, immutable"
 
+func defaultBodyLimitSkipper(c echo.Context) bool {
+	// Remeshing accepts generated GLBs that routinely exceed the default
+	// upload limit. The route has its own tighter, format-specific limit.
+	return c.Request().Method == http.MethodPost && c.Path() == "/3d/remesh"
+}
+
 // applyModelLoadCooldown maps a ModelLoadCooldownError anywhere in err's chain
 // to HTTP 503 with a Retry-After header (whole seconds, floor 1), so a client
 // polling a model whose load recently failed backs off instead of triggering a
@@ -123,7 +129,10 @@ func API(application *application.Application) (*echo.Echo, error) {
 
 	// Set body limit
 	if application.ApplicationConfig().UploadLimitMB > 0 {
-		e.Use(middleware.BodyLimit(fmt.Sprintf("%dM", application.ApplicationConfig().UploadLimitMB)))
+		e.Use(middleware.BodyLimitWithConfig(middleware.BodyLimitConfig{
+			Limit:   fmt.Sprintf("%dM", application.ApplicationConfig().UploadLimitMB),
+			Skipper: defaultBodyLimitSkipper,
+		}))
 	}
 
 	// SPA fallback handler, set later when React UI is available
@@ -305,14 +314,22 @@ func API(application *application.Application) (*echo.Echo, error) {
 		audioPath := filepath.Join(application.ApplicationConfig().GeneratedContentDir, "audio")
 		imagePath := filepath.Join(application.ApplicationConfig().GeneratedContentDir, "images")
 		videoPath := filepath.Join(application.ApplicationConfig().GeneratedContentDir, "videos")
+		threeDPath := filepath.Join(application.ApplicationConfig().GeneratedContentDir, "3d")
 
 		os.MkdirAll(audioPath, 0750)
 		os.MkdirAll(imagePath, 0750)
 		os.MkdirAll(videoPath, 0750)
+		_ = os.MkdirAll(threeDPath, 0750)
+
+		// Go's built-in MIME table has no .glb entry and minimal containers
+		// ship no /etc/mime.types, so generated GLBs would otherwise be
+		// served as application/octet-stream.
+		_ = mime.AddExtensionType(".glb", "model/gltf-binary")
 
 		e.Static("/generated-audio", audioPath)
 		e.Static("/generated-images", imagePath)
 		e.Static("/generated-videos", videoPath)
+		e.Static("/generated-3d", threeDPath)
 	}
 
 	// Usage recording is initialised in application/startup.go and
diff --git a/core/http/auth/features.go b/core/http/auth/features.go
index a43eef992..063b4d76e 100644
--- a/core/http/auth/features.go
+++ b/core/http/auth/features.go
@@ -91,6 +91,10 @@ var RouteFeatureRegistry = []RouteFeature{
 	// Video
 	{"POST", "/video", FeatureVideo},
 
+	// 3D generation
+	{"POST", "/3d/generations", Feature3D},
+	{"POST", "/3d/remesh", Feature3D},
+
 	// Sound generation
 	{"POST", "/v1/sound-generation", FeatureSound},
 
@@ -182,6 +186,7 @@ func APIFeatureMetas() []FeatureMeta {
 		{FeatureVAD, "Voice Activity Detection", true},
 		{FeatureDetection, "Detection", true},
 		{FeatureVideo, "Video Generation", true},
+		{Feature3D, "3D Generation", true},
 		{FeatureEmbeddings, "Embeddings", true},
 		{FeatureSound, "Sound Generation", true},
 		{FeatureRealtime, "Realtime", true},
diff --git a/core/http/auth/middleware.go b/core/http/auth/middleware.go
index c67954640..dd76bdf09 100644
--- a/core/http/auth/middleware.go
+++ b/core/http/auth/middleware.go
@@ -584,6 +584,7 @@ func isAPIPath(path string) bool {
 		strings.HasPrefix(path, "/tts") ||
 		strings.HasPrefix(path, "/vad") ||
 		strings.HasPrefix(path, "/video") ||
+		strings.HasPrefix(path, "/3d/") ||
 		strings.HasPrefix(path, "/stores/") ||
 		strings.HasPrefix(path, "/system") ||
 		strings.HasPrefix(path, "/ws/") ||
diff --git a/core/http/auth/middleware_test.go b/core/http/auth/middleware_test.go
index 5137851e1..94be9336a 100644
--- a/core/http/auth/middleware_test.go
+++ b/core/http/auth/middleware_test.go
@@ -156,6 +156,16 @@ var _ = Describe("Auth Middleware", func() {
 			Expect(rec.Code).To(Equal(http.StatusUnauthorized))
 		})
 
+		It("returns 401 for unauthenticated 3D generation requests", func() {
+			rec := doRequest(app, http.MethodPost, "/3d/generations")
+			Expect(rec.Code).To(Equal(http.StatusUnauthorized))
+		})
+
+		It("returns 401 for unauthenticated 3D remesh requests", func() {
+			rec := doRequest(app, http.MethodPost, "/3d/remesh")
+			Expect(rec.Code).To(Equal(http.StatusUnauthorized))
+		})
+
 		It("allows unauthenticated access to non-API paths when no legacy keys", func() {
 			rec := doRequest(app, http.MethodGet, "/app")
 			Expect(rec.Code).To(Equal(http.StatusOK))
diff --git a/core/http/auth/permissions.go b/core/http/auth/permissions.go
index 1795792f9..6c01e03a3 100644
--- a/core/http/auth/permissions.go
+++ b/core/http/auth/permissions.go
@@ -47,6 +47,7 @@ const (
 	FeatureVAD                 = "vad"
 	FeatureDetection           = "detection"
 	FeatureVideo               = "video"
+	Feature3D                  = "3d"
 	FeatureEmbeddings          = "embeddings"
 	FeatureSound               = "sound"
 	FeatureRealtime            = "realtime"
@@ -73,7 +74,7 @@ var GeneralFeatures = []string{FeatureFineTuning, FeatureQuantization}
 var APIFeatures = []string{
 	FeatureChat, FeatureImages, FeatureAudioSpeech, FeatureAudioTranscription,
 	FeatureAudioDiarization, FeatureAudioClassification,
-	FeatureVAD, FeatureDetection, FeatureVideo, FeatureEmbeddings, FeatureSound,
+	FeatureVAD, FeatureDetection, FeatureVideo, Feature3D, FeatureEmbeddings, FeatureSound,
 	FeatureRealtime, FeatureRerank, FeatureTokenize, FeatureMCP, FeatureStores,
 	FeatureFaceRecognition, FeatureVoiceRecognition, FeatureAudioTransform,
 	FeaturePIIFilter,
diff --git a/core/http/body_limit_test.go b/core/http/body_limit_test.go
new file mode 100644
index 000000000..7d398766c
--- /dev/null
+++ b/core/http/body_limit_test.go
@@ -0,0 +1,66 @@
+package http_test
+
+import (
+	"bytes"
+	"context"
+	"mime/multipart"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+
+	"github.com/mudler/LocalAI/core/application"
+	"github.com/mudler/LocalAI/core/config"
+	. "github.com/mudler/LocalAI/core/http"
+	"github.com/mudler/LocalAI/pkg/system"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("Request body limits", func() {
+	It("lets the remesh route apply its larger limit without weakening other routes", func() {
+		dir := GinkgoT().TempDir()
+		models := filepath.Join(dir, "models")
+		backends := filepath.Join(dir, "backends")
+		Expect(os.Mkdir(models, 0o750)).To(Succeed())
+		Expect(os.Mkdir(backends, 0o750)).To(Succeed())
+
+		state, err := system.GetSystemState(
+			system.WithModelPath(models),
+			system.WithBackendPath(backends),
+		)
+		Expect(err).NotTo(HaveOccurred())
+		ctx, cancel := context.WithCancel(context.Background())
+		defer cancel()
+		localApp, err := application.New(
+			config.WithContext(ctx),
+			config.WithSystemState(state),
+			config.WithUploadLimitMB(1),
+		)
+		Expect(err).NotTo(HaveOccurred())
+		defer func() { _ = localApp.Shutdown() }()
+		app, err := API(localApp)
+		Expect(err).NotTo(HaveOccurred())
+
+		body := new(bytes.Buffer)
+		writer := multipart.NewWriter(body)
+		Expect(writer.WriteField("model", "missing-model")).To(Succeed())
+		part, err := writer.CreateFormFile("mesh", "large.glb")
+		Expect(err).NotTo(HaveOccurred())
+		_, err = part.Write(bytes.Repeat([]byte{'x'}, 2<<20))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(writer.Close()).To(Succeed())
+
+		request := httptest.NewRequest(http.MethodPost, "/3d/remesh", body)
+		request.Header.Set("Content-Type", writer.FormDataContentType())
+		response := httptest.NewRecorder()
+		app.ServeHTTP(response, request)
+		Expect(response.Code).To(Equal(http.StatusNotFound), response.Body.String())
+
+		request = httptest.NewRequest(http.MethodPost, "/3d/generations", bytes.NewReader(make([]byte, 2<<20)))
+		request.Header.Set("Content-Type", "application/json")
+		response = httptest.NewRecorder()
+		app.ServeHTTP(response, request)
+		Expect(response.Code).To(Equal(http.StatusRequestEntityTooLarge), response.Body.String())
+	})
+})
diff --git a/core/http/endpoints/localai/api_instructions.go b/core/http/endpoints/localai/api_instructions.go
index 02ff85d1d..62f264cf0 100644
--- a/core/http/endpoints/localai/api_instructions.go
+++ b/core/http/endpoints/localai/api_instructions.go
@@ -81,6 +81,12 @@ var instructionDefs = []instructionDef{
 		Tags:        []string{"video"},
 		Intro:       "POST /video accepts start_image, end_image, and audio as public URL, base64, or data URI. Backend-specific tuning is passed as string values in params.",
 	},
+	{
+		Name:        "3d",
+		Description: "Image-to-3D asset generation (binary glTF / GLB) via TRELLIS.2",
+		Tags:        []string{"3d"},
+		Intro:       "POST /3d/generations accepts a conditioning image as public URL, base64, or data URI (no text prompt) and returns one .glb asset as a URL under /generated-3d or as b64_json. quality selects the mesh pipeline (auto|coarse|512|1024); background controls solid-background removal (auto|keep|black|white); step, texture_steps, and cfg_scale tune the flow sampling. POST /3d/remesh accepts multipart model, mesh (GLB), and a single detail percentage to return a watertight print-ready GLB; the enclosing offset is derived automatically.",
+	},
 	{
 		Name:        "face-recognition",
 		Description: "Face verification (1:1), identification (1:N), embedding, and demographic analysis",
diff --git a/core/http/endpoints/localai/api_instructions_test.go b/core/http/endpoints/localai/api_instructions_test.go
index 43caa9ee1..9648e674a 100644
--- a/core/http/endpoints/localai/api_instructions_test.go
+++ b/core/http/endpoints/localai/api_instructions_test.go
@@ -39,7 +39,7 @@ var _ = Describe("API Instructions Endpoints", func() {
 
 			instructions, ok := resp["instructions"].([]any)
 			Expect(ok).To(BeTrue())
-			Expect(instructions).To(HaveLen(17))
+			Expect(instructions).To(HaveLen(18))
 
 			// Verify each instruction has required fields and correct URL format
 			for _, s := range instructions {
@@ -79,6 +79,7 @@ var _ = Describe("API Instructions Endpoints", func() {
 				"middleware-admin",
 				"intelligent-routing",
 				"voice-library",
+				"3d",
 			))
 		})
 	})
@@ -123,6 +124,16 @@ var _ = Describe("API Instructions Endpoints", func() {
 			Expect(string(body)).To(ContainSubstring("stream"))
 		})
 
+		It("should advertise the LocalAI 3D generation path", func() {
+			req := httptest.NewRequest(http.MethodGet, "/api/instructions/3d", nil)
+			rec := httptest.NewRecorder()
+			app.ServeHTTP(rec, req)
+
+			body, _ := io.ReadAll(rec.Body)
+			Expect(string(body)).To(ContainSubstring("POST /3d/generations"))
+			Expect(string(body)).NotTo(ContainSubstring("/v1/3d/generations"))
+		})
+
 		It("should return JSON fragment when format=json", func() {
 			req := httptest.NewRequest(http.MethodGet, "/api/instructions/chat-inference?format=json", nil)
 			rec := httptest.NewRecorder()
diff --git a/core/http/endpoints/localai/model3d.go b/core/http/endpoints/localai/model3d.go
new file mode 100644
index 000000000..fd514c7db
--- /dev/null
+++ b/core/http/endpoints/localai/model3d.go
@@ -0,0 +1,186 @@
+package localai
+
+import (
+	"encoding/base64"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"net/url"
+	"os"
+	"path/filepath"
+	"slices"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/labstack/echo/v4"
+
+	"github.com/mudler/LocalAI/core/backend"
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/http/middleware"
+	"github.com/mudler/LocalAI/core/schema"
+
+	"github.com/mudler/xlog"
+
+	model "github.com/mudler/LocalAI/pkg/model"
+)
+
+// Conditioning images are single frames, so a much tighter cap than the
+// video-input limit is enough.
+const max3DInputBytes = 32 << 20
+
+var (
+	valid3DQualities   = []string{"", "auto", "coarse", "512", "1024"}
+	valid3DBackgrounds = []string{"", "auto", "keep", "black", "white"}
+)
+
+// Model3DEndpoint
+// @Summary Creates a 3D asset (binary glTF / GLB) from a conditioning image.
+// @Tags 3d
+// @Param request body schema.Model3DRequest true "query params"
+// @Success 200 {object} schema.OpenAIResponse "Response"
+// @Router /3d/generations [post]
+func Model3DEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+	return func(c echo.Context) error {
+		input, ok := c.Get(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST).(*schema.Model3DRequest)
+		if !ok || input.Model == "" {
+			xlog.Error("3D Endpoint - Invalid Input")
+			return echo.ErrBadRequest
+		}
+
+		config, ok := c.Get(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG).(*config.ModelConfig)
+		if !ok || config == nil {
+			xlog.Error("3D Endpoint - Invalid Config")
+			return echo.ErrBadRequest
+		}
+
+		if input.Image == "" {
+			return echo.NewHTTPError(http.StatusBadRequest, "image is required: 3D generation is image-conditioned")
+		}
+		// Reject unknown enum values here rather than surfacing an opaque
+		// backend error after a model load.
+		if !slices.Contains(valid3DQualities, input.Quality) {
+			return echo.NewHTTPError(http.StatusBadRequest, fmt.Sprintf("invalid quality %q: must be one of auto, coarse, 512, 1024", input.Quality))
+		}
+		if !slices.Contains(valid3DBackgrounds, input.Background) {
+			return echo.NewHTTPError(http.StatusBadRequest, fmt.Sprintf("invalid background %q: must be one of auto, keep, black, white", input.Background))
+		}
+
+		src, err := stageVideoMediaWithLimit(c.Request().Context(), appConfig.GeneratedContentDir, input.Image, max3DInputBytes)
+		if err != nil {
+			return echo.NewHTTPError(http.StatusBadRequest, fmt.Sprintf("invalid image: %v", err))
+		}
+		defer func() { _ = os.Remove(src) }()
+
+		xlog.Debug("Parameter Config", "config", config)
+
+		if config.Backend == "" {
+			config.Backend = model.Trellis2CppBackend
+		}
+
+		step := input.Step
+		if step == 0 && config.Step != 0 {
+			step = int32(config.Step)
+		}
+		cfgScale := input.CFGScale
+		if cfgScale == 0 && config.CFGScale != 0 {
+			cfgScale = config.CFGScale
+		}
+
+		b64JSON := input.ResponseFormat == "b64_json"
+
+		tempDir := ""
+		if !b64JSON {
+			tempDir = filepath.Join(appConfig.GeneratedContentDir, "3d")
+			if err := os.MkdirAll(tempDir, 0o750); err != nil {
+				return err
+			}
+		}
+		// Create a temporary file
+		outputFile, err := os.CreateTemp(tempDir, "b64")
+		if err != nil {
+			return err
+		}
+		if err := outputFile.Close(); err != nil {
+			_ = os.Remove(outputFile.Name())
+			return err
+		}
+
+		output := outputFile.Name() + ".glb"
+
+		// Rename the temporary file
+		err = os.Rename(outputFile.Name(), output)
+		if err != nil {
+			_ = os.Remove(outputFile.Name())
+			return err
+		}
+		preserveOutput := false
+		defer func() {
+			if !preserveOutput {
+				_ = os.Remove(output)
+			}
+		}()
+
+		baseURL := middleware.BaseURL(c)
+
+		xlog.Debug("Model3DEndpoint: Calling Model3DGeneration",
+			"quality", input.Quality,
+			"background", input.Background,
+			"cfg_scale", cfgScale,
+			"step", step,
+			"texture_steps", input.TextureSteps,
+			"seed", input.Seed)
+
+		fn, err := backend.Model3DGeneration(
+			backend.Model3DGenerationOptions{
+				Image:        src,
+				Destination:  output,
+				Seed:         input.Seed,
+				Step:         step,
+				CFGScale:     cfgScale,
+				TextureSteps: input.TextureSteps,
+				Quality:      input.Quality,
+				Background:   input.Background,
+				Params:       input.Params,
+			},
+			ml,
+			*config,
+			appConfig,
+		)
+		if err != nil {
+			return mapBackendError(err)
+		}
+		if err := fn(); err != nil {
+			return mapBackendError(err)
+		}
+
+		item := &schema.Item{}
+
+		if b64JSON {
+			data, err := os.ReadFile(output)
+			if err != nil {
+				return err
+			}
+			item.B64JSON = base64.StdEncoding.EncodeToString(data)
+		} else {
+			base := filepath.Base(output)
+			item.URL, err = url.JoinPath(baseURL, "generated-3d", base)
+			if err != nil {
+				return err
+			}
+			preserveOutput = true
+		}
+
+		id := uuid.New().String()
+		created := int(time.Now().Unix())
+		resp := &schema.OpenAIResponse{
+			ID:      id,
+			Created: created,
+			Data:    []schema.Item{*item},
+		}
+
+		jsonResult, _ := json.Marshal(resp)
+		xlog.Debug("Response", "response", string(jsonResult))
+
+		return c.JSON(200, resp)
+	}
+}
diff --git a/core/http/endpoints/localai/model3d_internal_test.go b/core/http/endpoints/localai/model3d_internal_test.go
new file mode 100644
index 000000000..4437a2a3a
--- /dev/null
+++ b/core/http/endpoints/localai/model3d_internal_test.go
@@ -0,0 +1,72 @@
+package localai
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"strings"
+
+	"github.com/labstack/echo/v4"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/http/middleware"
+	"github.com/mudler/LocalAI/core/schema"
+)
+
+// The validation branches all return before any model load, so the handler can
+// be driven with nil loaders.
+var _ = Describe("3D endpoint request validation", func() {
+	call := func(input *schema.Model3DRequest) error {
+		appConfig := &config.ApplicationConfig{GeneratedContentDir: GinkgoT().TempDir()}
+		handler := Model3DEndpoint(nil, nil, appConfig)
+
+		e := echo.New()
+		req := httptest.NewRequest(http.MethodPost, "/3d/generations", strings.NewReader("{}"))
+		c := e.NewContext(req, httptest.NewRecorder())
+		c.Set(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST, input)
+		c.Set(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG, &config.ModelConfig{Name: "test-3d"})
+		return handler(c)
+	}
+
+	expectBadRequest := func(err error, substr string) {
+		var httpErr *echo.HTTPError
+		Expect(err).To(BeAssignableToTypeOf(httpErr))
+		httpErr = err.(*echo.HTTPError)
+		Expect(httpErr.Code).To(Equal(http.StatusBadRequest))
+		Expect(httpErr.Message).To(ContainSubstring(substr))
+	}
+
+	It("requires a conditioning image", func() {
+		err := call(&schema.Model3DRequest{BasicModelRequest: schema.BasicModelRequest{Model: "m"}})
+		expectBadRequest(err, "image is required")
+	})
+
+	It("rejects unknown quality values", func() {
+		err := call(&schema.Model3DRequest{
+			BasicModelRequest: schema.BasicModelRequest{Model: "m"},
+			Image:             "aGk=",
+			Quality:           "2048",
+		})
+		expectBadRequest(err, "invalid quality")
+	})
+
+	It("rejects unknown background values", func() {
+		err := call(&schema.Model3DRequest{
+			BasicModelRequest: schema.BasicModelRequest{Model: "m"},
+			Image:             "aGk=",
+			Background:        "transparent",
+		})
+		expectBadRequest(err, "invalid background")
+	})
+
+	It("rejects undecodable image payloads", func() {
+		err := call(&schema.Model3DRequest{
+			BasicModelRequest: schema.BasicModelRequest{Model: "m"},
+			Image:             "not%%%base64",
+			Quality:           "512",
+			Background:        "auto",
+		})
+		expectBadRequest(err, "invalid image")
+	})
+})
diff --git a/core/http/endpoints/localai/model3d_remesh.go b/core/http/endpoints/localai/model3d_remesh.go
new file mode 100644
index 000000000..a0a2b004e
--- /dev/null
+++ b/core/http/endpoints/localai/model3d_remesh.go
@@ -0,0 +1,163 @@
+package localai
+
+import (
+	"fmt"
+	"io"
+	"math"
+	"net/http"
+	"os"
+	"path/filepath"
+	"strconv"
+
+	"github.com/labstack/echo/v4"
+
+	"github.com/mudler/LocalAI/core/backend"
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/http/middleware"
+	"github.com/mudler/LocalAI/core/schema"
+	model "github.com/mudler/LocalAI/pkg/model"
+)
+
+const (
+	max3DRemeshBytes       = 512 << 20
+	defaultRemeshDetailPct = float32(0.5)
+	minRemeshDetailPct     = float32(0.35)
+	maxRemeshDetailPct     = float32(2.5)
+)
+
+func normalizedRemeshDetail(detail float32) (float32, error) {
+	if detail == 0 {
+		return defaultRemeshDetailPct, nil
+	}
+	if math.IsNaN(float64(detail)) || math.IsInf(float64(detail), 0) || detail < minRemeshDetailPct || detail > maxRemeshDetailPct {
+		return 0, fmt.Errorf("detail must be between %.2f and %.2f percent", minRemeshDetailPct, maxRemeshDetailPct)
+	}
+	return detail, nil
+}
+
+func saveRemeshUpload(c echo.Context, dir string) (string, error) {
+	header, err := c.FormFile("mesh")
+	if err != nil {
+		return "", fmt.Errorf("mesh is required")
+	}
+	if header.Size > max3DRemeshBytes {
+		return "", fmt.Errorf("mesh exceeds the 512 MiB limit")
+	}
+	source, err := header.Open()
+	if err != nil {
+		return "", fmt.Errorf("opening mesh: %w", err)
+	}
+	defer func() { _ = source.Close() }()
+
+	if err := os.MkdirAll(dir, 0o750); err != nil {
+		return "", err
+	}
+	temp, err := os.CreateTemp(dir, "remesh-input-*.glb")
+	if err != nil {
+		return "", err
+	}
+	path := temp.Name()
+	defer func() {
+		if err != nil {
+			_ = os.Remove(path)
+		}
+	}()
+
+	written, copyErr := io.Copy(temp, io.LimitReader(source, max3DRemeshBytes+1))
+	closeErr := temp.Close()
+	if copyErr != nil {
+		err = fmt.Errorf("saving mesh: %w", copyErr)
+		return "", err
+	}
+	if closeErr != nil {
+		err = closeErr
+		return "", err
+	}
+	if written == 0 {
+		err = fmt.Errorf("mesh is empty")
+		return "", err
+	}
+	if written > max3DRemeshBytes {
+		err = fmt.Errorf("mesh exceeds the 512 MiB limit")
+		return "", err
+	}
+	return path, nil
+}
+
+// Model3DRemeshEndpoint rebuilds an existing generated GLB as a watertight mesh.
+// @Summary Applies watertight print remeshing to an existing 3D asset.
+// @Tags 3d
+// @Accept multipart/form-data
+// @Produce model/gltf-binary
+// @Param model formData string true "3D model name"
+// @Param mesh formData file true "Source GLB"
+// @Param detail formData number false "Detail size as percent of the source bounding-box diagonal (0.35–2.5; default 0.5)"
+// @Success 200 {file} binary "Remeshed GLB"
+// @Router /3d/remesh [post]
+func Model3DRemeshEndpoint(ml *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+	return func(c echo.Context) error {
+		input, ok := c.Get(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST).(*schema.Model3DRemeshRequest)
+		if !ok || input.Model == "" {
+			return echo.NewHTTPError(http.StatusBadRequest, "model is required")
+		}
+		modelConfig, ok := c.Get(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG).(*config.ModelConfig)
+		if !ok || modelConfig == nil {
+			return echo.ErrBadRequest
+		}
+		detail, err := normalizedRemeshDetail(input.Detail)
+		if err != nil {
+			return echo.NewHTTPError(http.StatusBadRequest, err.Error())
+		}
+
+		source, err := saveRemeshUpload(c, appConfig.GeneratedContentDir)
+		if err != nil {
+			return echo.NewHTTPError(http.StatusBadRequest, err.Error())
+		}
+		defer func() { _ = os.Remove(source) }()
+
+		outputFile, err := os.CreateTemp(appConfig.GeneratedContentDir, "remeshed-*.glb")
+		if err != nil {
+			return err
+		}
+		output := outputFile.Name()
+		if err := outputFile.Close(); err != nil {
+			_ = os.Remove(output)
+			return err
+		}
+		defer func() { _ = os.Remove(output) }()
+
+		if modelConfig.Backend == "" {
+			modelConfig.Backend = model.Trellis2CppBackend
+		}
+		fn, err := backend.Model3DGeneration(
+			backend.Model3DGenerationOptions{
+				Image:       source,
+				Destination: output,
+				Params: map[string]string{
+					"operation":      "print_remesh",
+					"alpha_ratio":    strconv.FormatFloat(float64(detail/100), 'g', -1, 32),
+					"detail_percent": strconv.FormatFloat(float64(detail), 'g', -1, 32),
+					"texture_size":   "2048",
+				},
+			},
+			ml,
+			*modelConfig,
+			appConfig,
+		)
+		if err != nil {
+			return mapBackendError(err)
+		}
+		if err := fn(); err != nil {
+			return mapBackendError(err)
+		}
+
+		file, err := os.Open(filepath.Clean(output))
+		if err != nil {
+			return err
+		}
+		defer func() { _ = file.Close() }()
+		c.Response().Header().Set(echo.HeaderContentDisposition, `attachment; filename="remeshed.glb"`)
+		c.Response().Header().Set(echo.HeaderCacheControl, "no-store")
+		return c.Stream(http.StatusOK, "model/gltf-binary", file)
+	}
+}
diff --git a/core/http/endpoints/localai/model3d_remesh_internal_test.go b/core/http/endpoints/localai/model3d_remesh_internal_test.go
new file mode 100644
index 000000000..d3e4108e2
--- /dev/null
+++ b/core/http/endpoints/localai/model3d_remesh_internal_test.go
@@ -0,0 +1,68 @@
+package localai
+
+import (
+	"bytes"
+	"math"
+	"mime/multipart"
+	"net/http"
+	"net/http/httptest"
+	"os"
+
+	"github.com/labstack/echo/v4"
+	"github.com/mudler/LocalAI/core/schema"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("3D print remeshing request handling", func() {
+	DescribeTable("normalizes the demo detail range",
+		func(input, expected float32, valid bool) {
+			value, err := normalizedRemeshDetail(input)
+			if valid {
+				Expect(err).NotTo(HaveOccurred())
+				Expect(value).To(BeNumerically("~", expected, 1e-6))
+			} else {
+				Expect(err).To(HaveOccurred())
+			}
+		},
+		Entry("default", float32(0), float32(0.5), true),
+		Entry("fine endpoint", float32(0.35), float32(0.35), true),
+		Entry("coarse endpoint", float32(2.5), float32(2.5), true),
+		Entry("too fine", float32(0.1), float32(0), false),
+		Entry("too coarse", float32(3), float32(0), false),
+		Entry("not a number", float32(math.NaN()), float32(0), false),
+	)
+
+	It("streams the multipart GLB to a bounded temporary file", func() {
+		body := new(bytes.Buffer)
+		writer := multipart.NewWriter(body)
+		Expect(writer.WriteField("model", "trellis-test-model")).To(Succeed())
+		Expect(writer.WriteField("detail", "0.35")).To(Succeed())
+		part, err := writer.CreateFormFile("mesh", "source.glb")
+		Expect(err).NotTo(HaveOccurred())
+		_, err = part.Write([]byte("glTF-test"))
+		Expect(err).NotTo(HaveOccurred())
+		Expect(writer.Close()).To(Succeed())
+
+		e := echo.New()
+		req := httptest.NewRequest(http.MethodPost, "/3d/remesh", body)
+		req.Header.Set(echo.HeaderContentType, writer.FormDataContentType())
+		ctx := e.NewContext(req, httptest.NewRecorder())
+		input := new(schema.Model3DRemeshRequest)
+		Expect(ctx.Bind(input)).To(Succeed())
+		Expect(input.Model).To(Equal("trellis-test-model"))
+		Expect(input.Detail).To(BeNumerically("~", 0.35, 1e-6))
+		path, err := saveRemeshUpload(ctx, GinkgoT().TempDir())
+		Expect(err).NotTo(HaveOccurred())
+		defer func() { _ = os.Remove(path) }()
+		Expect(os.ReadFile(path)).To(Equal([]byte("glTF-test")))
+	})
+
+	It("requires a mesh part", func() {
+		e := echo.New()
+		req := httptest.NewRequest(http.MethodPost, "/3d/remesh", bytes.NewReader(nil))
+		ctx := e.NewContext(req, httptest.NewRecorder())
+		_, err := saveRemeshUpload(ctx, GinkgoT().TempDir())
+		Expect(err).To(MatchError("mesh is required"))
+	})
+})
diff --git a/core/http/endpoints/localai/stores.go b/core/http/endpoints/localai/stores.go
index 8074da9e0..af9cf1569 100644
--- a/core/http/endpoints/localai/stores.go
+++ b/core/http/endpoints/localai/stores.go
@@ -9,7 +9,7 @@ import (
 	"github.com/mudler/LocalAI/pkg/store"
 )
 
-func StoresSetEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+func StoresSetEndpoint(sl *model.ModelLoader, cl *config.ModelConfigLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		input := new(schema.StoresSet)
 
@@ -17,7 +17,7 @@ func StoresSetEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConfi
 			return err
 		}
 
-		sb, err := backend.StoreBackend(sl, appConfig, input.Store, input.Backend)
+		sb, err := backend.StoreBackend(sl, appConfig, cl, input.Store, input.Backend)
 		if err != nil {
 			return err
 		}
@@ -36,7 +36,7 @@ func StoresSetEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConfi
 	}
 }
 
-func StoresDeleteEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+func StoresDeleteEndpoint(sl *model.ModelLoader, cl *config.ModelConfigLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		input := new(schema.StoresDelete)
 
@@ -44,7 +44,7 @@ func StoresDeleteEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationCo
 			return err
 		}
 
-		sb, err := backend.StoreBackend(sl, appConfig, input.Store, input.Backend)
+		sb, err := backend.StoreBackend(sl, appConfig, cl, input.Store, input.Backend)
 		if err != nil {
 			return err
 		}
@@ -57,7 +57,7 @@ func StoresDeleteEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationCo
 	}
 }
 
-func StoresGetEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+func StoresGetEndpoint(sl *model.ModelLoader, cl *config.ModelConfigLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		input := new(schema.StoresGet)
 
@@ -65,7 +65,7 @@ func StoresGetEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConfi
 			return err
 		}
 
-		sb, err := backend.StoreBackend(sl, appConfig, input.Store, input.Backend)
+		sb, err := backend.StoreBackend(sl, appConfig, cl, input.Store, input.Backend)
 		if err != nil {
 			return err
 		}
@@ -88,7 +88,7 @@ func StoresGetEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConfi
 	}
 }
 
-func StoresFindEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+func StoresFindEndpoint(sl *model.ModelLoader, cl *config.ModelConfigLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		input := new(schema.StoresFind)
 
@@ -96,7 +96,7 @@ func StoresFindEndpoint(sl *model.ModelLoader, appConfig *config.ApplicationConf
 			return err
 		}
 
-		sb, err := backend.StoreBackend(sl, appConfig, input.Store, input.Backend)
+		sb, err := backend.StoreBackend(sl, appConfig, cl, input.Store, input.Backend)
 		if err != nil {
 			return err
 		}
diff --git a/core/http/endpoints/openai/realtime.go b/core/http/endpoints/openai/realtime.go
index bee3b52ed..32b08b3fd 100644
--- a/core/http/endpoints/openai/realtime.go
+++ b/core/http/endpoints/openai/realtime.go
@@ -2163,6 +2163,11 @@ type liveResponse struct {
 	output  []types.MessageItemUnion
 	usage   backend.TokenUsage
 	outcome responseOutcome
+	// metadata is echoed back on response.created and response.done. It is the
+	// only thing tying a terminal event to the response.create that asked for
+	// it, which is what lets a client run an out-of-band response alongside the
+	// spoken conversation and still recognise its own answer.
+	metadata map[string]string
 }
 
 func (r *liveResponse) addItem(it types.MessageItemUnion) { r.output = append(r.output, it) }
@@ -2192,12 +2197,16 @@ func triggerResponse(ctx context.Context, session *Session, conv *Conversation,
 	// terminals the legacy code emitted (one response.done per turn, with empty
 	// Output/Usage) are gone; tool turns are now internal to this single response.
 	r := &liveResponse{id: generateUniqueID()}
+	if overrides != nil {
+		r.metadata = overrides.Metadata
+	}
 	sendEvent(t, types.ResponseCreatedEvent{
 		ServerEventBase: types.ServerEventBase{},
 		Response: types.Response{
-			ID:     r.id,
-			Object: "realtime.response",
-			Status: types.ResponseStatusInProgress,
+			ID:       r.id,
+			Object:   "realtime.response",
+			Status:   types.ResponseStatusInProgress,
+			Metadata: r.metadata,
 		},
 	})
 
@@ -2208,10 +2217,11 @@ func triggerResponse(ctx context.Context, session *Session, conv *Conversation,
 		sendEvent(t, types.ResponseDoneEvent{
 			ServerEventBase: types.ServerEventBase{},
 			Response: types.Response{
-				ID:     r.id,
-				Object: "realtime.response",
-				Status: types.ResponseStatusCancelled,
-				Output: r.output,
+				ID:       r.id,
+				Object:   "realtime.response",
+				Status:   types.ResponseStatusCancelled,
+				Output:   r.output,
+				Metadata: r.metadata,
 			},
 		})
 	case outcomeFailed:
@@ -2221,11 +2231,12 @@ func triggerResponse(ctx context.Context, session *Session, conv *Conversation,
 		sendEvent(t, types.ResponseDoneEvent{
 			ServerEventBase: types.ServerEventBase{},
 			Response: types.Response{
-				ID:     r.id,
-				Object: "realtime.response",
-				Status: types.ResponseStatusCompleted,
-				Output: r.output,
-				Usage:  responseUsage(r.usage),
+				ID:       r.id,
+				Object:   "realtime.response",
+				Status:   types.ResponseStatusCompleted,
+				Output:   r.output,
+				Usage:    responseUsage(r.usage),
+				Metadata: r.metadata,
 			},
 		})
 	}
diff --git a/core/http/endpoints/openai/realtime_stream_test.go b/core/http/endpoints/openai/realtime_stream_test.go
index 439f3240e..2d5d7d7a1 100644
--- a/core/http/endpoints/openai/realtime_stream_test.go
+++ b/core/http/endpoints/openai/realtime_stream_test.go
@@ -263,4 +263,64 @@ var _ = Describe("triggerResponse", func() {
 		Expect(done.Response.Usage.OutputTokens).To(Equal(3))
 		Expect(done.Response.Usage.TotalTokens).To(Equal(8))
 	})
+
+	// response.metadata is the only thing tying a terminal event back to the
+	// response.create that asked for it. Without the echo, a client running an
+	// out-of-band response alongside the spoken conversation cannot tell its own
+	// answer from the conversation's, and blocks until it times out.
+	It("echoes response.create metadata back on response.created and response.done", func() {
+		m := &fakeModel{
+			cfg:         &config.ModelConfig{},
+			predictResp: backend.LLMResponse{Response: "Hi there."},
+		}
+		session := &Session{
+			OutputSampleRate: 24000,
+			ModelInterface:   m,
+			ModelConfig:      &config.ModelConfig{},
+			OutputModalities: []types.Modality{types.ModalityText},
+		}
+		t := &fakeTransport{}
+
+		triggerResponse(context.Background(), session, &Conversation{}, t, &types.ResponseCreateParams{
+			Metadata: map[string]string{"client_run": "abc123"},
+		})
+
+		var created *types.ResponseCreatedEvent
+		var done *types.ResponseDoneEvent
+		for i := range t.events {
+			switch e := t.events[i].(type) {
+			case types.ResponseCreatedEvent:
+				created = &e
+			case types.ResponseDoneEvent:
+				done = &e
+			}
+		}
+		Expect(created).NotTo(BeNil())
+		Expect(created.Response.Metadata).To(HaveKeyWithValue("client_run", "abc123"))
+		Expect(done).NotTo(BeNil())
+		Expect(done.Response.Metadata).To(HaveKeyWithValue("client_run", "abc123"))
+	})
+
+	// Omitted rather than sent as an empty object, matching the omitempty tag.
+	It("sends no metadata when response.create carried none", func() {
+		m := &fakeModel{
+			cfg:         &config.ModelConfig{},
+			predictResp: backend.LLMResponse{Response: "Hi there."},
+		}
+		session := &Session{
+			OutputSampleRate: 24000,
+			ModelInterface:   m,
+			ModelConfig:      &config.ModelConfig{},
+			OutputModalities: []types.Modality{types.ModalityText},
+		}
+		t := &fakeTransport{}
+
+		triggerResponse(context.Background(), session, &Conversation{}, t, nil)
+
+		for i := range t.events {
+			if d, ok := t.events[i].(types.ResponseDoneEvent); ok {
+				Expect(d.Response.Metadata).To(BeEmpty())
+			}
+		}
+	})
 })
diff --git a/core/http/react-ui/e2e/page-render-smoke.spec.js b/core/http/react-ui/e2e/page-render-smoke.spec.js
index 27f006b98..3f24752bf 100644
--- a/core/http/react-ui/e2e/page-render-smoke.spec.js
+++ b/core/http/react-ui/e2e/page-render-smoke.spec.js
@@ -14,6 +14,7 @@ import { test, expect } from './coverage-fixtures.js'
 // not bounce to a gate redirect (/login or back to /app home).
 const PAGES = [
   ['/app/talk', 'Talk'],
+  ['/app/3d', '3D Generation'],
   ['/app/usage', 'Usage'],
   ['/app/account', 'Account'],
   ['/app/studio', 'Studio'],
diff --git a/core/http/react-ui/e2e/threed-gen.spec.js b/core/http/react-ui/e2e/threed-gen.spec.js
new file mode 100644
index 000000000..b10270e19
--- /dev/null
+++ b/core/http/react-ui/e2e/threed-gen.spec.js
@@ -0,0 +1,319 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// 3D generation page: mock the capabilities + generation endpoints, feed a
+// real (tiny) Form-B GLB through the parser/viewer, and exercise the
+// IndexedDB-backed history. All assertions are DOM/text — never pixels — so
+// the suite passes with or without working WebGL2 in headless Chromium.
+
+function mockCapabilities(page) {
+  return page.route('**/api/models/capabilities', (route) => {
+    route.fulfill({
+      contentType: 'application/json',
+      body: JSON.stringify({ data: [{ id: 'trellis-test-model', capabilities: ['FLAG_3D'] }] }),
+    })
+  })
+}
+
+// A valid one-triangle GLB in trellis2cpp's vertex-PBR form: POSITION/NORMAL
+// f32, COLOR_0 u16-normalized VEC4, _METALLIC_ROUGHNESS u8-normalized VEC2,
+// u32 indices — the layout mesh_export.cpp's write_vertex_glb emits.
+function buildTinyGlb() {
+  const positions = new Float32Array([0, 0, 0, 1, 0, 0, 0, 1, 0])
+  const normals = new Float32Array([0, 0, 1, 0, 0, 1, 0, 0, 1])
+  const colors = new Uint16Array([
+    65535, 0, 0, 65535,
+    0, 65535, 0, 65535,
+    0, 0, 65535, 65535,
+  ])
+  const metalRough = new Uint8Array([0, 153, 0, 153, 0, 153])
+  const indices = new Uint32Array([0, 1, 2])
+
+  const views = []
+  let binLength = 0
+  const addView = (typed) => {
+    const byteOffset = binLength
+    views.push({ buffer: 0, byteOffset, byteLength: typed.byteLength, target: 34962 })
+    binLength += typed.byteLength
+    binLength += (4 - (binLength % 4)) % 4
+    return views.length - 1
+  }
+  addView(positions); addView(normals); addView(colors); addView(metalRough)
+  const idxView = addView(indices)
+  views[idxView].target = 34963
+
+  const json = {
+    asset: { version: '2.0', generator: 'threed-gen.spec' },
+    scene: 0,
+    scenes: [{ nodes: [0] }],
+    nodes: [{ mesh: 0 }],
+    meshes: [{ primitives: [{ attributes: { POSITION: 0, NORMAL: 1, COLOR_0: 2, _METALLIC_ROUGHNESS: 3 }, indices: 4, material: 0 }] }],
+    materials: [{ pbrMetallicRoughness: { baseColorFactor: [1, 1, 1, 1], metallicFactor: 0, roughnessFactor: 0.6 }, doubleSided: true }],
+    accessors: [
+      { bufferView: 0, componentType: 5126, count: 3, type: 'VEC3', min: [0, 0, 0], max: [1, 1, 0] },
+      { bufferView: 1, componentType: 5126, count: 3, type: 'VEC3' },
+      { bufferView: 2, componentType: 5123, normalized: true, count: 3, type: 'VEC4' },
+      { bufferView: 3, componentType: 5121, normalized: true, count: 3, type: 'VEC2' },
+      { bufferView: 4, componentType: 5125, count: 3, type: 'SCALAR' },
+    ],
+    bufferViews: views,
+    buffers: [{ byteLength: binLength }],
+  }
+
+  const bin = Buffer.alloc(binLength)
+  const parts = [positions, normals, colors, metalRough, indices]
+  for (let i = 0; i < parts.length; i++) {
+    Buffer.from(parts[i].buffer, parts[i].byteOffset, parts[i].byteLength).copy(bin, views[i].byteOffset)
+  }
+
+  let jsonText = JSON.stringify(json)
+  while (jsonText.length % 4 !== 0) jsonText += ' '
+  const jsonBuf = Buffer.from(jsonText)
+
+  const total = 12 + 8 + jsonBuf.length + 8 + bin.length
+  const glb = Buffer.alloc(total)
+  glb.writeUInt32LE(0x46546c67, 0)   // magic 'glTF'
+  glb.writeUInt32LE(2, 4)
+  glb.writeUInt32LE(total, 8)
+  glb.writeUInt32LE(jsonBuf.length, 12)
+  glb.writeUInt32LE(0x4e4f534a, 16)  // 'JSON'
+  jsonBuf.copy(glb, 20)
+  const binHeader = 20 + jsonBuf.length
+  glb.writeUInt32LE(bin.length, binHeader)
+  glb.writeUInt32LE(0x004e4942, binHeader + 4) // 'BIN\0'
+  bin.copy(glb, binHeader + 8)
+  return glb
+}
+
+// 1x1 transparent PNG for the conditioning-image upload.
+const TINY_PNG = Buffer.from(
+  'iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNkYPhfDwAChwGA60e6kgAAAABJRU5ErkJggg==',
+  'base64',
+)
+
+function mockGeneration(page, onRequest) {
+  return page.route('**/3d/generations', (route) => {
+    if (route.request().method() !== 'POST') return route.continue()
+    onRequest?.(route.request().postDataJSON())
+    route.fulfill({
+      contentType: 'application/json',
+      body: JSON.stringify({
+        created: Math.floor(Date.now() / 1000),
+        data: [{ url: '/generated-3d/test.glb' }],
+      }),
+    })
+  })
+}
+
+function mockGlbDownload(page) {
+  return page.route('**/generated-3d/test.glb', (route) => {
+    route.fulfill({
+      status: 200,
+      headers: { 'Content-Type': 'model/gltf-binary' },
+      body: buildTinyGlb(),
+    })
+  })
+}
+
+async function generateOnce(page) {
+  await page.goto('/app/3d')
+  await expect(page.getByRole('button', { name: 'trellis-test-model' })).toBeVisible({ timeout: 10_000 })
+  await page.locator('#threed-image-file').setInputFiles({
+    name: 'input.png',
+    mimeType: 'image/png',
+    buffer: TINY_PNG,
+  })
+  await page.locator('button[type="submit"]').click()
+}
+
+test.describe('3D generation', () => {
+  test.beforeEach(async ({ page }) => {
+    await mockCapabilities(page)
+    await mockGlbDownload(page)
+  })
+
+  test('generates, shows mesh stats, and offers a GLB download', async ({ page }) => {
+    let requestBody = null
+    await mockGeneration(page, (body) => { requestBody = body })
+
+    await generateOnce(page)
+
+    // Stats render from the parsed GLB even without working GL.
+    await expect(page.getByTestId('glb-stats')).toContainText('3', { timeout: 15_000 })
+    await expect(page.getByTestId('glb-stats')).toContainText('1')
+    await expect(page.getByTestId('glb-stats')).toContainText('PBR')
+
+    const download = page.getByTestId('glb-download')
+    await expect(download).toBeVisible()
+    await expect(download).toHaveAttribute('href', /^blob:/)
+    await expect(download).toHaveAttribute('download', /\.glb$/)
+
+    expect(requestBody.model).toBe('trellis-test-model')
+    expect(requestBody.image).toBeTruthy()
+    expect(requestBody.quality).toBe('auto')
+    expect(requestBody.background).toBe('auto')
+    expect(requestBody.response_format).toBe('url')
+  })
+
+  test('advanced settings map to step/texture_steps/cfg_scale/seed', async ({ page }) => {
+    let requestBody = null
+    await mockGeneration(page, (body) => { requestBody = body })
+
+    await page.goto('/app/3d')
+    await expect(page.getByRole('button', { name: 'trellis-test-model' })).toBeVisible({ timeout: 10_000 })
+    await page.locator('#threed-image-file').setInputFiles({ name: 'input.png', mimeType: 'image/png', buffer: TINY_PNG })
+
+    await page.locator('select').first().selectOption('512')
+    await page.getByRole('button', { name: /Advanced Settings/ }).click()
+    const advanced = page.locator('#threed-advanced-options')
+    await advanced.locator('input').nth(0).fill('20')   // steps
+    await advanced.locator('input').nth(1).fill('8')    // texture steps
+    await advanced.locator('input').nth(2).fill('5.5')  // guidance
+    await advanced.locator('input').nth(3).fill('42')   // seed
+    await page.locator('button[type="submit"]').click()
+
+    await expect(page.getByTestId('glb-download')).toBeVisible({ timeout: 15_000 })
+    expect(requestBody.quality).toBe('512')
+    expect(requestBody.step).toBe(20)
+    expect(requestBody.texture_steps).toBe(8)
+    expect(requestBody.cfg_scale).toBe(5.5)
+    expect(requestBody.seed).toBe(42)
+  })
+
+  test('applies a single-detail watertight remesh and previews it before download', async ({ page }) => {
+    await mockGeneration(page)
+    let remeshRequest = null
+    await page.route('**/3d/remesh', (route) => {
+      remeshRequest = route.request()
+      route.fulfill({
+        status: 200,
+        headers: { 'Content-Type': 'model/gltf-binary' },
+        body: buildTinyGlb(),
+      })
+    })
+
+    await generateOnce(page)
+    await expect(page.getByTestId('glb-remesh')).toBeVisible({ timeout: 15_000 })
+    await page.getByLabel('Remesh detail').fill('100')
+    await page.getByTestId('glb-remesh').click()
+
+    await expect(page.getByTestId('glb-remesh')).toContainText('Show original')
+    await expect(page.getByTestId('glb-download')).toHaveAttribute('download', /-remeshed\.glb$/)
+    expect(remeshRequest).not.toBeNull()
+    expect(remeshRequest.method()).toBe('POST')
+    expect(remeshRequest.headers()['content-type']).toContain('multipart/form-data')
+    const multipart = remeshRequest.postDataBuffer().toString()
+    expect(multipart).toContain('trellis-test-model')
+    expect(multipart).toContain('0.35')
+
+    await page.getByTestId('glb-remesh').click()
+    await expect(page.getByTestId('glb-remesh')).toContainText('Apply remeshing')
+    await expect(page.getByTestId('glb-download')).not.toHaveAttribute('download', /-remeshed\.glb$/)
+  })
+
+  test('history entry persists across navigation and reloads into the viewer', async ({ page }) => {
+    await mockGeneration(page)
+    await generateOnce(page)
+    await expect(page.getByTestId('media-history-item')).toHaveCount(1, { timeout: 15_000 })
+
+    // IndexedDB persists within the browser context — navigate away and back.
+    await page.goto('/app')
+    await page.goto('/app/3d')
+    await expect(page.getByTestId('media-history-item')).toHaveCount(1, { timeout: 15_000 })
+
+    // Selecting the entry loads the stored Blob back into the viewer.
+    await page.getByTestId('media-history-item').click()
+    await expect(page.getByTestId('glb-stats')).toBeVisible({ timeout: 15_000 })
+    await expect(page.getByTestId('glb-download')).toHaveAttribute('href', /^blob:/)
+  })
+
+  test('deleting a history entry removes it', async ({ page }) => {
+    await mockGeneration(page)
+    await generateOnce(page)
+    await expect(page.getByTestId('media-history-item')).toHaveCount(1, { timeout: 15_000 })
+
+    await page.getByTestId('media-history-delete').click()
+    await expect(page.getByTestId('media-history-item')).toHaveCount(0)
+  })
+
+  test('API errors surface through the trace link error box', async ({ page }) => {
+    await page.route('**/3d/generations', (route) => {
+      route.fulfill({
+        status: 500,
+        contentType: 'application/json',
+        body: JSON.stringify({ error: { message: 'trellis2 pipeline load: missing files' } }),
+      })
+    })
+
+    await generateOnce(page)
+    await expect(page.locator('.media-result')).toContainText(/missing files|error/i, { timeout: 15_000 })
+  })
+
+  test('rejects oversized conditioning images before making an API request', async ({ page }) => {
+    let requested = false
+    await mockGeneration(page, () => { requested = true })
+    await page.goto('/app/3d')
+    await expect(page.getByRole('button', { name: 'trellis-test-model' })).toBeVisible({ timeout: 10_000 })
+
+    await page.locator('#threed-image-file').setInputFiles({
+      name: 'oversized.png',
+      mimeType: 'image/png',
+      buffer: Buffer.alloc(32 * 1024 * 1024 + 1),
+    })
+
+    await expect(page.locator('.toast')).toContainText('32 MiB limit')
+    expect(requested).toBe(false)
+  })
+
+  test('hides and guards 3D generation when the feature is disabled', async ({ page }) => {
+    await page.route('**/api/auth/status', (route) => {
+      route.fulfill({
+        contentType: 'application/json',
+        body: JSON.stringify({
+          authEnabled: true,
+          staticApiKeyRequired: false,
+          user: { id: 'restricted-user', role: 'user', permissions: { '3d': false } },
+        }),
+      })
+    })
+
+    await page.goto('/app/studio?tab=threed')
+    await expect(page.getByRole('button', { name: '3D', exact: true })).toHaveCount(0)
+    await expect(page.locator('.studio-tab', { hasText: 'Images' })).toHaveClass(/studio-tab-active/)
+
+    await page.goto('/app/3d')
+    await expect(page).toHaveURL(/\/app\/?$/)
+  })
+
+  test.describe('touch input', () => {
+    test.use({ hasTouch: true })
+
+    test('accepts two-finger touch gestures in the GLB viewer', async ({ page }) => {
+      await mockGeneration(page)
+      await generateOnce(page)
+      const canvas = page.getByTestId('glb-canvas')
+      await expect(canvas).toBeVisible({ timeout: 15_000 })
+      await canvas.scrollIntoViewIfNeeded()
+      const box = await canvas.boundingBox()
+      expect(box).not.toBeNull()
+
+      const client = await page.context().newCDPSession(page)
+      await client.send('Input.dispatchTouchEvent', {
+        type: 'touchStart',
+        touchPoints: [
+          { id: 1, x: box.x + box.width * 0.35, y: box.y + box.height * 0.5 },
+          { id: 2, x: box.x + box.width * 0.65, y: box.y + box.height * 0.5 },
+        ],
+      })
+      await client.send('Input.dispatchTouchEvent', {
+        type: 'touchMove',
+        touchPoints: [
+          { id: 1, x: box.x + box.width * 0.3, y: box.y + box.height * 0.45 },
+          { id: 2, x: box.x + box.width * 0.7, y: box.y + box.height * 0.55 },
+        ],
+      })
+      await client.send('Input.dispatchTouchEvent', { type: 'touchEnd', touchPoints: [] })
+
+      await expect(page.getByRole('button', { name: 'Auto-rotate' })).toHaveClass(/btn-secondary/)
+    })
+  })
+})
diff --git a/core/http/react-ui/public/locales/en/media.json b/core/http/react-ui/public/locales/en/media.json
index 7beefe10b..d3e9484f2 100644
--- a/core/http/react-ui/public/locales/en/media.json
+++ b/core/http/react-ui/public/locales/en/media.json
@@ -5,7 +5,8 @@
       "video": "Video",
       "tts": "TTS",
       "sound": "Sound",
-      "transform": "Transform"
+      "transform": "Transform",
+      "threed": "3D"
     }
   },
   "image": {
@@ -70,6 +71,59 @@
       "noResults": "No video generated"
     }
   },
+  "threed": {
+    "title": "3D Generation",
+    "labels": {
+      "model": "Model",
+      "image": "Input image",
+      "quality": "Quality",
+      "quality_auto": "Auto (best available)",
+      "quality_coarse": "Coarse preview (fast)",
+      "quality_512": "512³ fine",
+      "quality_1024": "1024³ cascade (slow)",
+      "background": "Background",
+      "background_auto": "Auto-remove solid background",
+      "background_keep": "Keep original",
+      "background_black": "Remove black",
+      "background_white": "Remove white",
+      "advanced": "Advanced Settings",
+      "steps": "Shape steps",
+      "textureSteps": "Material steps",
+      "guidance": "Guidance",
+      "seed": "Seed",
+      "seedPlaceholder": "Random"
+    },
+    "actions": {
+      "generate": "Generate",
+      "generating": "Generating...",
+      "remesh": "Apply remeshing",
+      "remeshing": "Applying remeshing...",
+      "showOriginal": "Show original",
+      "download": "Download GLB"
+    },
+    "remesh": {
+      "title": "Watertight print remesh",
+      "detail": "Remesh detail",
+      "coarser": "Coarser · faster",
+      "finer": "Finer · slower",
+      "hint": "Detail controls the smallest preserved features. The enclosing offset follows it automatically.",
+      "ready": "Previewing the watertight remeshed model. This is the version that will be downloaded."
+    },
+    "viewer": {
+      "wireframe": "Wireframe",
+      "autoRotate": "Auto-rotate",
+      "noWebgl": "WebGL2 is not available in this browser — download the GLB to view it elsewhere.",
+      "contextLost": "The 3D view lost the GPU context — reload the page to restore it.",
+      "stats": "{{verts}} vertices · {{tris}} triangles",
+      "hint": "drag: rotate · pinch/wheel: zoom · two-finger/right-drag: pan · double-click: reset"
+    },
+    "empty": "Generated 3D model will appear here",
+    "toasts": {
+      "noImage": "Please provide an input image",
+      "noModel": "Please select a model",
+      "noResults": "No model generated"
+    }
+  },
   "tts": {
     "title": "Text to Speech",
     "labels": {
diff --git a/core/http/react-ui/public/locales/en/models.json b/core/http/react-ui/public/locales/en/models.json
index c9a7ab4e9..2e2f21e57 100644
--- a/core/http/react-ui/public/locales/en/models.json
+++ b/core/http/react-ui/public/locales/en/models.json
@@ -29,6 +29,7 @@
     "llm": "Chat",
     "image": "Image",
     "video": "Video",
+    "threed": "3D",
     "multimodal": "Multimodal",
     "vision": "Vision",
     "tts": "TTS",
diff --git a/core/http/react-ui/src/App.css b/core/http/react-ui/src/App.css
index 68a7e4d5a..b27ec01a7 100644
--- a/core/http/react-ui/src/App.css
+++ b/core/http/react-ui/src/App.css
@@ -5166,6 +5166,45 @@ button.collapsible-header:focus-visible {
   border-radius: var(--radius-md);
 }
 
+.threed-remesh-controls {
+  display: flex;
+  flex-direction: column;
+  gap: var(--spacing-xs);
+  padding: var(--spacing-md);
+  background: var(--color-surface-raised);
+  border: 1px solid var(--color-border-subtle);
+  border-radius: var(--radius-md);
+}
+
+.threed-remesh-heading,
+.threed-remesh-scale {
+  display: flex;
+  justify-content: space-between;
+  gap: var(--spacing-md);
+}
+
+.threed-remesh-heading {
+  color: var(--color-text-primary);
+  font-size: var(--text-sm);
+  font-weight: var(--font-weight-medium);
+}
+
+.threed-remesh-heading output,
+.threed-remesh-scale,
+.threed-remesh-ready {
+  color: var(--color-text-muted);
+  font-size: var(--text-xs);
+}
+
+.threed-remesh-controls input[type="range"] {
+  width: 100%;
+}
+
+.threed-remesh-ready {
+  margin: var(--spacing-xs) 0 0;
+  color: var(--color-success);
+}
+
 .media-result-grid {
   display: grid;
   grid-template-columns: repeat(auto-fill, minmax(200px, 1fr));
diff --git a/core/http/react-ui/src/components/GlbViewer.jsx b/core/http/react-ui/src/components/GlbViewer.jsx
new file mode 100644
index 000000000..ead0c8a32
--- /dev/null
+++ b/core/http/react-ui/src/components/GlbViewer.jsx
@@ -0,0 +1,636 @@
+import { useEffect, useRef, useState } from 'react'
+import { useTranslation } from 'react-i18next'
+import { parseGlb } from '../utils/glb'
+
+/* ── WebGL2 GLB viewer ──────────────────────────────────────────────────────
+ * Ported from the trellis2cpp demo server's hand-rolled viewer
+ * (localai-org/trellis2cpp server/web/index.html): quaternion trackball,
+ * metallic-roughness PBR with a procedural environment, ACES tonemapping,
+ * hidden-line wireframe. Kept dependency-free — the renderer is tuned for
+ * the dense unoriented dual-grid meshes TRELLIS.2 produces, which generic
+ * GLTF viewers shade poorly.
+ *
+ * Deltas from the demo: input is a parsed GLB (see utils/glb.js) instead of
+ * the demo's T2MESH stream; COLOR_0 arrives LINEAR (no gamma decode — the
+ * demo's pow(2.2) would double-darken it); vertex attributes upload as
+ * normalized integers straight from the GLB buffers; an optional UV-texture
+ * path covers the atlas-baked form; the base orientation drops the demo's
+ * Z-up -> Y-up turn because baked GLBs are already Y-up.
+ */
+
+const VS = `#version 300 es
+layout(location=0) in vec3 pos;
+layout(location=1) in vec3 nrm;
+layout(location=2) in vec4 baseColor;
+layout(location=3) in vec2 metalRough;
+layout(location=4) in vec2 uv;
+uniform mat4 mvp, viewModel;
+out vec3 vN; out vec3 vV; out vec4 vCol; out vec2 vMR; out vec2 vUV;
+void main() {
+  gl_Position = mvp * vec4(pos, 1.0);
+  vec4 vp = viewModel * vec4(pos, 1.0);
+  vV  = -vp.xyz;                    // view-space: to-camera (camera at origin)
+  vN  = mat3(viewModel) * nrm;      // view-space normal (orbits with camera)
+  vCol = baseColor;
+  vMR = metalRough;
+  vUV = uv;
+}`
+
+// Lightweight metallic-roughness PBR in view space. The dual-grid mesh has
+// unoriented winding (faithful to TRELLIS.2), so every normal is faced toward
+// the camera. No IBL — a procedural environment + one studio key light;
+// ACES tonemapping at the end.
+const FS = `#version 300 es
+precision highp float;
+in vec3 vN; in vec3 vV; in vec4 vCol; in vec2 vMR; in vec2 vUV;
+uniform int wire;
+uniform int colorMode;  // 0 = untextured grey, 1 = vertex PBR (linear COLOR_0), 2 = UV textures
+uniform sampler2D baseColorTex;
+uniform sampler2D mrTex;
+uniform float uMetallicFactor, uRoughnessFactor;
+out vec4 frag;
+// procedural environment radiance in a direction (view space; +y = up)
+vec3 envColor(vec3 d) {
+  vec3 sky = mix(vec3(0.32, 0.40, 0.55), vec3(0.72, 0.80, 0.98), clamp(d.y, 0.0, 1.0));
+  vec3 gnd = vec3(0.14, 0.13, 0.12);
+  return mix(gnd, sky, smoothstep(-0.30, 0.12, d.y));
+}
+void main() {
+  if (wire == 1) { frag = vec4(0.30, 0.64, 1.00, 1.0); return; }
+  vec3 N = normalize(vN), V = normalize(vV);
+  if (dot(N, V) < 0.0) N = -N;   // face the camera (mesh is unoriented)
+  vec3 base = vec3(0.62, 0.66, 0.72);
+  float metal = 0.0, rough = 0.5, opacity = 1.0;
+  if (colorMode == 1) {
+    // COLOR_0 is stored linear in the GLB — use it directly.
+    base = clamp(vCol.rgb, 0.0, 1.0);
+    metal = clamp(vMR.x, 0.0, 1.0);
+    rough = clamp(vMR.y, 0.06, 1.0);
+    opacity = clamp(vCol.a, 0.0, 1.0);
+  } else if (colorMode == 2) {
+    vec4 bc = texture(baseColorTex, vUV);
+    base = pow(clamp(bc.rgb, 0.0, 1.0), vec3(2.2));  // baseColorTexture is sRGB
+    opacity = bc.a;
+    vec3 mr = texture(mrTex, vUV).rgb;               // G=roughness, B=metallic
+    metal = clamp(mr.b * uMetallicFactor, 0.0, 1.0);
+    rough = clamp(mr.g * uRoughnessFactor, 0.06, 1.0);
+  }
+  if (opacity < 0.01) discard;
+  float nv = max(dot(N, V), 1e-3);
+
+  // Schlick Fresnel (grazing reflectance rises to 1 - roughness for metals).
+  vec3 F0 = mix(vec3(0.04), base, metal);
+  vec3 F  = F0 + (max(vec3(1.0 - rough), F0) - F0) * pow(1.0 - nv, 5.0);
+
+  // diffuse: hemispheric environment irradiance (metals have no diffuse)
+  vec3 irr = envColor(N) * 0.55 + vec3(0.12);
+  vec3 diffuse = base * (1.0 - metal) * irr;
+
+  // specular: environment reflection, blurred toward a flat tint by roughness
+  vec3 refl = mix(envColor(reflect(-V, N)), vec3(0.34, 0.37, 0.44), rough * rough);
+  vec3 specular = refl * F;
+
+  // one crisp studio key light for a lively highlight
+  vec3 L = normalize(vec3(0.45, 0.70, 0.55)), H = normalize(L + V);
+  float shin = mix(8.0, 260.0, pow(1.0 - rough, 2.0));
+  float sp = pow(max(dot(N, H), 0.0), shin) * (shin + 2.0) / 6.2831853;
+  vec3  kc = vec3(1.00, 0.96, 0.88);
+  float ndl = max(dot(N, L), 0.0);
+  specular += kc * sp * F * ndl;
+  diffuse  += base * (1.0 - metal) * kc * ndl * 0.28;
+
+  vec3 color = diffuse + specular;
+  color = (color * (2.51 * color + 0.03)) / (color * (2.43 * color + 0.59) + 0.14); // ACES
+  frag = vec4(pow(clamp(color, 0.0, 1.0), vec3(1.0/2.2)), opacity);
+}`
+
+/* trackball orientation (quaternion): each drag composes a small rotation
+ * about the screen axes onto the current orientation — no gimbal lock. */
+const Q = {
+  axisAngle(x, y, z, a) { const h = a * 0.5, s = Math.sin(h); return [x * s, y * s, z * s, Math.cos(h)] },
+  mul(a, b) {   // Hamilton product a·b (apply b, then a)
+    return [
+      a[3] * b[0] + a[0] * b[3] + a[1] * b[2] - a[2] * b[1],
+      a[3] * b[1] - a[0] * b[2] + a[1] * b[3] + a[2] * b[0],
+      a[3] * b[2] + a[0] * b[1] - a[1] * b[0] + a[2] * b[3],
+      a[3] * b[3] - a[0] * b[0] - a[1] * b[1] - a[2] * b[2],
+    ]
+  },
+  norm(q) { const n = Math.hypot(q[0], q[1], q[2], q[3]) || 1; return [q[0] / n, q[1] / n, q[2] / n, q[3] / n] },
+  toMat4(q) {   // column-major rotation matrix
+    const [x, y, z, w] = q
+    const xx = x * x, yy = y * y, zz = z * z, xy = x * y, xz = x * z, yz = y * z, wx = w * x, wy = w * y, wz = w * z
+    return new Float32Array([
+      1 - 2 * (yy + zz), 2 * (xy + wz), 2 * (xz - wy), 0,
+      2 * (xy - wz), 1 - 2 * (xx + zz), 2 * (yz + wx), 0,
+      2 * (xz + wy), 2 * (yz - wx), 1 - 2 * (xx + yy), 0,
+      0, 0, 0, 1,
+    ])
+  },
+}
+
+// GLBs are already Y-up (the baker swaps axes on export), so unlike the demo
+// there is no Z-up correction here — just a gentle 3/4 default view.
+const QBASE = Q.norm(Q.mul(Q.axisAngle(1, 0, 0, -0.30), Q.axisAngle(0, 1, 0, 0.55)))
+
+/* minimal mat4 helpers (column-major) */
+const M = {
+  mul(a, b) {
+    const o = new Float32Array(16)
+    for (let c = 0; c < 4; ++c) for (let r = 0; r < 4; ++r)
+      o[c * 4 + r] = a[r] * b[c * 4] + a[4 + r] * b[c * 4 + 1] + a[8 + r] * b[c * 4 + 2] + a[12 + r] * b[c * 4 + 3]
+    return o
+  },
+  persp(fov, asp, near, far) {
+    const f = 1 / Math.tan(fov / 2), o = new Float32Array(16)
+    o[0] = f / asp; o[5] = f
+    o[10] = (far + near) / (near - far); o[11] = -1
+    o[14] = 2 * far * near / (near - far)
+    return o
+  },
+  trans(x, y, z) {
+    const o = new Float32Array(16)
+    o[0] = o[5] = o[10] = o[15] = 1
+    o[12] = x; o[13] = y; o[14] = z
+    return o
+  },
+  scale(s) {
+    const o = new Float32Array(16)
+    o[0] = o[5] = o[10] = s; o[15] = 1
+    return o
+  },
+}
+
+function glType(gl, array) {
+  if (array instanceof Uint8Array) return gl.UNSIGNED_BYTE
+  if (array instanceof Uint16Array) return gl.UNSIGNED_SHORT
+  return gl.FLOAT
+}
+
+export function createGlbViewer(canvas, { onContextLost } = {}) {
+  const gl = canvas.getContext('webgl2', { antialias: true })
+  if (!gl) return null
+
+  // If the GPU context is lost (driver reset / out of memory), preventDefault
+  // keeps it recoverable and the page shows a notice instead of a dead canvas.
+  const contextLost = (e) => {
+    e.preventDefault()
+    if (onContextLost) onContextLost()
+  }
+  canvas.addEventListener('webglcontextlost', contextLost, false)
+
+  function shader(type, src) {
+    const s = gl.createShader(type)
+    gl.shaderSource(s, src); gl.compileShader(s)
+    if (!gl.getShaderParameter(s, gl.COMPILE_STATUS)) throw new Error(gl.getShaderInfoLog(s))
+    return s
+  }
+  const prog = gl.createProgram()
+  gl.attachShader(prog, shader(gl.VERTEX_SHADER, VS))
+  gl.attachShader(prog, shader(gl.FRAGMENT_SHADER, FS))
+  gl.linkProgram(prog)
+  if (!gl.getProgramParameter(prog, gl.LINK_STATUS)) throw new Error(gl.getProgramInfoLog(prog))
+  const uMVP = gl.getUniformLocation(prog, 'mvp')
+  const uViewModel = gl.getUniformLocation(prog, 'viewModel')
+  const uWire = gl.getUniformLocation(prog, 'wire')
+  const uColorMode = gl.getUniformLocation(prog, 'colorMode')
+  const uBaseColorTex = gl.getUniformLocation(prog, 'baseColorTex')
+  const uMrTex = gl.getUniformLocation(prog, 'mrTex')
+  const uMetallicFactor = gl.getUniformLocation(prog, 'uMetallicFactor')
+  const uRoughnessFactor = gl.getUniformLocation(prog, 'uRoughnessFactor')
+
+  const vao = gl.createVertexArray()
+  const vbo = gl.createBuffer(), nbo = gl.createBuffer(), cbo = gl.createBuffer()
+  const mbo = gl.createBuffer(), ubo = gl.createBuffer(), ibo = gl.createBuffer()
+  const wireIbo = gl.createBuffer()
+  let nIndices = 0, nWire = 0
+  let colorMode = 0
+  let baseColorTexture = null, mrTexture = null
+  let metallicFactor = 1, roughnessFactor = 1
+  // fit-to-view: model = rotation * scale * translate(-center) keeps the demo's
+  // camera constants valid for any GLB extent (trellis meshes are ~unit cube).
+  let center = [0, 0, 0], fitScale = 1
+
+  let rot = QBASE.slice(), dist = 1.8, panX = 0, panY = 0
+  let wire = false, spin = true
+  let disposed = false
+
+  function makeTexture(bitmap) {
+    const tex = gl.createTexture()
+    gl.bindTexture(gl.TEXTURE_2D, tex)
+    // Matches the baker's sampler: linear filtering, clamp, no mipmaps.
+    gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_MIN_FILTER, gl.LINEAR)
+    gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_MAG_FILTER, gl.LINEAR)
+    gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_WRAP_S, gl.CLAMP_TO_EDGE)
+    gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_WRAP_T, gl.CLAMP_TO_EDGE)
+    gl.texImage2D(gl.TEXTURE_2D, 0, gl.RGBA, gl.RGBA, gl.UNSIGNED_BYTE, bitmap)
+    gl.bindTexture(gl.TEXTURE_2D, null)
+    return tex
+  }
+
+  function dropTextures() {
+    if (baseColorTexture) { gl.deleteTexture(baseColorTexture); baseColorTexture = null }
+    if (mrTexture) { gl.deleteTexture(mrTexture); mrTexture = null }
+  }
+
+  async function setMesh(mesh) {
+    gl.bindVertexArray(vao)
+    gl.bindBuffer(gl.ARRAY_BUFFER, vbo)
+    gl.bufferData(gl.ARRAY_BUFFER, mesh.positions, gl.STATIC_DRAW)
+    gl.enableVertexAttribArray(0)
+    gl.vertexAttribPointer(0, 3, gl.FLOAT, false, 0, 0)
+    if (mesh.normals) {
+      gl.bindBuffer(gl.ARRAY_BUFFER, nbo)
+      gl.bufferData(gl.ARRAY_BUFFER, mesh.normals, gl.STATIC_DRAW)
+      gl.enableVertexAttribArray(1)
+      gl.vertexAttribPointer(1, 3, gl.FLOAT, false, 0, 0)
+    } else {
+      gl.disableVertexAttribArray(1)
+      gl.vertexAttrib3f(1, 0, 1, 0)
+    }
+
+    dropTextures()
+    colorMode = 0
+    if (mesh.color0) {
+      // Upload COLOR_0 / _METALLIC_ROUGHNESS as normalized integers straight
+      // from the GLB buffers — no CPU conversion of multi-million-vertex data.
+      gl.bindBuffer(gl.ARRAY_BUFFER, cbo)
+      gl.bufferData(gl.ARRAY_BUFFER, mesh.color0.array, gl.STATIC_DRAW)
+      gl.enableVertexAttribArray(2)
+      gl.vertexAttribPointer(2, mesh.color0.size, glType(gl, mesh.color0.array), mesh.color0.normalized, 0, 0)
+      if (mesh.metalRough) {
+        gl.bindBuffer(gl.ARRAY_BUFFER, mbo)
+        gl.bufferData(gl.ARRAY_BUFFER, mesh.metalRough.array, gl.STATIC_DRAW)
+        gl.enableVertexAttribArray(3)
+        gl.vertexAttribPointer(3, mesh.metalRough.size, glType(gl, mesh.metalRough.array), mesh.metalRough.normalized, 0, 0)
+      } else {
+        gl.disableVertexAttribArray(3)
+        gl.vertexAttrib2f(3, 0, 0.6)
+      }
+      gl.disableVertexAttribArray(4)
+      colorMode = 1
+    } else if (mesh.uv && mesh.baseColorPng) {
+      gl.bindBuffer(gl.ARRAY_BUFFER, ubo)
+      gl.bufferData(gl.ARRAY_BUFFER, mesh.uv, gl.STATIC_DRAW)
+      gl.enableVertexAttribArray(4)
+      gl.vertexAttribPointer(4, 2, gl.FLOAT, false, 0, 0)
+      gl.disableVertexAttribArray(2)
+      gl.disableVertexAttribArray(3)
+      const bitmaps = await Promise.all([
+        createImageBitmap(new Blob([mesh.baseColorPng], { type: 'image/png' })),
+        mesh.metalRoughPng ? createImageBitmap(new Blob([mesh.metalRoughPng], { type: 'image/png' })) : null,
+      ])
+      if (disposed) return
+      gl.bindVertexArray(vao)
+      baseColorTexture = makeTexture(bitmaps[0])
+      mrTexture = bitmaps[1] ? makeTexture(bitmaps[1]) : makeTexture(bitmaps[0])
+      metallicFactor = mesh.material.metallicFactor
+      roughnessFactor = mesh.material.roughnessFactor
+      colorMode = 2
+    } else {
+      gl.disableVertexAttribArray(2)
+      gl.disableVertexAttribArray(3)
+      gl.disableVertexAttribArray(4)
+    }
+
+    gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, ibo)
+    gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, mesh.indices, gl.STATIC_DRAW)
+    nIndices = mesh.indices.length
+    // wireframe index buffer: 3 edges per triangle, bounded to WIRE_BUDGET —
+    // a full wireframe of a multi-million-triangle mesh would OOM the GPU and
+    // lose the context (sub-pixel lines also fill in as a solid mass).
+    const WIRE_BUDGET = 12_000_000               // ~6M segments, ~48 MB
+    const nTri = nIndices / 3
+    const wireStride = Math.max(1, Math.ceil(nTri * 6 / WIRE_BUDGET))
+    const wireIdx = new Uint32Array(Math.ceil(nTri / wireStride) * 6)
+    let o = 0
+    const idx = mesh.indices
+    for (let t = 0; t < nIndices; t += 3 * wireStride) {
+      wireIdx[o++] = idx[t]; wireIdx[o++] = idx[t + 1]
+      wireIdx[o++] = idx[t + 1]; wireIdx[o++] = idx[t + 2]
+      wireIdx[o++] = idx[t + 2]; wireIdx[o++] = idx[t]
+    }
+    gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, wireIbo)
+    gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, wireIdx.subarray(0, o), gl.STATIC_DRAW)
+    nWire = o
+    gl.bindVertexArray(null)
+
+    center = [
+      (mesh.bboxMin[0] + mesh.bboxMax[0]) / 2,
+      (mesh.bboxMin[1] + mesh.bboxMax[1]) / 2,
+      (mesh.bboxMin[2] + mesh.bboxMax[2]) / 2,
+    ]
+    const radius = Math.hypot(
+      mesh.bboxMax[0] - mesh.bboxMin[0],
+      mesh.bboxMax[1] - mesh.bboxMin[1],
+      mesh.bboxMax[2] - mesh.bboxMin[2],
+    ) / 2 || 1
+    fitScale = 0.866 / radius
+    resetView()
+  }
+
+  function clear() {
+    nIndices = 0
+    nWire = 0
+    dropTextures()
+  }
+
+  function resetView() {
+    rot = QBASE.slice(); dist = 1.8; panX = panY = 0
+  }
+
+  /* input */
+  let dragging = false, panning = false, lx = 0, ly = 0
+  let pinchDistance = 0, pinchX = 0, pinchY = 0
+  const pointers = new Map()
+  const pointerPair = () => Array.from(pointers.values()).slice(0, 2)
+  const beginPinch = () => {
+    const [a, b] = pointerPair()
+    if (!a || !b) return
+    pinchDistance = Math.hypot(b.x - a.x, b.y - a.y)
+    pinchX = (a.x + b.x) / 2
+    pinchY = (a.y + b.y) / 2
+  }
+  const stopSpin = () => {
+    spin = false
+    if (onSpinChange) onSpinChange(false)
+  }
+  const onPointerDown = (e) => {
+    e.preventDefault()
+    canvas.setPointerCapture(e.pointerId)
+    pointers.set(e.pointerId, { x: e.clientX, y: e.clientY })
+    if (pointers.size === 1) {
+      dragging = true
+      panning = e.button === 2 || e.shiftKey
+      lx = e.clientX; ly = e.clientY
+    } else {
+      dragging = false
+      beginPinch()
+      stopSpin()
+    }
+  }
+  const onPointerUp = (e) => {
+    pointers.delete(e.pointerId)
+    if (canvas.hasPointerCapture(e.pointerId)) canvas.releasePointerCapture(e.pointerId)
+    pinchDistance = 0
+    if (pointers.size === 1) {
+      const remaining = pointers.values().next().value
+      dragging = true
+      panning = false
+      lx = remaining.x; ly = remaining.y
+    } else {
+      dragging = false
+    }
+  }
+  const onPointerMove = (e) => {
+    if (!pointers.has(e.pointerId)) return
+    pointers.set(e.pointerId, { x: e.clientX, y: e.clientY })
+
+    if (pointers.size >= 2) {
+      const [a, b] = pointerPair()
+      const nextDistance = Math.hypot(b.x - a.x, b.y - a.y)
+      const nextX = (a.x + b.x) / 2
+      const nextY = (a.y + b.y) / 2
+      if (pinchDistance > 0 && nextDistance > 0) {
+        dist *= pinchDistance / nextDistance
+        dist = Math.max(0.3, Math.min(8, dist))
+        panX += (nextX - pinchX) * 0.0015 * dist
+        panY -= (nextY - pinchY) * 0.0015 * dist
+      }
+      pinchDistance = nextDistance
+      pinchX = nextX
+      pinchY = nextY
+      return
+    }
+
+    if (!dragging) return
+    const dx = e.clientX - lx, dy = e.clientY - ly
+    lx = e.clientX; ly = e.clientY
+    if (panning) {
+      panX += dx * 0.0015 * dist; panY -= dy * 0.0015 * dist
+    } else {
+      // Compose screen-axis turns onto the current orientation (fixed camera
+      // frame), so rotation stays screen-relative and never locks up.
+      const k = 0.008
+      rot = Q.norm(Q.mul(Q.axisAngle(1, 0, 0, dy * k), Q.mul(Q.axisAngle(0, 1, 0, dx * k), rot)))
+      stopSpin()
+    }
+  }
+  const onContextMenu = (e) => e.preventDefault()
+  const onWheel = (e) => {
+    e.preventDefault()
+    dist *= Math.exp(e.deltaY * 0.001)
+    dist = Math.max(0.3, Math.min(8, dist))
+  }
+  const onDblClick = () => resetView()
+  let onSpinChange = null
+
+  canvas.addEventListener('pointerdown', onPointerDown)
+  canvas.addEventListener('pointerup', onPointerUp)
+  canvas.addEventListener('pointercancel', onPointerUp)
+  canvas.addEventListener('pointermove', onPointerMove)
+  canvas.addEventListener('contextmenu', onContextMenu)
+  canvas.addEventListener('wheel', onWheel, { passive: false })
+  canvas.addEventListener('dblclick', onDblClick)
+
+  gl.enable(gl.DEPTH_TEST)
+  gl.enable(gl.BLEND)
+  gl.blendFunc(gl.SRC_ALPHA, gl.ONE_MINUS_SRC_ALPHA)
+  gl.clearColor(0.063, 0.078, 0.094, 1)
+
+  let rafId = 0
+  let last = performance.now()
+  function frame(now) {
+    if (disposed) return
+    const dt = (now - last) / 1000; last = now
+    // auto-rotate: a slow turn about the screen-vertical axis (turntable feel)
+    if (spin) rot = Q.norm(Q.mul(Q.axisAngle(0, 1, 0, dt * 0.4), rot))
+
+    const w = canvas.clientWidth, h = canvas.clientHeight
+    if (w > 0 && h > 0 && (canvas.width !== w * devicePixelRatio || canvas.height !== h * devicePixelRatio)) {
+      canvas.width = w * devicePixelRatio; canvas.height = h * devicePixelRatio
+    }
+    gl.viewport(0, 0, canvas.width, canvas.height)
+    gl.clear(gl.COLOR_BUFFER_BIT | gl.DEPTH_BUFFER_BIT)
+
+    if (nIndices) {
+      const rotation = Q.toMat4(rot)
+      const model = M.mul(rotation, M.mul(M.scale(fitScale), M.trans(-center[0], -center[1], -center[2])))
+      const view = M.trans(panX, panY, -dist)
+      const viewModel = M.mul(view, model)
+      const proj = M.persp(0.9, (w || 1) / (h || 1), 0.05, 100)
+      const mvp = M.mul(proj, viewModel)
+
+      gl.useProgram(prog)
+      gl.uniformMatrix4fv(uMVP, false, mvp)
+      gl.uniformMatrix4fv(uViewModel, false, viewModel)
+      gl.uniform1f(uMetallicFactor, metallicFactor)
+      gl.uniform1f(uRoughnessFactor, roughnessFactor)
+      if (colorMode === 2) {
+        gl.activeTexture(gl.TEXTURE0)
+        gl.bindTexture(gl.TEXTURE_2D, baseColorTexture)
+        gl.uniform1i(uBaseColorTex, 0)
+        gl.activeTexture(gl.TEXTURE1)
+        gl.bindTexture(gl.TEXTURE_2D, mrTexture)
+        gl.uniform1i(uMrTex, 1)
+      }
+      gl.bindVertexArray(vao)
+      if (wire) {
+        // hidden-line wireframe: a depth-only prepass (pushed back a hair via
+        // polygon offset) occludes back-facing edges, so the front surface's
+        // edges show instead of a see-through blob.
+        gl.enable(gl.POLYGON_OFFSET_FILL)
+        gl.polygonOffset(1.0, 1.0)
+        gl.colorMask(false, false, false, false)
+        gl.uniform1i(uWire, 0)
+        gl.uniform1i(uColorMode, 0)
+        gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, ibo)
+        gl.drawElements(gl.TRIANGLES, nIndices, gl.UNSIGNED_INT, 0)
+        gl.colorMask(true, true, true, true)
+        gl.disable(gl.POLYGON_OFFSET_FILL)
+        gl.uniform1i(uWire, 1)
+        // Front-surface edges sit at ~equal depth to the offset-back fill, so
+        // LEQUAL lets them pass while occluded edges (greater depth) fail.
+        gl.depthFunc(gl.LEQUAL)
+        gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, wireIbo)
+        gl.drawElements(gl.LINES, nWire, gl.UNSIGNED_INT, 0)
+        gl.depthFunc(gl.LESS)
+      } else {
+        gl.uniform1i(uWire, 0)
+        gl.uniform1i(uColorMode, colorMode)
+        gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, ibo)
+        gl.drawElements(gl.TRIANGLES, nIndices, gl.UNSIGNED_INT, 0)
+      }
+      gl.bindVertexArray(null)
+    }
+    rafId = requestAnimationFrame(frame)
+  }
+  rafId = requestAnimationFrame(frame)
+
+  function dispose() {
+    disposed = true
+    cancelAnimationFrame(rafId)
+    canvas.removeEventListener('pointerdown', onPointerDown)
+    canvas.removeEventListener('pointerup', onPointerUp)
+    canvas.removeEventListener('pointercancel', onPointerUp)
+    canvas.removeEventListener('pointermove', onPointerMove)
+    canvas.removeEventListener('contextmenu', onContextMenu)
+    canvas.removeEventListener('wheel', onWheel)
+    canvas.removeEventListener('dblclick', onDblClick)
+    canvas.removeEventListener('webglcontextlost', contextLost)
+    dropTextures()
+    gl.deleteBuffer(vbo); gl.deleteBuffer(nbo); gl.deleteBuffer(cbo)
+    gl.deleteBuffer(mbo); gl.deleteBuffer(ubo); gl.deleteBuffer(ibo)
+    gl.deleteBuffer(wireIbo)
+    gl.deleteVertexArray(vao)
+    gl.deleteProgram(prog)
+    gl.getExtension('WEBGL_lose_context')?.loseContext()
+  }
+
+  return {
+    setMesh,
+    clear,
+    dispose,
+    resetView,
+    setWire(v) { wire = v },
+    setSpin(v) { spin = v },
+    onSpinChanged(fn) { onSpinChange = fn },
+  }
+}
+
+export default function GlbViewer({ blob }) {
+  const { t } = useTranslation('media')
+  const canvasRef = useRef(null)
+  const viewerRef = useRef(null)
+  const [glError, setGlError] = useState(null) // 'no-webgl2' | 'context-lost' | parse error text
+  const [stats, setStats] = useState(null)
+  const [wire, setWire] = useState(false)
+  const [spin, setSpin] = useState(true)
+
+  useEffect(() => {   // GL lifecycle — once per mount
+    let viewer = null
+    try {
+      viewer = createGlbViewer(canvasRef.current, { onContextLost: () => setGlError('context-lost') })
+    } catch {
+      viewer = null
+    }
+    if (!viewer) {
+      setGlError('no-webgl2')
+      return undefined
+    }
+    viewer.onSpinChanged(setSpin)
+    viewerRef.current = viewer
+    return () => {
+      viewerRef.current = null
+      viewer.dispose()
+    }
+  }, [])
+
+  useEffect(() => {   // (re)load when the blob changes
+    if (!blob) {
+      viewerRef.current?.clear()
+      setStats(null)
+      return undefined
+    }
+    let cancelled = false
+    ;(async () => {
+      try {
+        // Parse before touching GL: stats and parse errors surface even when
+        // WebGL2 is unavailable (e.g. headless CI), and the download button
+        // keeps working either way.
+        const mesh = parseGlb(await blob.arrayBuffer())
+        if (cancelled) return
+        setStats({
+          nVerts: mesh.nVerts,
+          nTris: mesh.nTris,
+          pbr: !!(mesh.color0 || mesh.baseColorPng),
+        })
+        if (viewerRef.current) await viewerRef.current.setMesh(mesh)
+      } catch (err) {
+        if (!cancelled) setGlError(err.message)
+      }
+    })()
+    return () => { cancelled = true }
+  }, [blob])
+
+  const toggleWire = () => {
+    const next = !wire
+    setWire(next)
+    viewerRef.current?.setWire(next)
+  }
+  const toggleSpin = () => {
+    const next = !spin
+    setSpin(next)
+    viewerRef.current?.setSpin(next)
+  }
+
+  return (
+    <div className="glb-viewer" style={{ display: 'flex', flexDirection: 'column', gap: 'var(--spacing-sm)', width: '100%' }}>
+      <canvas
+        ref={canvasRef}
+        style={{ width: '100%', aspectRatio: '4 / 3', borderRadius: 'var(--radius-md)', background: '#101418', touchAction: 'none' }}
+        data-testid="glb-canvas"
+      />
+      <div style={{ display: 'flex', alignItems: 'center', gap: 'var(--spacing-sm)', flexWrap: 'wrap' }}>
+        <button type="button" className={`btn btn-sm ${wire ? 'btn-primary' : 'btn-secondary'}`} onClick={toggleWire}>
+          <i className="fas fa-border-none" /> {t('threed.viewer.wireframe')}
+        </button>
+        <button type="button" className={`btn btn-sm ${spin ? 'btn-primary' : 'btn-secondary'}`} onClick={toggleSpin}>
+          <i className="fas fa-rotate" /> {t('threed.viewer.autoRotate')}
+        </button>
+        {stats && (
+          <span style={{ color: 'var(--color-text-muted)', fontSize: '0.85em' }} data-testid="glb-stats">
+            {t('threed.viewer.stats', { verts: stats.nVerts.toLocaleString(), tris: stats.nTris.toLocaleString() })}
+            {stats.pbr ? ' · PBR' : ''}
+          </span>
+        )}
+      </div>
+      {glError === 'no-webgl2' && <p style={{ color: 'var(--color-text-muted)' }}>{t('threed.viewer.noWebgl')}</p>}
+      {glError === 'context-lost' && <p style={{ color: 'var(--color-text-muted)' }}>{t('threed.viewer.contextLost')}</p>}
+      {glError && glError !== 'no-webgl2' && glError !== 'context-lost' && (
+        <p style={{ color: 'var(--color-danger, #e5484d)' }}>{glError}</p>
+      )}
+      <p style={{ color: 'var(--color-text-muted)', fontSize: '0.8em', margin: 0 }}>{t('threed.viewer.hint')}</p>
+    </div>
+  )
+}
diff --git a/core/http/react-ui/src/components/ThreeDHistory.jsx b/core/http/react-ui/src/components/ThreeDHistory.jsx
new file mode 100644
index 000000000..7686669d5
--- /dev/null
+++ b/core/http/react-ui/src/components/ThreeDHistory.jsx
@@ -0,0 +1,76 @@
+import { memo, useState } from 'react'
+import { relativeTime } from '../utils/format'
+import { useTranslation } from 'react-i18next'
+
+// ThreeDHistory — sibling of MediaHistory for IndexedDB-backed 3D entries
+// (see use3DHistory). Reuses MediaHistory's markup, CSS classes, and testids
+// so styling and e2e helpers carry over; the differences are the entry shape
+// (input-image thumbnail, quality subtitle) and the Blob-backed source.
+// Deliberately a plain vertical list — no showcase/gallery mode.
+export default memo(function ThreeDHistory({ entries, selectedId, onSelect, onDelete, onClearAll }) {
+  const { t } = useTranslation('media')
+  const [expanded, setExpanded] = useState(true)
+
+  return (
+    <div className="media-history" data-testid="media-history">
+      <div
+        className={`collapsible-header ${expanded ? 'open' : ''}`}
+        onClick={() => setExpanded(!expanded)}
+        style={{ display: 'flex', alignItems: 'center' }}
+      >
+        <i className="fas fa-chevron-right" />
+        <span style={{ flex: 1 }}>{t('history.title')} ({entries.length})</span>
+        {entries.length > 0 && (
+          <button
+            className="media-history-clear-btn"
+            title={t('history.clearTitle')}
+            onClick={(e) => { e.stopPropagation(); onClearAll() }}
+          >
+            <i className="fas fa-trash" />
+          </button>
+        )}
+      </div>
+      {expanded && (
+        <div className="media-history-list">
+          {entries.length === 0 ? (
+            <div className="media-history-empty">{t('history.empty')}</div>
+          ) : (
+            entries.map(entry => (
+              <div
+                key={entry.id}
+                className={`media-history-item ${selectedId === entry.id ? 'active' : ''}`}
+                onClick={() => onSelect(entry.id)}
+                data-testid="media-history-item"
+              >
+                <div className="media-history-item-thumb">
+                  {entry.inputThumb ? (
+                    <img src={entry.inputThumb} alt="" />
+                  ) : (
+                    <i className="fas fa-cube" />
+                  )}
+                </div>
+                <div className="media-history-item-info">
+                  <div className="media-history-item-top">
+                    <span className="media-history-item-prompt">{entry.name || entry.model}</span>
+                    <span className="media-history-item-time">{relativeTime(entry.createdAt)}</span>
+                  </div>
+                  <div className="media-history-item-model">
+                    {entry.params?.quality ? `${entry.params.quality} · ` : ''}{entry.model}
+                  </div>
+                </div>
+                <button
+                  className="media-history-item-delete"
+                  title={t('history.deleteEntry')}
+                  onClick={(e) => { e.stopPropagation(); onDelete(entry.id) }}
+                  data-testid="media-history-delete"
+                >
+                  <i className="fas fa-times" />
+                </button>
+              </div>
+            ))
+          )}
+        </div>
+      )}
+    </div>
+  )
+})
diff --git a/core/http/react-ui/src/hooks/use3DHistory.js b/core/http/react-ui/src/hooks/use3DHistory.js
new file mode 100644
index 000000000..b15e331ae
--- /dev/null
+++ b/core/http/react-ui/src/hooks/use3DHistory.js
@@ -0,0 +1,132 @@
+import { useState, useEffect, useCallback, useMemo } from 'react'
+import { generateId } from '../utils/format'
+
+// use3DHistory — IndexedDB-backed history of 3D generations. Unlike
+// useMediaHistory (localStorage, URL-only), entries here carry the generated
+// GLB as a Blob: GLBs are multi-megabyte binaries the server may eventually
+// clean up, and IndexedDB is the only browser store that handles blobs of
+// that size. Blobs read back from IndexedDB are lazy handles (the bytes are
+// not materialized into JS memory until used), so a single store is fine.
+//
+// Entry: { id, createdAt, name, model,
+//          params: { seed, steps, textureSteps, guidance, quality, background },
+//          inputThumb,  // small dataURL of the conditioning image
+//          glb }        // Blob
+
+const DB_NAME = 'localai-3d-history'
+const DB_VERSION = 1
+const STORE = 'generations'
+const MAX_ENTRIES = 20
+
+function openDb() {
+  return new Promise((resolve, reject) => {
+    const req = indexedDB.open(DB_NAME, DB_VERSION)
+    req.onupgradeneeded = () => {
+      const db = req.result
+      if (!db.objectStoreNames.contains(STORE)) {
+        db.createObjectStore(STORE, { keyPath: 'id' }).createIndex('createdAt', 'createdAt')
+      }
+    }
+    req.onsuccess = () => resolve(req.result)
+    req.onerror = () => reject(req.error)
+  })
+}
+
+const txDone = (tx) => new Promise((resolve, reject) => {
+  tx.oncomplete = resolve
+  tx.onerror = () => reject(tx.error)
+  tx.onabort = () => reject(tx.error)
+})
+
+async function withStore(mode, fn) {
+  const db = await openDb()
+  try {
+    const tx = db.transaction(STORE, mode)
+    const result = fn(tx.objectStore(STORE))
+    await txDone(tx)
+    return result
+  } finally {
+    db.close()
+  }
+}
+
+async function idbGetAll() {
+  const req = await withStore('readonly', (store) => store.getAll())
+  return (req.result || []).sort((a, b) => b.createdAt - a.createdAt)
+}
+
+// Insert + keep-newest-N eviction in one transaction so a crash between the
+// two can't leave the store unbounded.
+async function idbPutAndEvict(entry) {
+  await withStore('readwrite', (store) => {
+    store.put(entry)
+    // getAllKeys on the createdAt index yields primary keys oldest-first.
+    const keysReq = store.index('createdAt').getAllKeys()
+    keysReq.onsuccess = () => {
+      const excess = keysReq.result.length - MAX_ENTRIES
+      for (let i = 0; i < excess; i++) store.delete(keysReq.result[i])
+    }
+  })
+}
+
+const idbDelete = (id) => withStore('readwrite', (store) => store.delete(id))
+const idbClear = () => withStore('readwrite', (store) => store.clear())
+
+export function use3DHistory() {
+  const [entries, setEntries] = useState([])
+  const [selectedId, setSelectedId] = useState(null)
+
+  const refresh = useCallback(async () => {
+    try {
+      setEntries(await idbGetAll())
+    } catch {
+      // IndexedDB unavailable (private mode etc.) — degrade to session-only.
+      setEntries((prev) => prev)
+    }
+  }, [])
+
+  useEffect(() => { refresh() }, [refresh])
+
+  const addEntry = useCallback(async ({ model, params, inputThumb, glb, name }) => {
+    const entry = { id: generateId(), createdAt: Date.now(), model, params, inputThumb, glb, name }
+    try {
+      await idbPutAndEvict(entry)
+      await refresh()
+    } catch {
+      setEntries((prev) => [entry, ...prev].slice(0, MAX_ENTRIES))
+    }
+    return entry
+  }, [refresh])
+
+  const deleteEntry = useCallback(async (id) => {
+    setSelectedId((prev) => (prev === id ? null : prev))
+    try {
+      await idbDelete(id)
+      await refresh()
+    } catch {
+      setEntries((prev) => prev.filter((e) => e.id !== id))
+    }
+  }, [refresh])
+
+  const clearAll = useCallback(async () => {
+    setSelectedId(null)
+    try {
+      await idbClear()
+    } catch {
+      // fall through to the local reset below
+    }
+    setEntries([])
+  }, [])
+
+  // Toggles: clicking the selected entry deselects it (back to latest result).
+  const selectEntry = useCallback((id) => {
+    setSelectedId((prev) => (prev === id ? null : id))
+  }, [])
+
+  const selectedEntry = useMemo(
+    () => entries.find((e) => e.id === selectedId) || null,
+    [entries, selectedId],
+  )
+
+  return { entries, addEntry, deleteEntry, clearAll, selectEntry, selectedId, selectedEntry }
+}
diff --git a/core/http/react-ui/src/pages/Models.jsx b/core/http/react-ui/src/pages/Models.jsx
index 53dfccf2f..f7491805a 100644
--- a/core/http/react-ui/src/pages/Models.jsx
+++ b/core/http/react-ui/src/pages/Models.jsx
@@ -56,6 +56,7 @@ const FILTERS = [
   { key: 'chat', labelKey: 'filters.llm', icon: 'fa-brain' },
   { key: 'image', labelKey: 'filters.image', icon: 'fa-image' },
   { key: 'video', labelKey: 'filters.video', icon: 'fa-video' },
+  { key: '3d', labelKey: 'filters.threed', icon: 'fa-cube' },
   { key: 'multimodal', labelKey: 'filters.multimodal', icon: 'fa-shapes' },
   { key: 'vision', labelKey: 'filters.vision', icon: 'fa-eye' },
   { key: 'tts', labelKey: 'filters.tts', icon: 'fa-microphone' },
diff --git a/core/http/react-ui/src/pages/Studio.jsx b/core/http/react-ui/src/pages/Studio.jsx
index 5926078c2..43692cea7 100644
--- a/core/http/react-ui/src/pages/Studio.jsx
+++ b/core/http/react-ui/src/pages/Studio.jsx
@@ -2,6 +2,7 @@ import { useSearchParams } from 'react-router-dom'
 import { useTranslation } from 'react-i18next'
 import ImageGen from './ImageGen'
 import VideoGen from './VideoGen'
+import ThreeDGen from './ThreeDGen'
 import TTS from './TTS'
 import Sound from './Sound'
 import AudioTransform from './AudioTransform'
@@ -10,6 +11,7 @@ import { useAuth } from '../context/AuthContext'
 const BASE_TABS = [
   { key: 'images', labelKey: 'studio.tabs.images', icon: 'fas fa-image' },
   { key: 'video', labelKey: 'studio.tabs.video', icon: 'fas fa-video' },
+  { key: 'threed', labelKey: 'studio.tabs.threed', icon: 'fas fa-cube' },
   { key: 'tts', labelKey: 'studio.tabs.tts', icon: 'fas fa-headphones' },
   { key: 'sound', labelKey: 'studio.tabs.sound', icon: 'fas fa-music' },
 ]
@@ -19,6 +21,7 @@ const TRANSFORM_TAB = { key: 'transform', labelKey: 'studio.tabs.transform', ico
 const TAB_COMPONENTS = {
   images: ImageGen,
   video: VideoGen,
+  threed: ThreeDGen,
   tts: TTS,
   sound: Sound,
   transform: AudioTransform,
@@ -28,19 +31,23 @@ export default function Studio() {
   const { t } = useTranslation('media')
   const { hasFeature } = useAuth()
   const [searchParams, setSearchParams] = useSearchParams()
-  const activeTab = searchParams.get('tab') || 'images'
+  const requestedTab = searchParams.get('tab') || 'images'
+  const threeDEnabled = hasFeature('3d')
+  const transformEnabled = hasFeature('audio_transform')
+  const activeTab =
+    ((requestedTab === 'threed' && !threeDEnabled) ||
+      (requestedTab === 'transform' && !transformEnabled))
+      ? 'images'
+      : requestedTab
 
-  // Transform is a distinct capability; only show its tab when enabled.
-  const tabs = hasFeature('audio_transform') ? [...BASE_TABS, TRANSFORM_TAB] : BASE_TABS
+  const enabledTabs = BASE_TABS.filter(tab => tab.key !== 'threed' || threeDEnabled)
+  const tabs = transformEnabled ? [...enabledTabs, TRANSFORM_TAB] : enabledTabs
 
   const setTab = (key) => {
     setSearchParams({ tab: key }, { replace: true })
   }
 
-  const ActiveComponent =
-    (activeTab === 'transform' && !hasFeature('audio_transform'))
-      ? ImageGen
-      : (TAB_COMPONENTS[activeTab] || ImageGen)
+  const ActiveComponent = TAB_COMPONENTS[activeTab] || ImageGen
 
   return (
     <div>
diff --git a/core/http/react-ui/src/pages/ThreeDGen.jsx b/core/http/react-ui/src/pages/ThreeDGen.jsx
new file mode 100644
index 000000000..f4a0e3775
--- /dev/null
+++ b/core/http/react-ui/src/pages/ThreeDGen.jsx
@@ -0,0 +1,285 @@
+import { useState } from 'react'
+import { useParams, useOutletContext } from 'react-router-dom'
+import { useTranslation } from 'react-i18next'
+import ModelSelector from '../components/ModelSelector'
+import PageHeader from '../components/PageHeader'
+import { CAP_3D } from '../utils/capabilities'
+import LoadingSpinner from '../components/LoadingSpinner'
+import GenerationProgress from '../components/GenerationProgress'
+import ErrorWithTraceLink from '../components/ErrorWithTraceLink'
+import ThreeDHistory from '../components/ThreeDHistory'
+import GlbViewer from '../components/GlbViewer'
+import MediaInput from '../components/biometrics/MediaInput'
+import { threeDApi } from '../utils/api'
+import { apiUrl } from '../utils/basePath'
+import { use3DHistory } from '../hooks/use3DHistory'
+import useObjectUrl from '../hooks/useObjectUrl'
+
+const QUALITIES = ['auto', 'coarse', '512', '1024']
+const BACKGROUNDS = ['auto', 'keep', 'black', 'white']
+const MAX_3D_INPUT_BYTES = 32 * 1024 * 1024
+const REMESH_DETAIL_COARSE = 2.5
+const REMESH_DETAIL_FINE = 0.35
+
+function remeshDetail(sliderValue) {
+  const position = Number(sliderValue) / 100
+  return REMESH_DETAIL_COARSE * Math.pow(REMESH_DETAIL_FINE / REMESH_DETAIL_COARSE, position)
+}
+
+function remeshedName(name = '3d-model.glb') {
+  return `${name.replace(/\.glb$/i, '')}-remeshed.glb`
+}
+
+// Small thumbnail of the conditioning image for the history list — full-size
+// data URLs would bloat every IndexedDB entry for no visual gain.
+async function makeThumb(dataUrl, size = 96) {
+  try {
+    const img = new Image()
+    await new Promise((resolve, reject) => {
+      img.onload = resolve
+      img.onerror = reject
+      img.src = dataUrl
+    })
+    const scale = size / Math.max(img.width, img.height, 1)
+    const canvas = document.createElement('canvas')
+    canvas.width = Math.max(1, Math.round(img.width * scale))
+    canvas.height = Math.max(1, Math.round(img.height * scale))
+    canvas.getContext('2d').drawImage(img, 0, 0, canvas.width, canvas.height)
+    return canvas.toDataURL('image/jpeg', 0.7)
+  } catch {
+    return null
+  }
+}
+
+export default function ThreeDGen() {
+  const { model: urlModel } = useParams()
+  const { addToast } = useOutletContext()
+  const { t } = useTranslation('media')
+  const [model, setModel] = useState(urlModel || '')
+  const [image, setImage] = useState(null)
+  const [quality, setQuality] = useState('auto')
+  const [background, setBackground] = useState('auto')
+  const [steps, setSteps] = useState('')
+  const [textureSteps, setTextureSteps] = useState('')
+  const [guidance, setGuidance] = useState('')
+  const [seed, setSeed] = useState('')
+  const [showAdvanced, setShowAdvanced] = useState(false)
+  const [loading, setLoading] = useState(false)
+  const [error, setError] = useState(null)
+  const [result, setResult] = useState(null) // { blob, name, model }
+  const [remeshSlider, setRemeshSlider] = useState(82)
+  const [remeshState, setRemeshState] = useState(null) // { sourceBlob, blob?, name?, error? }
+  const [remeshLoading, setRemeshLoading] = useState(false)
+  const { entries, addEntry, deleteEntry, clearAll, selectEntry, selectedId, selectedEntry } = use3DHistory()
+
+  const source = selectedEntry
+    ? { blob: selectedEntry.glb, name: selectedEntry.name, model: selectedEntry.model }
+    : result
+  const showingRemesh = !!source && remeshState?.sourceBlob === source.blob && !!remeshState.blob
+  const active = showingRemesh ? { blob: remeshState.blob, name: remeshState.name } : source
+  const remeshError = source && remeshState?.sourceBlob === source.blob ? remeshState.error : null
+  const detail = remeshDetail(remeshSlider)
+  const downloadUrl = useObjectUrl(active?.blob)
+
+  const handleGenerate = async (e) => {
+    e.preventDefault()
+    if (!image?.base64) { addToast(t('threed.toasts.noImage'), 'warning'); return }
+    if (!model) { addToast(t('threed.toasts.noModel'), 'warning'); return }
+
+    setLoading(true)
+    setResult(null)
+    setRemeshState(null)
+    setError(null)
+
+    const body = { model, image: image.base64, quality, background, response_format: 'url' }
+    if (steps) body.step = parseInt(steps)
+    if (textureSteps) body.texture_steps = parseInt(textureSteps)
+    if (guidance) body.cfg_scale = parseFloat(guidance)
+    if (seed) body.seed = parseInt(seed)
+
+    try {
+      const data = await threeDApi.generate(body)
+      const url = data?.data?.[0]?.url
+      if (!url) {
+        addToast(t('threed.toasts.noResults'), 'warning')
+        return
+      }
+      const glbResp = await fetch(apiUrl(url))
+      if (!glbResp.ok) throw new Error(`fetching the generated GLB failed: HTTP ${glbResp.status}`)
+      const glb = await glbResp.blob()
+      const name = url.split('/').pop()
+      setResult({ blob: glb, name, model })
+      selectEntry(null)
+      const inputThumb = image.dataUrl ? await makeThumb(image.dataUrl) : null
+      await addEntry({
+        model,
+        params: { quality, background, steps, textureSteps, guidance, seed },
+        inputThumb,
+        glb,
+        name,
+      })
+    } catch (err) {
+      setError(err.message)
+    } finally {
+      setLoading(false)
+    }
+  }
+
+  const handleRemesh = async () => {
+    if (!source?.blob || !source.model) return
+    if (showingRemesh) {
+      setRemeshState(null)
+      return
+    }
+
+    const sourceBlob = source.blob
+    setRemeshLoading(true)
+    setRemeshState({ sourceBlob, error: null })
+    try {
+      const blob = await threeDApi.remesh(sourceBlob, source.model, detail)
+      setRemeshState({ sourceBlob, blob, name: remeshedName(source.name), error: null })
+    } catch (err) {
+      setRemeshState({ sourceBlob, error: err.message })
+    } finally {
+      setRemeshLoading(false)
+    }
+  }
+
+  const handleRemeshDetail = (value) => {
+    setRemeshSlider(value)
+    if (source && remeshState?.sourceBlob === source.blob) setRemeshState(null)
+  }
+
+  return (
+    <div className="media-layout">
+      <div className="media-controls">
+        <PageHeader title={<><i className="fas fa-cube" /> {t('threed.title')}</>} />
+
+        <form onSubmit={handleGenerate}>
+          <div className="form-group">
+            <label className="form-label">{t('threed.labels.model')}</label>
+            <ModelSelector value={model} onChange={setModel} capability={CAP_3D} />
+          </div>
+
+          <MediaInput
+            mode="image"
+            label={t('threed.labels.image')}
+            value={image}
+            onChange={setImage}
+            onError={(err) => addToast(err.message, 'error')}
+            maxBytes={MAX_3D_INPUT_BYTES}
+            idPrefix="threed"
+          />
+
+          <div className="form-grid-2col">
+            <div className="form-group">
+              <label className="form-label">{t('threed.labels.quality')}</label>
+              <select className="input btn-full" value={quality} onChange={(e) => setQuality(e.target.value)}>
+                {QUALITIES.map(q => <option key={q} value={q}>{t(`threed.labels.quality_${q}`)}</option>)}
+              </select>
+            </div>
+            <div className="form-group">
+              <label className="form-label">{t('threed.labels.background')}</label>
+              <select className="input btn-full" value={background} onChange={(e) => setBackground(e.target.value)}>
+                {BACKGROUNDS.map(b => <option key={b} value={b}>{t(`threed.labels.background_${b}`)}</option>)}
+              </select>
+            </div>
+          </div>
+
+          <button
+            type="button"
+            className={`collapsible-header ${showAdvanced ? 'open' : ''}`}
+            aria-expanded={showAdvanced}
+            aria-controls="threed-advanced-options"
+            onClick={() => setShowAdvanced(!showAdvanced)}
+          >
+            <i className="fas fa-chevron-right" aria-hidden="true" /> {t('threed.labels.advanced')}
+          </button>
+          {showAdvanced && (
+            <div id="threed-advanced-options" className="form-grid-2col">
+              <div className="form-group"><label className="form-label">{t('threed.labels.steps')}</label><input className="input" type="number" min="1" value={steps} onChange={(e) => setSteps(e.target.value)} placeholder="12" /></div>
+              <div className="form-group"><label className="form-label">{t('threed.labels.textureSteps')}</label><input className="input" type="number" min="1" value={textureSteps} onChange={(e) => setTextureSteps(e.target.value)} placeholder="12" /></div>
+              <div className="form-group"><label className="form-label">{t('threed.labels.guidance')}</label><input className="input" type="number" step="0.1" value={guidance} onChange={(e) => setGuidance(e.target.value)} placeholder="7.5" /></div>
+              <div className="form-group"><label className="form-label">{t('threed.labels.seed')}</label><input className="input" type="number" value={seed} onChange={(e) => setSeed(e.target.value)} placeholder={t('threed.labels.seedPlaceholder')} /></div>
+            </div>
+          )}
+
+          <button type="submit" className="btn btn-primary btn-full" disabled={loading}>
+            {loading ? <><LoadingSpinner size="sm" /> {t('threed.actions.generating')}</> : <><i className="fas fa-cube" /> {t('threed.actions.generate')}</>}
+          </button>
+        </form>
+        <ThreeDHistory
+          entries={entries}
+          selectedId={selectedId}
+          onSelect={selectEntry}
+          onDelete={deleteEntry}
+          onClearAll={clearAll}
+        />
+      </div>
+
+      <div className="media-preview">
+        <div className="media-result">
+          {loading ? (
+            <GenerationProgress label={t('threed.actions.generating')} />
+          ) : error ? (
+            <ErrorWithTraceLink message={error} />
+          ) : active?.blob ? (
+            <div style={{ display: 'flex', flexDirection: 'column', gap: 'var(--spacing-md)', width: '100%' }}>
+              <GlbViewer blob={active.blob} />
+              <div className="threed-remesh-controls">
+                <div className="threed-remesh-heading">
+                  <span>{t('threed.remesh.title')}</span>
+                  <output htmlFor="threed-remesh-detail">{detail.toFixed(2)}%</output>
+                </div>
+                <input
+                  id="threed-remesh-detail"
+                  type="range"
+                  min="0"
+                  max="100"
+                  step="1"
+                  value={remeshSlider}
+                  onChange={(e) => handleRemeshDetail(e.target.value)}
+                  disabled={remeshLoading}
+                  aria-label={t('threed.remesh.detail')}
+                />
+                <div className="threed-remesh-scale" aria-hidden="true">
+                  <span>{t('threed.remesh.coarser')}</span>
+                  <span>{t('threed.remesh.finer')}</span>
+                </div>
+                <p className="form-hint">{t('threed.remesh.hint')}</p>
+                <button
+                  type="button"
+                  className="btn btn-secondary btn-full"
+                  onClick={handleRemesh}
+                  disabled={remeshLoading}
+                  data-testid="glb-remesh"
+                >
+                  {remeshLoading
+                    ? <><LoadingSpinner size="sm" /> {t('threed.actions.remeshing')}</>
+                    : showingRemesh
+                      ? <><i className="fas fa-rotate-left" /> {t('threed.actions.showOriginal')}</>
+                      : <><i className="fas fa-cubes-stacked" /> {t('threed.actions.remesh')}</>}
+                </button>
+                {remeshError && <p className="form-error" role="alert">{remeshError}</p>}
+                {showingRemesh && <p className="threed-remesh-ready">{t('threed.remesh.ready')}</p>}
+              </div>
+              <a
+                className="btn btn-secondary"
+                href={downloadUrl}
+                download={active.name || `3d-${model || 'model'}.glb`}
+                data-testid="glb-download"
+              >
+                <i className="fas fa-download" /> {t('threed.actions.download')}
+              </a>
+            </div>
+          ) : (
+            <div style={{ textAlign: 'center', color: 'var(--color-text-muted)' }}>
+              <i className="fas fa-cube" style={{ fontSize: '3rem', marginBottom: 'var(--spacing-md)', opacity: 0.4 }} />
+              <p>{t('threed.empty')}</p>
+            </div>
+          )}
+        </div>
+      </div>
+    </div>
+  )
+}
diff --git a/core/http/react-ui/src/pages/Traces.jsx b/core/http/react-ui/src/pages/Traces.jsx
index aff887d4a..fc9d50bea 100644
--- a/core/http/react-ui/src/pages/Traces.jsx
+++ b/core/http/react-ui/src/pages/Traces.jsx
@@ -74,6 +74,8 @@ const TYPE_COLORS = {
   transcription: { bg: 'var(--color-warning-light)', color: 'var(--color-data-4)' },
   image_generation: { bg: 'var(--color-success-light)', color: 'var(--color-data-5)' },
   video_generation: { bg: 'var(--color-accent-light)', color: 'var(--color-data-7)' },
+  '3d_generation': { bg: 'var(--color-success-light)', color: 'var(--color-data-5)' },
+  '3d_remesh': { bg: 'var(--color-accent-light)', color: 'var(--color-data-7)' },
   tts: { bg: 'var(--color-warning-light)', color: 'var(--color-data-6)' },
   sound_generation: { bg: 'var(--color-info-light)', color: 'var(--color-data-8)' },
   rerank: { bg: 'var(--color-primary-light)', color: 'var(--color-data-1)' },
diff --git a/core/http/react-ui/src/router.jsx b/core/http/react-ui/src/router.jsx
index 47be7e72f..a221e0180 100644
--- a/core/http/react-ui/src/router.jsx
+++ b/core/http/react-ui/src/router.jsx
@@ -38,6 +38,7 @@ const Models = page('models', () => import('./pages/Models'))
 const Manage = page('manage', () => import('./pages/Manage'))
 const ImageGen = page('image', () => import('./pages/ImageGen'))
 const VideoGen = page('video', () => import('./pages/VideoGen'))
+const ThreeDGen = page('3d', () => import('./pages/ThreeDGen'))
 const TTS = page('tts', () => import('./pages/TTS'))
 const Sound = page('sound', () => import('./pages/Sound'))
 const AudioTransform = page('transform', () => import('./pages/AudioTransform'))
@@ -110,6 +111,8 @@ const appChildren = [
   { path: 'image/:model', element: <ImageGen /> },
   { path: 'video', element: <VideoGen /> },
   { path: 'video/:model', element: <VideoGen /> },
+  { path: '3d', element: <Feature feature="3d"><ThreeDGen /></Feature> },
+  { path: '3d/:model', element: <Feature feature="3d"><ThreeDGen /></Feature> },
   { path: 'tts', element: <TTS /> },
   { path: 'tts/:model', element: <TTS /> },
   { path: 'sound', element: <Sound /> },
diff --git a/core/http/react-ui/src/utils/api.js b/core/http/react-ui/src/utils/api.js
index bd0640963..0c9eb94c3 100644
--- a/core/http/react-ui/src/utils/api.js
+++ b/core/http/react-ui/src/utils/api.js
@@ -274,6 +274,22 @@ export const videoApi = {
   generate: (body) => postJSON(API_CONFIG.endpoints.video, body),
 }
 
+export const threeDApi = {
+  generate: (body) => postJSON(API_CONFIG.endpoints.threeDGenerations, body),
+  remesh: async (mesh, model, detail) => {
+    const form = new FormData()
+    form.append('model', model)
+    form.append('detail', String(detail))
+    form.append('mesh', mesh, 'source.glb')
+    const response = await fetch(apiUrl(API_CONFIG.endpoints.threeDRemesh), {
+      method: 'POST',
+      body: form,
+    })
+    await handleResponse(response)
+    return response.blob()
+  },
+}
+
 // parseAudioBlobResponse — shared response handling for audio-blob endpoints.
 // Throws on non-2xx (with the API error message when present); returns the
 // blob plus the parsed Content-Disposition filename mapped to the server's
diff --git a/core/http/react-ui/src/utils/capabilities.js b/core/http/react-ui/src/utils/capabilities.js
index 5d30a472d..b790d95d2 100644
--- a/core/http/react-ui/src/utils/capabilities.js
+++ b/core/http/react-ui/src/utils/capabilities.js
@@ -17,6 +17,11 @@ export const CAP_VAD = 'FLAG_VAD'
 export const CAP_DIARIZATION = 'FLAG_DIARIZATION'
 export const CAP_SOUND_CLASSIFICATION = 'FLAG_SOUND_CLASSIFICATION'
 export const CAP_VIDEO = 'FLAG_VIDEO'
+// Wire format note: /api/models/capabilities serves KnownUsecaseStrings, which
+// syncKnownUsecasesFromString rewrites to the UPPERCASE keys of
+// GetAllModelConfigUsecases() — so this must be FLAG_3D, not the lowercase
+// "3d" served by the OpenAI-style /v1/models/capabilities endpoint.
+export const CAP_3D = 'FLAG_3D'
 export const CAP_DETECTION = 'FLAG_DETECTION'
 export const CAP_FACE_RECOGNITION = 'FLAG_FACE_RECOGNITION'
 export const CAP_SPEAKER_RECOGNITION = 'FLAG_SPEAKER_RECOGNITION'
diff --git a/core/http/react-ui/src/utils/config.js b/core/http/react-ui/src/utils/config.js
index 66f136619..6ba619a1a 100644
--- a/core/http/react-ui/src/utils/config.js
+++ b/core/http/react-ui/src/utils/config.js
@@ -108,6 +108,8 @@ export const API_CONFIG = {
     // LocalAI-specific
     tts: '/tts',
     video: '/video',
+    threeDGenerations: '/3d/generations',
+    threeDRemesh: '/3d/remesh',
     backendMonitor: '/backend/monitor',
     backendShutdown: '/backend/shutdown',
     backendLoad: '/backend/load',
diff --git a/core/http/react-ui/src/utils/glb.js b/core/http/react-ui/src/utils/glb.js
new file mode 100644
index 000000000..310484371
--- /dev/null
+++ b/core/http/react-ui/src/utils/glb.js
@@ -0,0 +1,144 @@
+// Minimal GLB (binary glTF 2.0) parser for the two forms trellis2cpp's
+// t2_bake_glb emits (see mesh_export.cpp in localai-org/trellis2cpp):
+//
+//  Form B (default, dense vertex PBR):
+//    POSITION (VEC3 f32), NORMAL (VEC3 f32),
+//    COLOR_0 (VEC4 u16 normalized, LINEAR color + alpha),
+//    _METALLIC_ROUGHNESS (VEC2 u8 normalized: metallic, roughness),
+//    indices (SCALAR u32); material carries average metallic/roughness factors.
+//
+//  Form A (opt-in via T2GLB_XATLAS, UV atlas):
+//    POSITION/NORMAL/TEXCOORD_0 f32 + u32 indices,
+//    baseColorTexture (sRGB PNG) + metallicRoughnessTexture
+//    (glTF convention: G=roughness, B=metallic).
+//
+// Deliberately unsupported (the baker never emits them; anything else throws
+// so the page can show the error while the download button still works):
+// sparse accessors, interleaved bufferViews (byteStride), Draco, multiple
+// meshes/primitives.
+
+const GLB_MAGIC = 0x46546c67
+const CHUNK_JSON = 0x4e4f534a
+const CHUNK_BIN = 0x004e4942
+
+const COMPONENT_ARRAYS = {
+  5121: Uint8Array,
+  5123: Uint16Array,
+  5125: Uint32Array,
+  5126: Float32Array,
+}
+
+const TYPE_SIZES = { SCALAR: 1, VEC2: 2, VEC3: 3, VEC4: 4 }
+
+export function parseGlb(buf) {
+  if (!(buf instanceof ArrayBuffer) || buf.byteLength < 20) throw new Error('not a GLB file')
+  const dv = new DataView(buf)
+  if (dv.getUint32(0, true) !== GLB_MAGIC) throw new Error('not a GLB file')
+
+  let offset = 12
+  let json = null
+  let binOffset = -1
+  let binLength = 0
+  while (offset + 8 <= buf.byteLength) {
+    const len = dv.getUint32(offset, true)
+    const type = dv.getUint32(offset + 4, true)
+    const payload = offset + 8
+    if (type === CHUNK_JSON && !json) {
+      json = JSON.parse(new TextDecoder().decode(new Uint8Array(buf, payload, len)))
+    } else if (type === CHUNK_BIN && binOffset < 0) {
+      binOffset = payload
+      binLength = len
+    }
+    offset = payload + len + ((4 - (len % 4)) % 4)
+  }
+  if (!json) throw new Error('GLB has no JSON chunk')
+
+  const accessor = (index) => {
+    const a = json.accessors?.[index]
+    if (!a) throw new Error(`missing accessor ${index}`)
+    if (a.sparse) throw new Error('sparse accessors not supported')
+    const view = json.bufferViews?.[a.bufferView]
+    if (!view) throw new Error(`missing bufferView ${a.bufferView}`)
+    if (view.byteStride) throw new Error('interleaved bufferViews not supported')
+    const ArrayType = COMPONENT_ARRAYS[a.componentType]
+    const size = TYPE_SIZES[a.type]
+    if (!ArrayType || !size) throw new Error(`unsupported accessor layout ${a.componentType}/${a.type}`)
+    const start = binOffset + (view.byteOffset || 0) + (a.byteOffset || 0)
+    if (binOffset < 0 || start + a.count * size * ArrayType.BYTES_PER_ELEMENT > binOffset + binLength) {
+      throw new Error('accessor outside the BIN chunk')
+    }
+    return {
+      array: new ArrayType(buf, start, a.count * size),
+      size,
+      normalized: !!a.normalized,
+      min: a.min,
+      max: a.max,
+      count: a.count,
+    }
+  }
+
+  const prim = json.meshes?.[0]?.primitives?.[0]
+  if (!prim) throw new Error('GLB has no mesh')
+  const attrs = prim.attributes || {}
+  if (attrs.POSITION === undefined) throw new Error('GLB mesh has no POSITION attribute')
+
+  const position = accessor(attrs.POSITION)
+  const normal = attrs.NORMAL !== undefined ? accessor(attrs.NORMAL) : null
+
+  let indices
+  if (prim.indices !== undefined) {
+    const idx = accessor(prim.indices)
+    indices = idx.array instanceof Uint32Array ? idx.array : Uint32Array.from(idx.array)
+  } else {
+    indices = new Uint32Array(position.count)
+    for (let i = 0; i < indices.length; i++) indices[i] = i
+  }
+
+  const color0 = attrs.COLOR_0 !== undefined ? accessor(attrs.COLOR_0) : null
+  const metalRough = attrs._METALLIC_ROUGHNESS !== undefined ? accessor(attrs._METALLIC_ROUGHNESS) : null
+  const uv = attrs.TEXCOORD_0 !== undefined ? accessor(attrs.TEXCOORD_0) : null
+
+  const material = json.materials?.[prim.material]?.pbrMetallicRoughness || {}
+  const imageBytes = (textureIndex) => {
+    if (textureIndex === undefined) return null
+    const source = json.textures?.[textureIndex]?.source
+    const view = json.bufferViews?.[json.images?.[source]?.bufferView]
+    if (!view) return null
+    return new Uint8Array(buf, binOffset + (view.byteOffset || 0), view.byteLength)
+  }
+
+  // POSITION min/max are mandatory in glTF, but compute a fallback so a
+  // technically-invalid file still frames correctly.
+  let bboxMin = position.min
+  let bboxMax = position.max
+  if (!bboxMin || !bboxMax) {
+    bboxMin = [Infinity, Infinity, Infinity]
+    bboxMax = [-Infinity, -Infinity, -Infinity]
+    for (let i = 0; i < position.array.length; i += 3) {
+      for (let k = 0; k < 3; k++) {
+        const v = position.array[i + k]
+        if (v < bboxMin[k]) bboxMin[k] = v
+        if (v > bboxMax[k]) bboxMax[k] = v
+      }
+    }
+  }
+
+  return {
+    positions: position.array,
+    normals: normal ? normal.array : null,
+    indices,
+    color0,
+    metalRough,
+    uv: uv ? uv.array : null,
+    baseColorPng: imageBytes(material.baseColorTexture?.index),
+    metalRoughPng: imageBytes(material.metallicRoughnessTexture?.index),
+    material: {
+      metallicFactor: material.metallicFactor ?? 1,
+      roughnessFactor: material.roughnessFactor ?? 1,
+    },
+    bboxMin,
+    bboxMax,
+    nVerts: position.count,
+    nTris: Math.floor(indices.length / 3),
+  }
+}
diff --git a/core/http/react-ui/vite.config.js b/core/http/react-ui/vite.config.js
index 0e1296f74..094800066 100644
--- a/core/http/react-ui/vite.config.js
+++ b/core/http/react-ui/vite.config.js
@@ -56,6 +56,8 @@ export default defineConfig({
       '/generated-audio': backendUrl,
       '/generated-images': backendUrl,
       '/generated-videos': backendUrl,
+      '/generated-3d': backendUrl,
+      '/3d': backendUrl,
       '/version': backendUrl,
       '/system': backendUrl,
     },
diff --git a/core/http/routes/localai.go b/core/http/routes/localai.go
index 4a83e9009..498e29e8f 100644
--- a/core/http/routes/localai.go
+++ b/core/http/routes/localai.go
@@ -2,6 +2,7 @@ package routes
 
 import (
 	"github.com/labstack/echo/v4"
+	echomiddleware "github.com/labstack/echo/v4/middleware"
 	"github.com/mudler/LocalAI/core/application"
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/http/endpoints/localai"
@@ -193,10 +194,10 @@ func RegisterLocalAIRoutes(router *echo.Echo,
 		requestExtractor.SetModelAndConfig(func() schema.LocalAIRequest { return new(schema.VADRequest) }))
 
 	// Stores
-	router.POST("/stores/set", localai.StoresSetEndpoint(ml, appConfig))
-	router.POST("/stores/delete", localai.StoresDeleteEndpoint(ml, appConfig))
-	router.POST("/stores/get", localai.StoresGetEndpoint(ml, appConfig))
-	router.POST("/stores/find", localai.StoresFindEndpoint(ml, appConfig))
+	router.POST("/stores/set", localai.StoresSetEndpoint(ml, cl, appConfig))
+	router.POST("/stores/delete", localai.StoresDeleteEndpoint(ml, cl, appConfig))
+	router.POST("/stores/get", localai.StoresGetEndpoint(ml, cl, appConfig))
+	router.POST("/stores/find", localai.StoresFindEndpoint(ml, cl, appConfig))
 
 	if !appConfig.DisableMetrics {
 		router.GET("/metrics", localai.LocalAIMetricsEndpoint(), adminMiddleware)
@@ -208,6 +209,17 @@ func RegisterLocalAIRoutes(router *echo.Echo,
 		requestExtractor.BuildFilteredFirstAvailableDefaultModel(config.BuildUsecaseFilterFn(config.FLAG_VIDEO)),
 		requestExtractor.SetModelAndConfig(func() schema.LocalAIRequest { return new(schema.VideoRequest) }))
 
+	model3dHandler := localai.Model3DEndpoint(cl, ml, appConfig)
+	router.POST("/3d/generations",
+		model3dHandler,
+		requestExtractor.BuildFilteredFirstAvailableDefaultModel(config.BuildUsecaseFilterFn(config.FLAG_3D)),
+		requestExtractor.SetModelAndConfig(func() schema.LocalAIRequest { return new(schema.Model3DRequest) }))
+	router.POST("/3d/remesh",
+		localai.Model3DRemeshEndpoint(ml, appConfig),
+		echomiddleware.BodyLimit("513M"),
+		requestExtractor.BuildFilteredFirstAvailableDefaultModel(config.BuildUsecaseFilterFn(config.FLAG_3D)),
+		requestExtractor.SetModelAndConfig(func() schema.LocalAIRequest { return new(schema.Model3DRemeshRequest) }))
+
 	// Backend Statistics Module
 	// TODO: Should these use standard middlewares? Refactor later, they are extremely simple.
 	backendMonitorService := monitoring.NewBackendMonitorService(ml, cl, appConfig) // Split out for now
@@ -333,6 +345,7 @@ func RegisterLocalAIRoutes(router *echo.Echo,
 					"voice_profiles": "/api/voice-profiles",
 					"vad":            "/vad",
 					"video":          "/video",
+					"3d_generation":  "/3d/generations",
 					"detection":      "/v1/detection",
 					"tokenize":       "/v1/tokenize",
 				},
diff --git a/core/http/routes/ui_api.go b/core/http/routes/ui_api.go
index e2f2cc236..300cc0a3b 100644
--- a/core/http/routes/ui_api.go
+++ b/core/http/routes/ui_api.go
@@ -46,6 +46,7 @@ var usecaseFilters = map[string]config.ModelConfigUsecase{
 	config.UsecaseChat:                config.FLAG_CHAT,
 	config.UsecaseImage:               config.FLAG_IMAGE,
 	config.UsecaseVideo:               config.FLAG_VIDEO,
+	config.Usecase3D:                  config.FLAG_3D,
 	config.UsecaseVision:              config.FLAG_VISION,
 	config.UsecaseTTS:                 config.FLAG_TTS,
 	config.UsecaseTranscript:          config.FLAG_TRANSCRIPT,
diff --git a/core/http/routes/ui_api_internal_test.go b/core/http/routes/ui_api_internal_test.go
new file mode 100644
index 000000000..f63a9a067
--- /dev/null
+++ b/core/http/routes/ui_api_internal_test.go
@@ -0,0 +1,13 @@
+package routes
+
+import (
+	"testing"
+
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/onsi/gomega"
+)
+
+func TestUsecaseFiltersIncludes3D(t *testing.T) {
+	g := gomega.NewWithT(t)
+	g.Expect(usecaseFilters[config.Usecase3D]).To(gomega.Equal(config.FLAG_3D))
+}
diff --git a/core/schema/localai.go b/core/schema/localai.go
index 6513a7013..cead74f02 100644
--- a/core/schema/localai.go
+++ b/core/schema/localai.go
@@ -70,6 +70,29 @@ type VideoRequest struct {
 	Params         map[string]string `json:"params,omitempty" yaml:"params,omitempty"`         // backend-specific generation parameters
 }
 
+// @Description 3D asset generation request body. Generation is image-conditioned
+// (TRELLIS.2 has no text-prompt path); the response is a binary glTF (.glb).
+type Model3DRequest struct {
+	BasicModelRequest
+	Image          string            `json:"image" yaml:"image"`                                           // conditioning image: URL, base64, or data URI (required)
+	Seed           int32             `json:"seed,omitempty" yaml:"seed,omitempty"`                         // random seed; <=0 picks a random seed
+	Step           int32             `json:"step,omitempty" yaml:"step,omitempty"`                         // flow sampling steps (backend default 12)
+	CFGScale       float32           `json:"cfg_scale,omitempty" yaml:"cfg_scale,omitempty"`               // classifier-free guidance scale (backend default 7.5)
+	TextureSteps   int32             `json:"texture_steps,omitempty" yaml:"texture_steps,omitempty"`       // texture flow sampling steps (backend default 12)
+	Quality        string            `json:"quality,omitempty" yaml:"quality,omitempty"`                   // mesh pipeline: auto|coarse|512|1024
+	Background     string            `json:"background,omitempty" yaml:"background,omitempty"`             // background handling: auto|keep|black|white
+	ResponseFormat string            `json:"response_format,omitempty" yaml:"response_format,omitempty"`   // output format (url or b64_json)
+	Params         map[string]string `json:"params,omitempty" yaml:"params,omitempty"`                     // backend-specific generation parameters
+}
+
+// @Description Print-remesh an existing trellis2.cpp GLB. The multipart mesh
+// is wrapped into a watertight manifold; detail is a percentage of the source
+// bounding-box diagonal and the enclosing offset is derived automatically.
+type Model3DRemeshRequest struct {
+	BasicModelRequest
+	Detail float32 `json:"detail,omitempty" yaml:"detail,omitempty" form:"detail"` // detail size in percent (0.35–2.5; default 0.5)
+}
+
 // @Description TTS request body
 type TTSRequest struct {
 	BasicModelRequest
diff --git a/core/schema/request.go b/core/schema/request.go
index 1e4eed156..75b8b9db8 100644
--- a/core/schema/request.go
+++ b/core/schema/request.go
@@ -7,7 +7,7 @@ type LocalAIRequest interface {
 
 // @Description BasicModelRequest contains the basic model request fields
 type BasicModelRequest struct {
-	Model string `json:"model,omitempty" yaml:"model,omitempty"`
+	Model string `json:"model,omitempty" yaml:"model,omitempty" form:"model"`
 	// TODO: Should this also include the following fields from the OpenAI side of the world?
 	// If so, changes should be made to core/http/middleware/request.go to match
 
diff --git a/core/services/nodes/file_staging_3d_test.go b/core/services/nodes/file_staging_3d_test.go
new file mode 100644
index 000000000..f67de6682
--- /dev/null
+++ b/core/services/nodes/file_staging_3d_test.go
@@ -0,0 +1,46 @@
+package nodes
+
+import (
+	"context"
+	"errors"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+
+	grpc "github.com/mudler/LocalAI/pkg/grpc"
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+	ggrpc "google.golang.org/grpc"
+)
+
+type capturing3DBackend struct {
+	grpc.Backend
+	request *pb.Generate3DRequest
+}
+
+func (b *capturing3DBackend) Generate3D(_ context.Context, request *pb.Generate3DRequest, _ ...ggrpc.CallOption) (*pb.Result, error) {
+	b.request = request
+	return &pb.Result{Success: true}, nil
+}
+
+type failingFetchStager struct {
+	fakeFileStager
+}
+
+func (f *failingFetchStager) FetchRemote(_ context.Context, _, _, _ string) error {
+	return errors.New("transfer failed")
+}
+
+var _ = Describe("FileStagingClient 3D output", func() {
+	It("returns an error when the generated asset cannot be retrieved", func(ctx SpecContext) {
+		backend := &capturing3DBackend{}
+		stager := &failingFetchStager{}
+		client := NewFileStagingClient(backend, stager, "worker-1")
+		request := &pb.Generate3DRequest{Dst: "/data/generated/asset.glb"}
+
+		result, err := client.Generate3D(ctx, request)
+
+		Expect(result).To(Equal(&pb.Result{Success: true}))
+		Expect(err).To(MatchError(ContainSubstring("retrieving generated 3D asset: transfer failed")))
+		Expect(backend.request.Dst).To(Equal("/remote/tmp"))
+	})
+})
diff --git a/core/services/nodes/file_staging_client.go b/core/services/nodes/file_staging_client.go
index 7ecf6a02e..bfc202c82 100644
--- a/core/services/nodes/file_staging_client.go
+++ b/core/services/nodes/file_staging_client.go
@@ -209,6 +209,42 @@ func (f *FileStagingClient) GenerateVideo(ctx context.Context, in *pb.GenerateVi
 	return result, nil
 }
 
+func (f *FileStagingClient) Generate3D(ctx context.Context, in *pb.Generate3DRequest, opts ...ggrpc.CallOption) (*pb.Result, error) {
+	reqID := requestID()
+
+	// Stage the conditioning image or existing GLB used by 3D post-processing.
+	if in.Src != "" && isFilePath(in.Src) {
+		backendPath, _, err := f.stageInputFile(ctx, reqID, in.Src, "inputs")
+		if err != nil {
+			return nil, fmt.Errorf("staging 3D input asset: %w", err)
+		}
+		in.Src = backendPath
+	}
+
+	// Handle output destination
+	frontendDst := in.Dst
+	if frontendDst != "" {
+		tmpPath, err := f.stager.AllocRemoteTemp(ctx, f.nodeID)
+		if err != nil {
+			return nil, fmt.Errorf("allocating temp for 3D output: %w", err)
+		}
+		in.Dst = tmpPath
+	}
+
+	result, err := f.Backend.Generate3D(ctx, in, opts...)
+	if err != nil {
+		return result, err
+	}
+
+	if frontendDst != "" && in.Dst != frontendDst {
+		if err := f.retrieveOutputFile(ctx, in.Dst, frontendDst); err != nil {
+			return result, fmt.Errorf("retrieving generated 3D asset: %w", err)
+		}
+	}
+
+	return result, nil
+}
+
 func (f *FileStagingClient) TTS(ctx context.Context, in *pb.TTSRequest, opts ...ggrpc.CallOption) (*pb.Result, error) {
 	reqID := requestID()
 
diff --git a/core/services/nodes/health_mock_test.go b/core/services/nodes/health_mock_test.go
index 429087f52..5e2571194 100644
--- a/core/services/nodes/health_mock_test.go
+++ b/core/services/nodes/health_mock_test.go
@@ -157,6 +157,9 @@ func (c *fakeBackendClient) GenerateImage(_ context.Context, _ *pb.GenerateImage
 func (c *fakeBackendClient) GenerateVideo(_ context.Context, _ *pb.GenerateVideoRequest, _ ...ggrpc.CallOption) (*pb.Result, error) {
 	return nil, nil
 }
+func (c *fakeBackendClient) Generate3D(_ context.Context, _ *pb.Generate3DRequest, _ ...ggrpc.CallOption) (*pb.Result, error) {
+	return nil, nil
+}
 func (c *fakeBackendClient) TTS(_ context.Context, _ *pb.TTSRequest, _ ...ggrpc.CallOption) (*pb.Result, error) {
 	return nil, nil
 }
diff --git a/core/services/nodes/inflight.go b/core/services/nodes/inflight.go
index f4bb8988f..de851121c 100644
--- a/core/services/nodes/inflight.go
+++ b/core/services/nodes/inflight.go
@@ -144,6 +144,12 @@ func (c *InFlightTrackingClient) GenerateVideo(ctx context.Context, in *pb.Gener
 	return res, c.reconcile(err)
 }
 
+func (c *InFlightTrackingClient) Generate3D(ctx context.Context, in *pb.Generate3DRequest, opts ...ggrpc.CallOption) (*pb.Result, error) {
+	defer c.track(ctx)()
+	res, err := c.inner.Generate3D(ctx, in, opts...)
+	return res, c.reconcile(err)
+}
+
 func (c *InFlightTrackingClient) TTS(ctx context.Context, in *pb.TTSRequest, opts ...ggrpc.CallOption) (*pb.Result, error) {
 	defer c.track(ctx)()
 	res, err := c.inner.TTS(ctx, in, opts...)
diff --git a/core/services/nodes/inflight_test.go b/core/services/nodes/inflight_test.go
index 120d2e312..a79237a02 100644
--- a/core/services/nodes/inflight_test.go
+++ b/core/services/nodes/inflight_test.go
@@ -87,6 +87,10 @@ func (f *fakeGRPCBackend) GenerateVideo(_ context.Context, _ *pb.GenerateVideoRe
 	return &pb.Result{}, nil
 }
 
+func (f *fakeGRPCBackend) Generate3D(_ context.Context, _ *pb.Generate3DRequest, _ ...ggrpc.CallOption) (*pb.Result, error) {
+	return &pb.Result{}, nil
+}
+
 func (f *fakeGRPCBackend) TTS(_ context.Context, _ *pb.TTSRequest, _ ...ggrpc.CallOption) (*pb.Result, error) {
 	return &pb.Result{}, nil
 }
diff --git a/core/trace/backend_trace.go b/core/trace/backend_trace.go
index 051b471bf..951351f8c 100644
--- a/core/trace/backend_trace.go
+++ b/core/trace/backend_trace.go
@@ -24,6 +24,8 @@ const (
 	BackendTraceTranscription   BackendTraceType = "transcription"
 	BackendTraceImageGeneration BackendTraceType = "image_generation"
 	BackendTraceVideoGeneration BackendTraceType = "video_generation"
+	BackendTrace3DGeneration    BackendTraceType = "3d_generation"
+	BackendTrace3DRemesh        BackendTraceType = "3d_remesh"
 	BackendTraceTTS             BackendTraceType = "tts"
 	BackendTraceSoundGeneration BackendTraceType = "sound_generation"
 	BackendTraceRerank          BackendTraceType = "rerank"
diff --git a/docs/content/advanced/reverse-proxy-tls.md b/docs/content/advanced/reverse-proxy-tls.md
index d36a64ae4..8ac94050a 100644
--- a/docs/content/advanced/reverse-proxy-tls.md
+++ b/docs/content/advanced/reverse-proxy-tls.md
@@ -108,25 +108,48 @@ server {
     ssl_certificate /path/to/cert.pem;
     ssl_certificate_key /path/to/key.pem;
     
-    # Set the X-Forwarded-Proto header
-    proxy_set_header X-Forwarded-Proto $scheme;
-    
-    # Pass the original host
-    proxy_set_header X-Forwarded-Host $host;
-    
-    # If serving under a sub-path
-    # proxy_set_header X-Forwarded-Prefix /localai;
-    
-    # Other proxy settings
-    proxy_pass http://127.0.0.1:8080;
-    proxy_http_version 1.1;
-    proxy_set_header Upgrade $http_upgrade;
-    proxy_set_header Connection "upgrade";
-    proxy_set_header Host $host;
-    proxy_cache_bypass $http_upgrade;
+    # Allow multimodal requests with images or other large inputs.
+    client_max_body_size 50m;
+
+    location / {
+        proxy_pass http://127.0.0.1:8080;
+        proxy_http_version 1.1;
+
+        # Set the externally visible scheme and host.
+        proxy_set_header X-Forwarded-Proto $scheme;
+        proxy_set_header X-Forwarded-Host $host;
+        proxy_set_header Host $host;
+
+        # If serving under a sub-path
+        # proxy_set_header X-Forwarded-Prefix /localai;
+
+        # Local inference can take much longer than common proxy defaults.
+        proxy_connect_timeout 60s;
+        proxy_send_timeout 60m;
+        proxy_read_timeout 60m;
+
+        # Stream responses as LocalAI emits them. Disabling request buffering is
+        # optional and can help when clients upload large multimodal inputs.
+        proxy_buffering off;
+        proxy_request_buffering off;
+
+        proxy_set_header Upgrade $http_upgrade;
+        proxy_set_header Connection "upgrade";
+        proxy_cache_bypass $http_upgrade;
+    }
 }
 ```
 
+Adjust the `60m` values for the slowest request you expect to serve. The
+important setting for long non-streaming requests is `proxy_read_timeout`: if
+the proxy stops waiting before LocalAI finishes, clients receive a proxy-generated
+`504 Gateway Time-out` even though inference may still be running. OpenResty,
+Nginx Proxy Manager, Caddy, Traefik, and HAProxy have equivalent upstream
+response timeout settings.
+
+For bulk jobs on a trusted private network, you can also bypass the public
+reverse proxy and connect directly to LocalAI.
+
 ## Serving Under a Sub-Path
 
 If you serve LocalAI under a sub-path (e.g., `https://your-domain.com/localai`), you need to:
@@ -165,6 +188,19 @@ proxy_set_header X-Forwarded-Prefix /localai;
 - Check that your proxy is not adding duplicate headers
 - Verify `X-Forwarded-Proto` is not being set to both `http` and `https`
 
+### 504 Gateway Time-out During Inference
+
+An HTML `504 Gateway Time-out` page branded by Nginx or OpenResty means the
+reverse proxy stopped waiting for LocalAI. Increase the proxy's upstream read
+and send timeouts beyond the worst-case inference time. Also check its request
+body limit when sending images, audio, or other large inputs.
+
+This is separate from LocalAI's optional busy watchdog. If busy watchdog checks
+are enabled, set `LOCALAI_WATCHDOG_BUSY_TIMEOUT` beyond the longest expected
+request or disable the busy check for workloads that legitimately run longer.
+The default busy timeout is 5 minutes, but it is only enforced when the busy
+watchdog is enabled.
+
 ## Security Note
 
 When using reverse proxies, ensure your proxy only accepts connections from trusted sources and properly validates SSL certificates. Never expose LocalAI directly to the internet without TLS termination.
diff --git a/docs/content/features/3d-generation.md b/docs/content/features/3d-generation.md
new file mode 100644
index 000000000..67663dd5c
--- /dev/null
+++ b/docs/content/features/3d-generation.md
@@ -0,0 +1,122 @@
++++
+disableToc = false
+title = "3D Generation"
+weight = 19
+url = "/features/3d-generation/"
++++
+
+LocalAI can generate textured 3D meshes from a single conditioning image via the `/3d/generations` endpoint, powered by the `trellis2cpp` backend — a C++/GGML port of [Microsoft TRELLIS.2](https://github.com/microsoft/TRELLIS.2) ([trellis2.cpp](https://github.com/localai-org/trellis2cpp)). The output is a binary glTF (`.glb`) asset with PBR materials.
+
+Generation is image-conditioned only — there is no text-prompt path. Provide a photo or rendering of a single object (ideally on a plain background) and TRELLIS.2 reconstructs a full 3D mesh from it.
+
+## Setup
+
+Install a model from the gallery:
+
+```bash
+local-ai run trellis2-4b          # full pipeline: 1024³ cascade + PBR textures (~18 GB)
+# or
+local-ai run trellis2-4b-geometry # 512³ untextured geometry only (~7 GB)
+```
+
+The backend detects which component GGUFs are present and degrades gracefully: without the texture models it produces untextured geometry, and without the fine-flow models it falls back to a coarse marching-cubes preview.
+
+## API
+
+- **Method:** `POST`
+- **Endpoint:** `/3d/generations`
+
+### Request
+
+The request body is JSON with the following fields:
+
+| Parameter         | Type     | Required | Default | Description                                                        |
+|-------------------|----------|----------|---------|--------------------------------------------------------------------|
+| `model`           | `string` | Yes      |         | Model name to use                                                  |
+| `image`           | `string` | Yes      |         | Conditioning image as base64, a data URI, or a public URL          |
+| `quality`         | `string` | No       | `auto`  | Mesh pipeline: `auto`, `coarse`, `512`, or `1024`                  |
+| `background`      | `string` | No       | `auto`  | Background handling: `auto`, `keep`, `black`, or `white`           |
+| `step`            | `int`    | No       | 12      | Flow sampling steps for the shape                                  |
+| `texture_steps`   | `int`    | No       | 12      | Flow sampling steps for the PBR material                           |
+| `cfg_scale`       | `float`  | No       | 7.5     | Classifier-free guidance scale                                     |
+| `seed`            | `int`    | No       | random  | Random seed for reproducibility                                    |
+| `response_format` | `string` | No       | `url`   | `url` to return a file URL, `b64_json` for base64 output           |
+| `params`          | `object` | No       |         | Backend-specific string parameters (`texture_size`, `components`)  |
+
+`quality` selects the mesh resolution: `coarse` is a fast marching-cubes preview, `512` the fine dual-grid mesh, `1024` the high-resolution cascade (slow — several minutes, roughly 10 GB VRAM), and `auto` picks the best pipeline the installed model set supports.
+
+`background` controls solid-background removal on the conditioning image before generation: `auto` detects border-connected near-black/near-white, `keep` preserves the image alpha exactly, and `black`/`white` force removal of that colour.
+
+Backend-specific `params`: `texture_size` (UV-atlas resolution hint when atlas baking is enabled) and `components` (`tiny` removes small islands, `largest` keeps only the biggest connected component, `all` — the default — keeps everything).
+
+### Response
+
+Returns a JSON response using LocalAI's OpenAI-style generation envelope:
+
+| Field             | Type     | Description                                                    |
+|-------------------|----------|----------------------------------------------------------------|
+| `created`         | `int`    | Unix timestamp of generation                                   |
+| `id`              | `string` | Unique identifier (UUID)                                       |
+| `data`            | `array`  | Array with the generated asset                                 |
+| `data[].url`      | `string` | URL path to the `.glb` under `/generated-3d` (if `url`)        |
+| `data[].b64_json` | `string` | Base64-encoded GLB (if `response_format` is `b64_json`)        |
+
+### Watertight print remeshing
+
+`POST /3d/remesh` applies the same post-generation CGAL Alpha Wrap workflow as the trellis2.cpp demo. It accepts `multipart/form-data` and returns the remeshed GLB directly as `model/gltf-binary`:
+
+| Field    | Type     | Required | Default | Description |
+|----------|----------|----------|---------|-------------|
+| `model`  | `string` | Yes      |         | Installed TRELLIS.2 model name |
+| `mesh`   | `file`   | Yes      |         | Source GLB produced by TRELLIS.2 |
+| `detail` | `float`  | No       | `0.5`   | Smallest preserved detail as a percentage of the source bounding-box diagonal (`0.35`–`2.5`) |
+
+There is intentionally no independent offset control. The enclosing offset follows the trellis2.cpp demo and is derived as `detail / 30`; independent tuning tends to produce puffy or degenerate wraps. Lower detail percentages retain finer features but take longer and generally produce more triangles. The output is watertight, oriented, intersection-free, and 2-manifold. For textured sources, LocalAI unwraps the replacement mesh and reprojects its PBR material onto a new UV atlas.
+
+Source GLBs may be up to 512 MiB. This route uses its own upload limit because fine TRELLIS.2 meshes commonly exceed LocalAI's default `--upload-limit`.
+
+```bash
+curl http://localhost:8080/3d/remesh \
+  -F model=trellis2-4b \
+  -F mesh=@generated.glb \
+  -F detail=0.5 \
+  --output printable.glb
+```
+
+## Usage
+
+### Generate a 3D model from an image
+
+```bash
+curl http://localhost:8080/3d/generations \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "trellis2-4b",
+    "image": "https://example.com/photo-of-a-chair.png",
+    "quality": "512"
+  }'
+```
+
+The response contains a URL such as `/generated-3d/b64123456789.glb`; fetch it from the same server. The GLB is standard glTF 2.0 and opens in Blender, three.js, `<model-viewer>`, and most engines.
+
+### Base64 input and output
+
+```bash
+curl http://localhost:8080/3d/generations \
+  -H "Content-Type: application/json" \
+  -d "{
+    \"model\": \"trellis2-4b\",
+    \"image\": \"$(base64 -w0 chair.png)\",
+    \"response_format\": \"b64_json\"
+  }" | jq -r '.data[0].b64_json' | base64 -d > chair.glb
+```
+
+## WebUI
+
+The React UI includes a 3D tab in the Studio (and a `/3d` page) with an interactive PBR viewer: upload an image, pick the quality, and preview the generated mesh with orbit/pan/zoom and a wireframe toggle. Past generations are kept in the browser (IndexedDB). After generation, a single Detail slider and **Apply remeshing** button replace the preview with the exact watertight model that the GLB download exports; **Show original** switches back without regenerating.
+
+## Notes
+
+- The 512³ pipeline takes roughly two minutes on a modern GPU; the 1024³ cascade takes around five minutes and needs about 10 GB VRAM plus a temporary host-RAM spike.
+- `TRELLIS2_DEVICE=cpu` forces CPU inference (slow; mainly for debugging).
+- The generated mesh has unoriented winding (faithful to TRELLIS.2) and is exported Y-up with vertex-PBR materials; a UV-atlas texture bake can be enabled in the backend via the `T2GLB_XATLAS` environment variable.
diff --git a/docs/content/features/GPU-acceleration.md b/docs/content/features/GPU-acceleration.md
index 1b698da93..647aa84e1 100644
--- a/docs/content/features/GPU-acceleration.md
+++ b/docs/content/features/GPU-acceleration.md
@@ -166,12 +166,58 @@ Due to the nature of ROCm it is best to run all implementations in containers as
 - Make sure to do not use GPU assigned for compute for desktop rendering.
 - Ensure at least 100GB of free space on disk hosting container runtime and storing images prior to installation.
 
+### AMD Strix Halo / gfx1151 (RDNA 3.5)
+
+AMD Ryzen AI MAX+ (Strix Halo) APUs with an integrated Radeon 8060S (gfx1151 / RDNA 3.5) are
+supported with ROCm 7.11.0+. These systems provide up to 96 GB of unified VRAM accessible by the GPU.
+
+Tested on: Geekom A9 Mega (AMD Ryzen AI MAX+ 395, ROCm 7.11.0, Ubuntu 24.04, kernel 6.14).
+
+**Required kernel boot parameters** (add to `GRUB_CMDLINE_LINUX` in `/etc/default/grub`, then run `update-grub`):
+```
+iommu=pt amdgpu.gttsize=126976 ttm.pages_limit=32505856
+```
+
+**Required environment variables** for gfx1151 (set automatically in the ROCm/hipblas image):
+
+| Variable | Value | Purpose |
+|----------|-------|---------|
+| `HSA_OVERRIDE_GFX_VERSION` | `11.5.1` | Tells the HSA runtime to use gfx1151 code objects |
+| `ROCBLAS_USE_HIPBLASLT` | `1` | Prefer hipBLASLt over rocBLAS for GEMM (required for gfx1151) |
+| `HSA_XNACK` | `1` | Enable XNACK (memory-fault retry) for APU unified memory |
+| `HSA_ENABLE_SDMA` | `0` | Disable SDMA engine — causes hangs on APU/iGPU configs |
+
+> **Warning:** Do **not** set `GGML_CUDA_ENABLE_UNIFIED_MEMORY`. The C-level check is
+> `getenv(...) != nullptr`, so even `=0` activates `hipMallocManaged` (allocates from
+> system RAM instead of the 96 GB VRAM pool).
+
+**Running LocalAI on gfx1151** (use the standard ROCm/hipblas image — there is only one ROCm image, and it ships with ROCm 7.x by default):
+```yaml
+    image: quay.io/go-skynet/local-ai:master-gpu-hipblas
+    environment:
+      - HSA_OVERRIDE_GFX_VERSION=11.5.1
+      - ROCBLAS_USE_HIPBLASLT=1
+      - HSA_XNACK=1
+      - HSA_ENABLE_SDMA=0
+    devices:
+      - /dev/dri
+      - /dev/kfd
+    group_add:
+      - video
+```
+
+> **Note:** When updating the image, always recreate the container (`docker compose up --force-recreate`)
+> rather than just restarting it. `docker compose restart` preserves the old container environment
+> and will not pick up updated env vars from the image.
+
+For llama.cpp models, enable flash attention (`--flash-attention`) and disable mmap (`--no-mmap`) for best performance on APU systems.
+
 ### Limitations
 
 Ongoing verification testing of ROCm compatibility with integrated backends.
 Please note the following list of verified backends and devices.
 
-LocalAI hipblas images are built against the following targets: gfx908, gfx90a, gfx942, gfx950, gfx1030, gfx1100, gfx1101, gfx1102, gfx1200, gfx1201
+LocalAI hipblas images are built against the following targets: gfx908, gfx90a, gfx942, gfx950, gfx1030, gfx1100, gfx1101, gfx1102, gfx1151, gfx1200, gfx1201
 
 **Note:** Starting with ROCm 6.4, AMD removed rocBLAS kernel support for older architectures (gfx803, gfx900, gfx906). Since llama.cpp and other backends depend on rocBLAS for matrix operations, these GPUs (e.g. Radeon VII) are no longer supported in pre-built images.
 
@@ -181,14 +227,15 @@ If your device is not one of the above targets, you must specify the correspondi
 
 The devices in the following list have been tested with `hipblas` images.
 
-| Backend | Verified | Devices |
-| ---- | ---- | ---- |
-| llama.cpp | yes | MI100 (gfx908), MI210/250 (gfx90a) |
-| diffusers | yes | MI100 (gfx908), MI210/250 (gfx90a) |
-| whisper | no | none |
-| coqui | no | none |
-| transformers | no | none |
-| vllm | no | none |
+| Backend | Verified | Devices | ROCm Version |
+| ---- | ---- | ---- | ---- |
+| llama.cpp | yes | MI100 (gfx908), MI210/250 (gfx90a) | 7.x |
+| llama.cpp | yes | Radeon 8060S / gfx1151 (Strix Halo) | 7.11.0 |
+| diffusers | yes | MI100 (gfx908), MI210/250 (gfx90a) | 7.x |
+| whisper | no | none | - |
+| coqui | no | none | - |
+| transformers | no | none | - |
+| vllm | no | none | - |
 
 **You can help by expanding this list.**
 
diff --git a/docs/content/features/backends.md b/docs/content/features/backends.md
index 646fe4474..2ed52cf06 100644
--- a/docs/content/features/backends.md
+++ b/docs/content/features/backends.md
@@ -135,8 +135,9 @@ LocalAI supports various types of backends:
 - **Sound Generation Backends**: For music and audio generation (e.g., ACE-Step)
 - **Sound Classification Backends**: For sound-event classification / audio tagging - identifying everyday sounds like baby cry, glass breaking, alarms (e.g., ced.cpp)
 - **Image & Video Generation Backends**: For diffusion and audio-conditioned avatar models (e.g., stable-diffusion.cpp, diffusers, vLLM-Omni, [LongCat-Video]({{%relref "features/video-generation" %}}))
+- **3D Generation Backends**: For image-to-3D mesh generation ([trellis2.cpp]({{%relref "features/3d-generation" %}}) — Microsoft TRELLIS.2, producing GLB assets with PBR textures)
 - **Vision & Detection Backends**: For object detection, segmentation, depth, and face/voice recognition (e.g., rf-detr.cpp, locate-anything.cpp, sam3.cpp, insightface)
 - **Audio Processing Backends**: For voice activity detection and audio enhancement (e.g., Silero VAD, LocalVQE)
-- **Utility Backends**: For reranking, PII/NER token classification, fine-tuning, quantization, and vector storage (e.g., rerankers, privacy-filter.cpp, TRL, local-store)
+- **Utility Backends**: For reranking, PII/NER token classification, fine-tuning, quantization, and vector storage (e.g., rerankers, privacy-filter.cpp, TRL, local-store, valkey-store)
 
 See the [Backend & Model Compatibility Table]({{%relref "reference/compatibility-table" %}}) for the full catalog.
diff --git a/docs/content/features/openai-realtime.md b/docs/content/features/openai-realtime.md
index 4a139c687..bf2a50b6f 100644
--- a/docs/content/features/openai-realtime.md
+++ b/docs/content/features/openai-realtime.md
@@ -347,6 +347,31 @@ By default a realtime session responds with audio plus a transcript. To make the
 
 The GA `output_modalities` wins when both are present. A response-level value overrides the session-level one, and when neither is set the session falls back to `["audio"]`.
 
+### Out-of-band responses and `metadata`
+
+A response can be created outside the default conversation by setting `conversation` to `none`. The reply is not added to the conversation history, which is what makes it usable for a side channel — answering a chat message or a webhook while a spoken conversation is in progress.
+
+Because such a response arrives on the same socket as everything else, attach `metadata` to correlate it. LocalAI echoes the map back verbatim on both `response.created` and `response.done`:
+
+```json
+{"type": "response.create", "response": {
+  "conversation": "none",
+  "output_modalities": ["text"],
+  "metadata": {"client_run": "abc123"},
+  "input": [{"type": "message", "role": "user",
+             "content": [{"type": "input_text", "text": "Is the oven still on?"}]}]
+}}
+```
+
+```json
+{"type": "response.done", "response": {
+  "id": "resp_...", "status": "completed",
+  "metadata": {"client_run": "abc123"}
+}}
+```
+
+Keys are strings up to 64 characters, values up to 512. A response created without `metadata` omits the field rather than sending an empty object.
+
 ## Gating a realtime pipeline with voice recognition
 
 A pipeline realtime model can require speaker verification before it responds. Add a `voice_recognition` block under `pipeline`. When present, each committed utterance is verified against authorized speakers; unauthorized utterances are dropped before the LLM runs (no LLM call, no tool execution, no TTS). The session stays open.
diff --git a/docs/content/features/stores.md b/docs/content/features/stores.md
index de902b69b..1d2a3d851 100644
--- a/docs/content/features/stores.md
+++ b/docs/content/features/stores.md
@@ -44,8 +44,107 @@ Keys are arrays of floating point numbers with a maximum width of 32bits. Values
 The key vectors must all be the same length and it's best for search performance if they are normalized. When
 addings keys it will be detected if they are not normalized and what length they are.
 
-All endpoints accept a `store` field which specifies which store to operate on. Presently they are created
-on the fly and there is only one store backend so no configuration is required.
+All endpoints accept a `store` field which specifies which store to operate on. Stores are created
+on the fly. By default the in-memory `local-store` backend is used and no configuration is required,
+but you can select a different store backend per request (see [Backends](#backends) below).
+
+## Backends
+
+Each `/stores/*` request accepts an optional `backend` field selecting the store implementation.
+Two backends ship with LocalAI:
+
+| Backend | `backend` value | Persistence | Notes |
+|---------|-----------------|-------------|-------|
+| Local (default) | `local-store` (alias `embedded-store`) | In-memory, lost on restart | Exact cosine similarity, zero configuration. |
+| Valkey Search | `valkey-store` (alias `valkey`) | Durable (Valkey RDB/AOF) | Backed by a Valkey Search (`FT.*`) server; survives restarts and supports opt-in HNSW. |
+
+### Valkey store backend
+
+The `valkey-store` backend persists vectors in a [Valkey Search](https://valkey.io/) server, so
+the data survives a LocalAI restart — unlike the in-memory default. It requires a reachable server
+that ships the Valkey Search module (for example the `valkey/valkey-bundle` image).
+
+Select it by passing `"backend": "valkey-store"` (or the `"valkey"` alias) on any `/stores/*` request:
+
+```
+curl -X POST http://localhost:8080/stores/set \
+     -H "Content-Type: application/json" \
+     -d '{"backend": "valkey-store", "store": "my-vectors", "keys": [[0.1, 0.2], [0.3, 0.4]], "values": ["foo", "bar"]}'
+```
+
+The connection and index are configured through a **model config** named after the store (the
+`store` field on the request, which is the store's model ID). Create a YAML in your models
+directory whose `name` matches the store, set `backend: valkey-store`, and put the connection /
+index settings in the `options:` list as `key:value` strings. Because each store resolves its own
+config, different stores can point at different Valkey servers or use different index settings
+within one LocalAI process:
+
+```yaml
+name: my-vectors
+backend: valkey-store
+options:
+  - addr:valkey.internal:6379
+  - username_env:MY_VALKEY_USER
+  - password_env:MY_VALKEY_PASSWORD
+  - index_algo:HNSW
+  - distance_metric:COSINE
+```
+
+The `username_env` / `password_env` options name an environment variable that holds the actual
+credential rather than putting the secret directly in the YAML. This mirrors `cloud-proxy`'s
+`api_key_env` pattern and lets distinct store configs each reference their own credentials.
+The direct `username` / `password` options still work for backward compatibility, and take
+precedence when both are set.
+
+When no config exists for a store, the backend connects to `localhost:6379` with the defaults
+below (so the zero-config experience still works).
+
+| Option | Default | Description |
+|--------|---------|-------------|
+| `addr` | `localhost:6379` | Valkey server address (`host:port`). |
+| `username` | *(empty)* | Optional ACL username (plaintext in config). |
+| `password` | *(empty)* | Optional password / ACL secret (plaintext in config). |
+| `username_env` | *(empty)* | Name of an env var that holds the username. Preferred over `username` for secrets — keeps credentials out of the model YAML. |
+| `password_env` | *(empty)* | Name of an env var that holds the password. Preferred over `password` for secrets. |
+| `tls` | `false` | Enable TLS (required by many managed deployments). |
+| `tls_ca_cert` | *(empty)* | Path to a PEM CA bundle used to verify the server certificate (self-signed / private CA). |
+| `tls_skip_verify` | `false` | Skip TLS certificate verification. Insecure — for testing only. |
+| `client_name` | `localai-valkey-store` | Connection name reported by `CLIENT LIST`. Always set. |
+| `db` | `0` | Logical Valkey DB index (`SELECT n`). Namespace prefixing already isolates keyspaces on a shared DB. |
+| `index_algo` | `FLAT` | `FLAT` (exact, default) or `HNSW` (approximate ANN for large corpora). |
+| `hnsw_m` | `16` | HNSW graph degree (only when `index_algo:HNSW`). |
+| `hnsw_ef_construction` | `200` | HNSW build-time candidate list (HNSW only). |
+| `hnsw_ef_runtime` | `10` | HNSW query-time candidate list (HNSW only). |
+| `distance_metric` | `COSINE` | `COSINE` (default), `L2` or `IP`. |
+| `request_timeout_ms` | `5000` | Per-command timeout in milliseconds. |
+
+For `COSINE` the returned `similarities` follow the same convention as the local store
+(`1.0` = identical, `-1.0` = opposite); internally Valkey returns a cosine *distance* which the
+backend converts with `similarity = 1 - distance`. For `L2` and `IP` the raw Valkey score is
+returned in `similarities` (for `L2`, smaller means closer — the opposite ordering of `COSINE`);
+nearest-first ordering of the results is preserved in all cases.
+
+{{% notice note %}}
+Valkey Search updates its vector index **asynchronously** after a write. A `find` issued
+immediately after `set` may not yet see the new vectors — poll `find` briefly (or retry) until the
+expected results appear. `get` and `delete` are synchronous and unaffected.
+{{% /notice %}}
+
+{{% notice note %}}
+This backend targets a **standalone** Valkey Search server (one server per namespace/model). Valkey
+Cluster is not a supported target yet — index coordination across shards is out of scope for this
+backend.
+{{% /notice %}}
+
+{{% notice warning %}}
+`tls` defaults to `false` (plaintext). Set `tls:true` whenever the Valkey server is
+not on `localhost` or a `password`/`username` is configured, otherwise the credentials
+and the stored vectors travel the network unencrypted. The TLS `ServerName` (SNI) is derived from
+the host portion of `addr`, so certificate verification works for both hostname and
+IP-addressed endpoints. For a self-signed / private CA, point `tls_ca_cert` at the PEM
+bundle; `tls_skip_verify:true` disables verification entirely and should only be used for
+local testing.
+{{% /notice %}}
 
 ## Set
 
diff --git a/docs/content/reference/compatibility-table.md b/docs/content/reference/compatibility-table.md
index 4ac4cec77..8e739f93e 100644
--- a/docs/content/reference/compatibility-table.md
+++ b/docs/content/reference/compatibility-table.md
@@ -120,6 +120,7 @@ All backends listed here can be installed on demand from the [Backend Gallery]({
 | [rerankers](https://github.com/AnswerDotAI/rerankers) | Document reranking for RAG | CUDA 12, ROCm, Intel SYCL, Metal |
 | [privacy-filter.cpp](https://github.com/localai-org/privacy-filter.cpp) | Standalone GGML engine for the openai-privacy-filter PII/NER token-classification model family (powers LocalAI's PII redaction tier) | CPU, CUDA 13, Vulkan |
 | [local-store](https://github.com/mudler/LocalAI) | Local-first vector database for embeddings | CPU, Metal |
+| [valkey-store](https://github.com/mudler/LocalAI) | Durable vector store for embeddings backed by Valkey Search (FLAT or HNSW) | CPU, Metal |
 | [TRL](https://github.com/huggingface/trl) | Fine-tuning (SFT, DPO, GRPO, RLOO, KTO, ORPO) | CPU, CUDA 12/13 |
 | [llama.cpp quantization](https://github.com/ggml-org/llama.cpp) | HuggingFace → GGUF model conversion and quantization | CPU, Metal |
 
diff --git a/gallery/index.yaml b/gallery/index.yaml
index 1144c7e4b..9776061a2 100644
--- a/gallery/index.yaml
+++ b/gallery/index.yaml
@@ -1,4 +1,230 @@
 ---
+- &kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev"
+  variants:
+    - model: kat-coder-v2.5-dev-q8
+    - model: kat-coder-v2.5-dev-apex-i-quality
+    - model: kat-coder-v2.5-dev-apex-i-balanced
+    - model: kat-coder-v2.5-dev-apex-i-compact
+    - model: kat-coder-v2.5-dev-apex-i-mini
+    - model: kat-coder-v2.5-dev-apex-quality
+    - model: kat-coder-v2.5-dev-apex-balanced
+    - model: kat-coder-v2.5-dev-apex-compact
+  url: "github:mudler/LocalAI/gallery/virtual.yaml@master"
+  urls:
+    - https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev
+    - https://huggingface.co/bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF
+    - https://huggingface.co/mudler/KAT-Coder-V2.5-Dev-APEX-GGUF
+  description: |
+    KAT-Coder-V2.5-Dev is an Apache-2.0 agentic coding model from Kwaipilot,
+    post-trained from Qwen3.6-35B-A3B. It has 35 billion total parameters with
+    3 billion activated per token, a 262K-token context window, and text-only
+    weights tuned for repository-level coding and tool use.
+
+    This entry offers standard Q4_K_M and Q8_0 GGUF quantizations alongside
+    APEX mixed-precision variants with quality, balanced, compact, and mini
+    profiles. The APEX I-profiles use importance-matrix calibration.
+  license: "apache-2.0"
+  icon: https://huggingface.co/Kwaipilot/KAT-Coder-V2.5-Dev/resolve/main/kat_logo_hd.png
+  tags:
+    - llm
+    - gguf
+    - qwen
+    - coding
+  last_checked: "2026-07-29"
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF/Kwaipilot_KAT-Coder-V2.5-Dev-Q4_K_M.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF/Kwaipilot_KAT-Coder-V2.5-Dev-Q4_K_M.gguf
+      sha256: 4221c26e5663502d1c96fc901c9967d0e70ce2dcfaa5a9fb9280a46bd19e3c07
+      uri: huggingface://bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF/Kwaipilot_KAT-Coder-V2.5-Dev-Q4_K_M.gguf
+- !!merge <<: *kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev-q8"
+  variants: []
+  description: |
+    KAT-Coder-V2.5-Dev is an Apache-2.0 agentic coding model from Kwaipilot,
+    post-trained from Qwen3.6-35B-A3B. It has 35 billion total parameters with
+    3 billion activated per token, a 262K-token context window, and text-only
+    weights tuned for repository-level coding and tool use.
+
+    This entry uses the higher-quality Q8_0 GGUF quantization.
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF/Kwaipilot_KAT-Coder-V2.5-Dev-Q8_0.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF/Kwaipilot_KAT-Coder-V2.5-Dev-Q8_0.gguf
+      sha256: 5fa510f44779b0e3d38a6678985f417a1c65e3000405ca5d6dcf7fd065e47a15
+      uri: huggingface://bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-GGUF/Kwaipilot_KAT-Coder-V2.5-Dev-Q8_0.gguf
+- !!merge <<: *kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev-apex-i-quality"
+  variants: []
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Quality.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Quality.gguf
+      sha256: e1cf7f33e13ee787a8557effee41eef5261b24696e283ef9d320830ba39f6784
+      uri: huggingface://mudler/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Quality.gguf
+- !!merge <<: *kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev-apex-i-balanced"
+  variants: []
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Balanced.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Balanced.gguf
+      sha256: ee6e0ec15964c42ba91831d13e9709239f1b74da3dc49dd3edec4ad6aed8029f
+      uri: huggingface://mudler/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Balanced.gguf
+- !!merge <<: *kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev-apex-i-compact"
+  variants: []
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Compact.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Compact.gguf
+      sha256: 5235ac39e7989d9fcf08078acfa755611d42a1f80060d226e4aba04a3595d0d8
+      uri: huggingface://mudler/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Compact.gguf
+- !!merge <<: *kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev-apex-i-mini"
+  variants: []
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Mini.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Mini.gguf
+      sha256: 9d901bf1c44946840e15e6f73a66782f56ae2f1a41d6508995ddcd976e9af878
+      uri: huggingface://mudler/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-I-Mini.gguf
+- !!merge <<: *kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev-apex-quality"
+  variants: []
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Quality.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Quality.gguf
+      sha256: 849737baf59183ed518d34f2460f0dcd86190953ccc3213072bdcb1d7f8d2882
+      uri: huggingface://mudler/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Quality.gguf
+- !!merge <<: *kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev-apex-balanced"
+  variants: []
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Balanced.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Balanced.gguf
+      sha256: 9a5d31b110a95bc085d9851c420dd7a1ffeaec7c9263073c8850d8fb81c4f8ba
+      uri: huggingface://mudler/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Balanced.gguf
+- !!merge <<: *kat-coder-v2-5-dev
+  name: "kat-coder-v2.5-dev-apex-compact"
+  variants: []
+  overrides:
+    backend: llama-cpp
+    function:
+      automatic_tool_parsing_fallback: true
+      grammar:
+        disable: true
+    known_usecases:
+      - chat
+    options:
+      - use_jinja:true
+    parameters:
+      model: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Compact.gguf
+    template:
+      use_tokenizer_template: true
+  files:
+    - filename: llama-cpp/models/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Compact.gguf
+      sha256: 98dfee53102bbf01e67a768066543a707b40f365ff71e35132ac432c58ad99fd
+      uri: huggingface://mudler/KAT-Coder-V2.5-Dev-APEX-GGUF/KAT-Coder-V2.5-Dev-APEX-Compact.gguf
 - name: "inkling"
   url: "github:mudler/LocalAI/gallery/virtual.yaml@master"
   urls:
@@ -31062,6 +31288,101 @@
     - filename: clip_vision_h.safetensors
       sha256: ""
       uri: https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors
+- name: trellis2-4b
+  url: github:mudler/LocalAI/gallery/trellis2cpp.yaml@master
+  icon: https://cdn-avatars.huggingface.co/v1/production/uploads/1583646260758-5e64858c87403103f9f1055d.png
+  urls:
+    - https://huggingface.co/microsoft/TRELLIS.2-4B
+    - https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF
+    - https://huggingface.co/LocalAI-io/TRELLIS-image-large-GGUF
+    - https://huggingface.co/LocalAI-io/dinov3-vitl16-pretrain-lvd1689m-GGUF
+  description: |
+    Microsoft TRELLIS.2-4B image-to-3D generation via trellis2.cpp (C++/ggml).
+    Generates a textured 3D mesh (binary glTF / GLB with PBR materials) from a
+    single conditioning image. This is the full pipeline: coarse preview, 512
+    fine dual-grid, 1024 cascade, and PBR texturing (~18 GB of f16 GGUFs;
+    the 1024 textured path needs roughly 10 GB VRAM).
+  license: mit
+  tags:
+    - trellis
+    - image-to-3d
+    - 3d-generation
+    - gguf
+  last_checked: "2026-07-16"
+  overrides:
+    parameters:
+      model: ss_flow_f16.gguf
+  files:
+    - filename: dino_f16.gguf
+      sha256: 385d8186a38a2328ec740fb2ac1f33f9194d8774efc7ccafd4aa2e51cf5f6450
+      uri: huggingface://LocalAI-io/dinov3-vitl16-pretrain-lvd1689m-GGUF/dino_f16.gguf
+    - filename: ss_flow_f16.gguf
+      sha256: 1dded5b74237d24e6876a642a26f90b43742e3554418573860f810e3bbe61e8c
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/ss_flow_f16.gguf
+    - filename: ss_dec_f16.gguf
+      sha256: 9c2210b7ed830fdc8286961a8189878ff5bcfd3bfc83ab4eacee005d293d2185
+      uri: huggingface://LocalAI-io/TRELLIS-image-large-GGUF/ss_dec_f16.gguf
+    - filename: slat_flow_f16.gguf
+      sha256: 2f94bad7b1c524ad8c01943bc38fcc0c314e7d482ce896f3c6e96eb6e7cec15c
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/slat_flow_f16.gguf
+    - filename: slat_flow_1024_f16.gguf
+      sha256: b6a2270131e2e9235e9b6cb525193eb85ae132fa5af3274322aacd39e40a6bc5
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/slat_flow_1024_f16.gguf
+    - filename: shape_dec_f16.gguf
+      sha256: 6fe53f1d7763dabf7c8d72bc38f4053d87fde6f65bf17a9d378d27edb39d3530
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/shape_dec_f16.gguf
+    - filename: shape_enc_f16.gguf
+      sha256: 3ec80ff580987fcdb9bc594fc8b6fda890d63101ca442eb2b26f5dc315e8696c
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/shape_enc_f16.gguf
+    - filename: tex_dec_f16.gguf
+      sha256: afd304f4dfcb8c94df851b85519b415b99f04070f7d29de1320c50631b1be4e0
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/tex_dec_f16.gguf
+    - filename: tex_slat_flow_512_f16.gguf
+      sha256: 89a081b7f5487a5b31f03d240e4d959a56db0cc2c46c327230097a2554da52ae
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/tex_slat_flow_512_f16.gguf
+    - filename: tex_slat_flow_1024_f16.gguf
+      sha256: bbb55b0910c7929aac5e0612a9bb15113837a2c674cafb9f0f170eda8b5558a8
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/tex_slat_flow_1024_f16.gguf
+- name: trellis2-4b-geometry
+  url: github:mudler/LocalAI/gallery/trellis2cpp.yaml@master
+  icon: https://cdn-avatars.huggingface.co/v1/production/uploads/1583646260758-5e64858c87403103f9f1055d.png
+  urls:
+    - https://huggingface.co/microsoft/TRELLIS.2-4B
+    - https://huggingface.co/LocalAI-io/TRELLIS.2-4B-GGUF
+    - https://huggingface.co/LocalAI-io/TRELLIS-image-large-GGUF
+    - https://huggingface.co/LocalAI-io/dinov3-vitl16-pretrain-lvd1689m-GGUF
+  description: |
+    Microsoft TRELLIS.2-4B image-to-3D generation via trellis2.cpp (C++/ggml) —
+    geometry-only variant (~7 GB): coarse preview plus the 512 fine dual-grid
+    mesh, without the 1024 cascade or PBR texturing. The backend detects the
+    reduced model set automatically; install trellis2-4b instead for textured
+    output.
+  license: mit
+  tags:
+    - trellis
+    - image-to-3d
+    - 3d-generation
+    - gguf
+  last_checked: "2026-07-16"
+  overrides:
+    parameters:
+      model: ss_flow_f16.gguf
+  files:
+    - filename: dino_f16.gguf
+      sha256: 385d8186a38a2328ec740fb2ac1f33f9194d8774efc7ccafd4aa2e51cf5f6450
+      uri: huggingface://LocalAI-io/dinov3-vitl16-pretrain-lvd1689m-GGUF/dino_f16.gguf
+    - filename: ss_flow_f16.gguf
+      sha256: 1dded5b74237d24e6876a642a26f90b43742e3554418573860f810e3bbe61e8c
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/ss_flow_f16.gguf
+    - filename: ss_dec_f16.gguf
+      sha256: 9c2210b7ed830fdc8286961a8189878ff5bcfd3bfc83ab4eacee005d293d2185
+      uri: huggingface://LocalAI-io/TRELLIS-image-large-GGUF/ss_dec_f16.gguf
+    - filename: slat_flow_f16.gguf
+      sha256: 2f94bad7b1c524ad8c01943bc38fcc0c314e7d482ce896f3c6e96eb6e7cec15c
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/slat_flow_f16.gguf
+    - filename: shape_dec_f16.gguf
+      sha256: 6fe53f1d7763dabf7c8d72bc38f4053d87fde6f65bf17a9d378d27edb39d3530
+      uri: huggingface://LocalAI-io/TRELLIS.2-4B-GGUF/shape_dec_f16.gguf
 - name: sd-1.5-ggml
   url: github:mudler/LocalAI/gallery/sd-ggml.yaml@master
   urls:
diff --git a/gallery/trellis2cpp.yaml b/gallery/trellis2cpp.yaml
new file mode 100644
index 000000000..02d056691
--- /dev/null
+++ b/gallery/trellis2cpp.yaml
@@ -0,0 +1,9 @@
+---
+name: "trellis2cpp"
+
+config_file: |
+    backend: trellis2cpp
+    known_usecases:
+    - 3d
+    step: 12
+    cfg_scale: 7.5
diff --git a/go.mod b/go.mod
index f2af28655..8982bf9f7 100644
--- a/go.mod
+++ b/go.mod
@@ -65,6 +65,8 @@ require (
 	github.com/testcontainers/testcontainers-go/modules/nats v0.42.0
 	github.com/testcontainers/testcontainers-go/modules/postgres v0.42.0
 	github.com/timbutler/zxcvbn v1.0.4
+	github.com/valkey-io/valkey-go v1.0.76
+	github.com/valkey-io/valkey-go/mock v1.0.76
 	github.com/yalue/onnxruntime_go v1.11.0
 	go.opentelemetry.io/otel v1.44.0
 	go.opentelemetry.io/otel/exporters/prometheus v0.66.0
@@ -328,7 +330,7 @@ require (
 	github.com/yosida95/uritemplate/v3 v3.0.2 // indirect
 	go.opentelemetry.io/auto/sdk v1.2.1 // indirect
 	go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.67.0 // indirect
-	go.uber.org/mock v0.5.2 // indirect
+	go.uber.org/mock v0.6.0
 	go.yaml.in/yaml/v2 v2.4.4
 	go.yaml.in/yaml/v3 v3.0.4 // indirect
 	golang.org/x/image v0.38.0 // indirect
diff --git a/go.sum b/go.sum
index 6b32a4dcb..fbde4ca44 100644
--- a/go.sum
+++ b/go.sum
@@ -1351,6 +1351,10 @@ github.com/ulikunitz/xz v0.5.9/go.mod h1:nbz6k7qbPmH4IRqmfOplQw/tblSgqTqBwxkY0oW
 github.com/ulikunitz/xz v0.5.14 h1:uv/0Bq533iFdnMHZdRBTOlaNMdb1+ZxXIlHDZHIHcvg=
 github.com/ulikunitz/xz v0.5.14/go.mod h1:nbz6k7qbPmH4IRqmfOplQw/tblSgqTqBwxkY0oWt/14=
 github.com/urfave/cli v1.22.10/go.mod h1:Gos4lmkARVdJ6EkW0WaNv/tZAAMe9V7XWyB60NtXRu0=
+github.com/valkey-io/valkey-go v1.0.76 h1:Rcown7FFseVhG9b0+4MWfMs4xWu8otPzHjrsK044ET4=
+github.com/valkey-io/valkey-go v1.0.76/go.mod h1:6X581PhgfeMkJmyfjIsa2eFdq6dy3Qkkg9zwjM1p42M=
+github.com/valkey-io/valkey-go/mock v1.0.76 h1:9pyITMqVKiQtmdMh3P0HIeiwJhTzAhxzWTYs2goStfA=
+github.com/valkey-io/valkey-go/mock v1.0.76/go.mod h1:kmM1d1YT1Vrm7xjHjyDjcq+IUeFlzYMlboUWF5izvNw=
 github.com/valyala/bytebufferpool v1.0.0 h1:GqA5TC/0021Y/b9FG4Oi9Mr3q7XYx6KllzawFIhcdPw=
 github.com/valyala/bytebufferpool v1.0.0/go.mod h1:6bBcMArwyJ5K/AmCkWv1jt77kVWyCJ6HpOuEn7z0Csc=
 github.com/valyala/fasthttp v1.68.0 h1:v12Nx16iepr8r9ySOwqI+5RBJ/DqTxhOy1HrHoDFnok=
@@ -1445,8 +1449,8 @@ go.uber.org/fx v1.24.0 h1:wE8mruvpg2kiiL1Vqd0CC+tr0/24XIB10Iwp2lLWzkg=
 go.uber.org/fx v1.24.0/go.mod h1:AmDeGyS+ZARGKM4tlH4FY2Jr63VjbEDJHtqXTGP5hbo=
 go.uber.org/goleak v1.3.0 h1:2K3zAYmnTNqV73imy9J1T3WC+gmCePx2hEGkimedGto=
 go.uber.org/goleak v1.3.0/go.mod h1:CoHD4mav9JJNrW/WLlf7HGZPjdw8EucARQHekz1X6bE=
-go.uber.org/mock v0.5.2 h1:LbtPTcP8A5k9WPXj54PPPbjcI4Y6lhyOZXn+VS7wNko=
-go.uber.org/mock v0.5.2/go.mod h1:wLlUxC2vVTPTaE3UD51E0BGOAElKrILxhVSDYQLld5o=
+go.uber.org/mock v0.6.0 h1:hyF9dfmbgIX5EfOdasqLsWD6xqpNZlXblLB/Dbnwv3Y=
+go.uber.org/mock v0.6.0/go.mod h1:KiVJ4BqZJaMj4svdfmHM0AUx4NJYO8ZNpPnZn1Z+BBU=
 go.uber.org/multierr v1.5.0/go.mod h1:FeouvMocqHpRaaGuG9EjoKcStLC43Zu/fmqdUMPcKYU=
 go.uber.org/multierr v1.6.0/go.mod h1:cdWPpRnG4AhwMwsgIHip0KRBQjJy5kYEpYjJxpXp9iU=
 go.uber.org/multierr v1.11.0 h1:blXXJkSxSSfBVBlC76pxqeO+LN3aDfLQo+309xJstO0=
diff --git a/pkg/grpc/backend.go b/pkg/grpc/backend.go
index de4c51488..00e449193 100644
--- a/pkg/grpc/backend.go
+++ b/pkg/grpc/backend.go
@@ -73,6 +73,7 @@ type InferenceBackend interface {
 	Predict(ctx context.Context, in *pb.PredictOptions, opts ...grpc.CallOption) (*pb.Reply, error)
 	GenerateImage(ctx context.Context, in *pb.GenerateImageRequest, opts ...grpc.CallOption) (*pb.Result, error)
 	GenerateVideo(ctx context.Context, in *pb.GenerateVideoRequest, opts ...grpc.CallOption) (*pb.Result, error)
+	Generate3D(ctx context.Context, in *pb.Generate3DRequest, opts ...grpc.CallOption) (*pb.Result, error)
 	TTS(ctx context.Context, in *pb.TTSRequest, opts ...grpc.CallOption) (*pb.Result, error)
 	TTSStream(ctx context.Context, in *pb.TTSRequest, f func(reply *pb.Reply), opts ...grpc.CallOption) error
 	SoundGeneration(ctx context.Context, in *pb.SoundGenerationRequest, opts ...grpc.CallOption) (*pb.Result, error)
diff --git a/pkg/grpc/base/base.go b/pkg/grpc/base/base.go
index d1aef3cb6..0cf8d9162 100644
--- a/pkg/grpc/base/base.go
+++ b/pkg/grpc/base/base.go
@@ -59,6 +59,10 @@ func (llm *Base) GenerateVideo(*pb.GenerateVideoRequest) error {
 	return fmt.Errorf("unimplemented")
 }
 
+func (llm *Base) Generate3D(*pb.Generate3DRequest) error {
+	return fmt.Errorf("unimplemented")
+}
+
 func (llm *Base) AudioTranscription(context.Context, *pb.TranscriptRequest) (pb.TranscriptResult, error) {
 	return pb.TranscriptResult{}, fmt.Errorf("unimplemented")
 }
diff --git a/pkg/grpc/client.go b/pkg/grpc/client.go
index 5ea6d8c62..6ec3e57a8 100644
--- a/pkg/grpc/client.go
+++ b/pkg/grpc/client.go
@@ -247,6 +247,23 @@ func (c *Client) GenerateVideo(ctx context.Context, in *pb.GenerateVideoRequest,
 	return client.GenerateVideo(ctx, in, opts...)
 }
 
+func (c *Client) Generate3D(ctx context.Context, in *pb.Generate3DRequest, opts ...grpc.CallOption) (*pb.Result, error) {
+	if !c.parallel {
+		c.opMutex.Lock()
+		defer c.opMutex.Unlock()
+	}
+	c.setBusy(true)
+	defer c.setBusy(false)
+	defer c.wdMark()()
+	conn, err := c.dial()
+	if err != nil {
+		return nil, err
+	}
+	defer func() { _ = conn.Close() }()
+	client := pb.NewBackendClient(conn)
+	return client.Generate3D(ctx, in, opts...)
+}
+
 func (c *Client) TTS(ctx context.Context, in *pb.TTSRequest, opts ...grpc.CallOption) (*pb.Result, error) {
 	if !c.parallel {
 		c.opMutex.Lock()
diff --git a/pkg/grpc/embed.go b/pkg/grpc/embed.go
index 265297669..4c0bb2404 100644
--- a/pkg/grpc/embed.go
+++ b/pkg/grpc/embed.go
@@ -53,6 +53,10 @@ func (e *embedBackend) GenerateVideo(ctx context.Context, in *pb.GenerateVideoRe
 	return e.s.GenerateVideo(ctx, in)
 }
 
+func (e *embedBackend) Generate3D(ctx context.Context, in *pb.Generate3DRequest, opts ...grpc.CallOption) (*pb.Result, error) {
+	return e.s.Generate3D(ctx, in)
+}
+
 func (e *embedBackend) TTS(ctx context.Context, in *pb.TTSRequest, opts ...grpc.CallOption) (*pb.Result, error) {
 	return e.s.TTS(ctx, in)
 }
diff --git a/pkg/grpc/interface.go b/pkg/grpc/interface.go
index c968e7461..25df96233 100644
--- a/pkg/grpc/interface.go
+++ b/pkg/grpc/interface.go
@@ -18,6 +18,7 @@ type AIModel interface {
 	Embeddings(*pb.PredictOptions) ([]float32, error)
 	GenerateImage(*pb.GenerateImageRequest) error
 	GenerateVideo(*pb.GenerateVideoRequest) error
+	Generate3D(*pb.Generate3DRequest) error
 	Detect(*pb.DetectOptions) (pb.DetectResponse, error)
 	Depth(*pb.DepthRequest) (pb.DepthResponse, error)
 	FaceVerify(*pb.FaceVerifyRequest) (pb.FaceVerifyResponse, error)
diff --git a/pkg/grpc/server.go b/pkg/grpc/server.go
index 9896a76dc..f9d1020e1 100644
--- a/pkg/grpc/server.go
+++ b/pkg/grpc/server.go
@@ -165,6 +165,18 @@ func (s *server) GenerateVideo(ctx context.Context, in *pb.GenerateVideoRequest)
 	return &pb.Result{Message: "Video generated", Success: true}, nil
 }
 
+func (s *server) Generate3D(ctx context.Context, in *pb.Generate3DRequest) (*pb.Result, error) {
+	if s.llm.Locking() {
+		s.llm.Lock()
+		defer s.llm.Unlock()
+	}
+	err := s.llm.Generate3D(in)
+	if err != nil {
+		return &pb.Result{Message: fmt.Sprintf("Error generating 3D asset: %s", err.Error()), Success: false}, err
+	}
+	return &pb.Result{Message: "3D asset generated", Success: true}, nil
+}
+
 func (s *server) TTS(ctx context.Context, in *pb.TTSRequest) (*pb.Result, error) {
 	if err := s.checkModelIdentity(in); err != nil {
 		return nil, err
diff --git a/pkg/model/connection_evicting_client.go b/pkg/model/connection_evicting_client.go
index 0053df3dc..00d42d200 100644
--- a/pkg/model/connection_evicting_client.go
+++ b/pkg/model/connection_evicting_client.go
@@ -72,6 +72,12 @@ func (c *ConnectionEvictingClient) GenerateVideo(ctx context.Context, in *pb.Gen
 	return result, err
 }
 
+func (c *ConnectionEvictingClient) Generate3D(ctx context.Context, in *pb.Generate3DRequest, opts ...ggrpc.CallOption) (*pb.Result, error) {
+	result, err := c.Backend.Generate3D(ctx, in, opts...)
+	c.checkErr(err)
+	return result, err
+}
+
 func (c *ConnectionEvictingClient) TTS(ctx context.Context, in *pb.TTSRequest, opts ...ggrpc.CallOption) (*pb.Result, error) {
 	result, err := c.Backend.TTS(ctx, in, opts...)
 	c.checkErr(err)
diff --git a/pkg/model/initializers.go b/pkg/model/initializers.go
index c42b7ef68..1eb1a7a39 100644
--- a/pkg/model/initializers.go
+++ b/pkg/model/initializers.go
@@ -26,6 +26,7 @@ var Aliases = map[string]string{
 	"ik_llama":               IKLLamaCPP,
 	"ik-llama":               IKLLamaCPP,
 	"embedded-store":         LocalStoreBackend,
+	"valkey":                 ValkeyStoreBackend,
 	"huggingface-embeddings": TransformersBackend,
 	"transformers-musicgen":  TransformersBackend,
 	"sentencetransformers":   TransformersBackend,
@@ -43,9 +44,11 @@ var TypeAlias = map[string]string{
 const (
 	WhisperBackend             = "whisper"
 	StableDiffusionGGMLBackend = "stablediffusion-ggml"
+	Trellis2CppBackend         = "trellis2cpp"
 
 	TransformersBackend = "transformers"
 	LocalStoreBackend   = "local-store"
+	ValkeyStoreBackend  = "valkey-store"
 )
 
 // starts the grpcModelProcess for the backend, and returns a grpc client
diff --git a/scripts/changed-backends.js b/scripts/changed-backends.js
index 1c87d0247..08a024fc6 100644
--- a/scripts/changed-backends.js
+++ b/scripts/changed-backends.js
@@ -6,6 +6,7 @@ import {
   getAllBackendPaths,
   filterMatrix,
   BACKEND_MATRIX_FILE,
+  BACKEND_PROTO_FILE,
 } from "./lib/backend-filter.mjs";
 
 // Matrix data lives in a small data-only YAML so both backend.yml (master push)
@@ -116,6 +117,42 @@ async function getPreviousMatrix(event) {
   }
 }
 
+// backend.proto at the base revision plus the checked-out copy, so filterMatrix
+// can tell an additive edit (a new field, message or RPC, which invalidates no
+// existing image) from a breaking one. Returning null means "rebuild
+// everything", the same posture as an unresolvable matrix diff.
+//
+// Only called when the changed-file list names the proto, so the common path
+// costs no extra API request.
+async function getProtoRevisions(event) {
+  const ref = event.pull_request ? event.pull_request.base.sha : event.before;
+  if (!ref || /^0+$/.test(ref)) return null;
+  const owner = event.repository.owner.login;
+  const repo = event.repository.name;
+  try {
+    const res = await octokit.request('GET /repos/{owner}/{repo}/contents/{path}', {
+      owner,
+      repo,
+      path: BACKEND_PROTO_FILE,
+      ref,
+      mediaType: { format: 'raw' },
+    });
+    const previous = typeof res.data === 'string'
+      ? res.data
+      : Buffer.from(res.data.content, 'base64').toString('utf8');
+    return {
+      previous,
+      current: fs.readFileSync(BACKEND_PROTO_FILE, "utf8"),
+    };
+  } catch (err) {
+    console.log(
+      `could not read ${BACKEND_PROTO_FILE} at ${ref}, falling back to run-all:`,
+      err.message
+    );
+    return null;
+  }
+}
+
 // Group matrix entries by tag-suffix and emit a merge-matrix entry per group.
 // Both multi-leg groups (per-arch fan-out) and singletons get one entry each:
 // the build job pushes by digest only with no tags applied, so every backend
@@ -240,7 +277,7 @@ function emitFullMatrix() {
   }
 }
 
-function emitFilteredMatrix(changedFiles, previousMatrix) {
+function emitFilteredMatrix(changedFiles, previousMatrix, protoRevisions) {
   console.log("Changed files:", changedFiles);
 
   const { filtered, filteredDarwin, changedBackends } = filterMatrix({
@@ -248,6 +285,7 @@ function emitFilteredMatrix(changedFiles, previousMatrix) {
     includesDarwin,
     changedFiles,
     previousMatrix,
+    protoRevisions,
   });
 
   console.log("Filtered files:", filtered);
@@ -306,5 +344,9 @@ function emitFilteredMatrix(changedFiles, previousMatrix) {
     ? await getPreviousMatrix(event)
     : null;
 
-  emitFilteredMatrix(changedFiles, previousMatrix);
+  const protoRevisions = changedFiles.includes(BACKEND_PROTO_FILE)
+    ? await getProtoRevisions(event)
+    : null;
+
+  emitFilteredMatrix(changedFiles, previousMatrix, protoRevisions);
 })();
diff --git a/scripts/lib/backend-filter.mjs b/scripts/lib/backend-filter.mjs
index 50edd1726..c6e4bd7c2 100644
--- a/scripts/lib/backend-filter.mjs
+++ b/scripts/lib/backend-filter.mjs
@@ -50,6 +50,12 @@ export function inferBackendPath(item) {
   if (item.backend === "magpie-tts-cpp") {
     return `backend/go/magpie-tts-cpp/`;
   }
+  // trellis2cpp is a Go backend (Dockerfile.golang) wrapping the trellis2.cpp
+  // ggml port via purego, living in backend/go/trellis2cpp/. Keep the mapping
+  // explicit so a future dockerfile-suffix change cannot break path filtering.
+  if (item.backend === "trellis2cpp") {
+    return `backend/go/trellis2cpp/`;
+  }
   if (item.dockerfile.endsWith("golang")) {
     return `backend/go/${item.backend}/`;
   }
@@ -178,6 +184,126 @@ const GO_BACKEND_PKG_PREFIXES = [
   "pkg/utils/",
 ];
 
+export const BACKEND_PROTO_FILE = "backend/backend.proto";
+const PROTO_RULE_ID = "backend-proto";
+
+// Split a .proto into a map of symbol -> fingerprint so two revisions can be
+// compared structurally instead of textually. A comment reflow, a reindent or a
+// reordered field must not read as a change, and a renumbered or retyped field
+// must.
+//
+// The scanner is deliberately syntax-light: it tracks brace depth to build a
+// container path and records one entry per declaration (`message`, `enum`,
+// `service`, `oneof`, `rpc`) and one per statement (fields, enum values,
+// `option`, `reserved`). It never needs to understand types, only to notice
+// when the text describing one stops being identical.
+function protoSymbols(text) {
+  const symbols = new Map();
+  const stack = [];
+  const norm = s => s.trim().replace(/\s+/g, " ");
+
+  // A declaration is identified by its kind and name, so that changing its body
+  // shows up as changed members rather than as a wholesale replacement.
+  const declKey = header => {
+    const rpc = header.match(/^rpc\s+([A-Za-z_]\w*)/);
+    if (rpc) return `rpc ${rpc[1]}`;
+    const decl = header.match(/^(message|enum|service|oneof|extend)\s+([A-Za-z_]\w*)/);
+    if (decl) return `${decl[1]} ${decl[2]}`;
+    return header;
+  };
+
+  // `bool cache_prompt = 8` is identified by `cache_prompt`, so renumbering or
+  // retyping it changes the fingerprint under a stable key, while renaming it
+  // reads as a removal plus an addition. Statements with no `=` (`reserved 4;`)
+  // are their own identity.
+  const stmtKey = stmt => {
+    const eq = stmt.indexOf("=");
+    if (eq === -1) return stmt;
+    const lhs = norm(stmt.slice(0, eq)).split(" ");
+    return lhs[lhs.length - 1] || stmt;
+  };
+
+  let buf = "";
+  let i = 0;
+  while (i < text.length) {
+    const c = text[i];
+
+    // String literals first: `option go_package = "github.com/..."` contains a
+    // `//` that is not a comment.
+    if (c === '"' || c === "'") {
+      buf += c;
+      i++;
+      while (i < text.length) {
+        if (text[i] === "\\") {
+          buf += text.slice(i, i + 2);
+          i += 2;
+          continue;
+        }
+        buf += text[i];
+        i++;
+        if (text[i - 1] === c) break;
+      }
+      continue;
+    }
+    if (c === "/" && text[i + 1] === "/") {
+      while (i < text.length && text[i] !== "\n") i++;
+      continue;
+    }
+    if (c === "/" && text[i + 1] === "*") {
+      i += 2;
+      while (i < text.length && !(text[i] === "*" && text[i + 1] === "/")) i++;
+      i += 2;
+      continue;
+    }
+    if (c === "{") {
+      const header = norm(buf);
+      buf = "";
+      i++;
+      const key = header ? declKey(header) : "";
+      if (header) symbols.set(`${stack.join("/")}|${key}`, header);
+      stack.push(key);
+      continue;
+    }
+    if (c === "}") {
+      stack.pop();
+      buf = "";
+      i++;
+      continue;
+    }
+    if (c === ";") {
+      const stmt = norm(buf);
+      buf = "";
+      i++;
+      if (stmt) symbols.set(`${stack.join("/")}|${stmtKey(stmt)}`, stmt);
+      continue;
+    }
+    buf += c;
+    i++;
+  }
+  return symbols;
+}
+
+// True when every symbol the previous revision declared survives unchanged into
+// the current one. New symbols are free: a backend that never references a new
+// field, message or RPC produces the same behavior with or without it, so its
+// image does not need rebuilding.
+//
+// Anything else (a removed, renumbered, retyped or renamed field, a dropped
+// RPC, a changed option) is treated as breaking and rebuilds the full matrix.
+// Either text being unresolvable is breaking too, matching the run-all posture
+// changed-backends.js takes for a diff it cannot compute.
+export function protoChangeIsAdditive(previousText, currentText) {
+  if (typeof previousText !== "string" || typeof currentText !== "string") {
+    return false;
+  }
+  const before = protoSymbols(previousText);
+  const after = protoSymbols(currentText);
+  for (const [key, fingerprint] of before) {
+    if (after.get(key) !== fingerprint) return false;
+  }
+  return true;
+}
+
 // Shared build inputs: files that end up in, or decide the contents of, images
 // belonging to backends whose own directory they do not live under. The
 // per-backend prefix match in filterMatrix() structurally cannot see these, so
@@ -200,7 +326,14 @@ export const SHARED_BUILD_INPUTS = [
     // backends regenerate their stubs from it via `make protogen-go`, the C++
     // CMakeLists compile backend.pb.cc from it, and the Rust crate Makefile
     // copies it in before build.rs runs.
-    matches: file => file === "backend/backend.proto",
+    //
+    // Which is why this rule is always/always, and why it is the single most
+    // expensive entry in the table: 417 Linux plus 56 Darwin builds. It fires
+    // on 1.3% of commits (10 of 767 over six months), and every one of those
+    // ten was purely additive. filterMatrix() suppresses this rule for an
+    // additive-only diff, so `id` exists to identify it there.
+    id: PROTO_RULE_ID,
+    matches: file => file === BACKEND_PROTO_FILE,
     linux: always,
     darwin: always,
   },
@@ -362,8 +495,19 @@ export function filterMatrix({
   includesDarwin,
   changedFiles,
   previousMatrix,
+  protoRevisions,
 }) {
-  const sharedRules = matchedSharedRules(changedFiles);
+  // An additive-only backend.proto edit invalidates no existing image, so drop
+  // its always/always rule. Every other matched rule still applies: a PR that
+  // touches the proto and scripts/build/ is still a full rebuild.
+  const protoAdditiveOnly =
+    changedFiles.includes(BACKEND_PROTO_FILE) &&
+    !!protoRevisions &&
+    protoChangeIsAdditive(protoRevisions.previous, protoRevisions.current);
+
+  const sharedRules = matchedSharedRules(changedFiles).filter(
+    rule => !(rule.id === PROTO_RULE_ID && protoAdditiveOnly)
+  );
 
   const matrixFileChanged = changedFiles.includes(BACKEND_MATRIX_FILE);
   // The matrix file changed but we could not resolve what it used to say (API
diff --git a/scripts/lib/backend-filter_test.mjs b/scripts/lib/backend-filter_test.mjs
index 0d99e1e10..5f5ebd28d 100644
--- a/scripts/lib/backend-filter_test.mjs
+++ b/scripts/lib/backend-filter_test.mjs
@@ -8,7 +8,17 @@
 import test from "node:test";
 import assert from "node:assert/strict";
 
-import { filterMatrix } from "./backend-filter.mjs";
+import { filterMatrix, inferBackendPath } from "./backend-filter.mjs";
+
+test("trellis2cpp maps to its Go backend source directory", () => {
+  assert.equal(
+    inferBackendPath({
+      backend: "trellis2cpp",
+      dockerfile: "./backend/Dockerfile.golang",
+    }),
+    "backend/go/trellis2cpp/"
+  );
+});
 
 // A representative slice of .github/backend-matrix.yml: one entry per build
 // path that the filter treats differently. Kept inline rather than parsed from
@@ -385,3 +395,146 @@ test("an unavailable previous matrix conservatively rebuilds everything", () =>
   assert.equal(filtered.length, includes.length);
   assert.equal(filteredDarwin.length, includesDarwin.length);
 });
+
+// --- backend.proto: additive changes must not rebuild the world -------------
+//
+// backend/backend.proto is consumed by every language, so the SHARED_BUILD_INPUTS
+// rule for it is always/always: a full 417-entry Linux matrix plus all 56 Darwin
+// entries. It fires on 1.3% of commits, and in practice every one of those has
+// been purely additive (a new field with an unused number, a new message, a new
+// RPC). Adding `bool cache_prompt = 8;` (PR #11158) cannot change the behavior of
+// a backend that never reads it, yet it rebuilt all 473 images.
+//
+// So the rule becomes content-aware rather than path-aware, using the same shape
+// as previousMatrix above: changed-backends.js resolves the base revision and
+// hands the two texts in, and everything here stays pure.
+
+const protoWith = body => `
+syntax = "proto3";
+
+package backend;
+
+service Backend {
+  rpc Health(HealthMessage) returns (Reply) {}
+  rpc Predict(PredictOptions) returns (Reply) {}
+}
+
+message HealthMessage {}
+
+message PredictOptions {
+${body}
+}
+`;
+
+const BASE_FIELDS = `  string Prompt = 1;
+  int32 Tokens = 2;
+  bool UseTokenizerTemplate = 3;`;
+
+const runProto = (previous, current) =>
+  filterMatrix({
+    includes,
+    includesDarwin,
+    changedFiles: ["backend/backend.proto"],
+    protoRevisions: { previous, current },
+  });
+
+test("an added proto field rebuilds nothing", () => {
+  // The real PR #11158 diff: six added lines, one new field, unused number.
+  const { filtered, filteredDarwin, changedBackends } = runProto(
+    protoWith(BASE_FIELDS),
+    protoWith(`${BASE_FIELDS}\n  bool cache_prompt = 8;`)
+  );
+
+  assert.deepEqual(filtered, []);
+  assert.deepEqual(filteredDarwin, []);
+  assert.equal(changedBackends.size, 0);
+});
+
+test("an added proto message and RPC rebuild nothing", () => {
+  const current = protoWith(BASE_FIELDS).replace(
+    "message HealthMessage {}",
+    "message HealthMessage {}\n\nmessage ScoreRequest {\n  string Text = 1;\n}"
+  ).replace(
+    "  rpc Predict(PredictOptions) returns (Reply) {}",
+    "  rpc Predict(PredictOptions) returns (Reply) {}\n  rpc Score(ScoreRequest) returns (Reply) {}"
+  );
+
+  const { filtered, filteredDarwin } = runProto(protoWith(BASE_FIELDS), current);
+
+  assert.deepEqual(filtered, []);
+  assert.deepEqual(filteredDarwin, []);
+});
+
+test("a removed proto field rebuilds every backend on every OS", () => {
+  const { filtered, filteredDarwin } = runProto(
+    protoWith(BASE_FIELDS),
+    protoWith(`  string Prompt = 1;\n  bool UseTokenizerTemplate = 3;`)
+  );
+
+  assert.equal(filtered.length, includes.length);
+  assert.equal(filteredDarwin.length, includesDarwin.length);
+});
+
+test("a renumbered proto field rebuilds every backend on every OS", () => {
+  // Wire-incompatible: an old backend reading field 2 gets nothing.
+  const { filtered, filteredDarwin } = runProto(
+    protoWith(BASE_FIELDS),
+    protoWith(`  string Prompt = 1;\n  int32 Tokens = 9;\n  bool UseTokenizerTemplate = 3;`)
+  );
+
+  assert.equal(filtered.length, includes.length);
+  assert.equal(filteredDarwin.length, includesDarwin.length);
+});
+
+test("a retyped proto field rebuilds every backend on every OS", () => {
+  const { filtered, filteredDarwin } = runProto(
+    protoWith(BASE_FIELDS),
+    protoWith(`  string Prompt = 1;\n  int64 Tokens = 2;\n  bool UseTokenizerTemplate = 3;`)
+  );
+
+  assert.equal(filtered.length, includes.length);
+  assert.equal(filteredDarwin.length, includesDarwin.length);
+});
+
+test("a renamed proto field rebuilds every backend on every OS", () => {
+  // Same number and type, but every generated accessor changes name.
+  const { filtered, filteredDarwin } = runProto(
+    protoWith(BASE_FIELDS),
+    protoWith(`  string Prompt = 1;\n  int32 MaxTokens = 2;\n  bool UseTokenizerTemplate = 3;`)
+  );
+
+  assert.equal(filtered.length, includes.length);
+  assert.equal(filteredDarwin.length, includesDarwin.length);
+});
+
+test("a removed proto RPC rebuilds every backend on every OS", () => {
+  const { filtered, filteredDarwin } = runProto(
+    protoWith(BASE_FIELDS),
+    protoWith(BASE_FIELDS).replace(
+      "  rpc Predict(PredictOptions) returns (Reply) {}\n",
+      ""
+    )
+  );
+
+  assert.equal(filtered.length, includes.length);
+  assert.equal(filteredDarwin.length, includesDarwin.length);
+});
+
+test("a comment-only proto change rebuilds nothing", () => {
+  const { filtered, filteredDarwin } = runProto(
+    protoWith(BASE_FIELDS),
+    protoWith(`  string Prompt = 1;\n  // how many tokens to emit\n  int32 Tokens = 2;\n  bool UseTokenizerTemplate = 3;`)
+  );
+
+  assert.deepEqual(filtered, []);
+  assert.deepEqual(filteredDarwin, []);
+});
+
+test("unresolvable proto revisions conservatively rebuild everything", () => {
+  // Same posture as the previousMatrix fallback: if we cannot resolve what the
+  // proto used to say, we must not claim the change was additive.
+  const { filtered, filteredDarwin } = runProto(null, protoWith(BASE_FIELDS));
+
+  assert.equal(filtered.length, includes.length);
+  assert.equal(filteredDarwin.length, includesDarwin.length);
+});
diff --git a/swagger/docs.go b/swagger/docs.go
index 83314766e..452dc6516 100644
--- a/swagger/docs.go
+++ b/swagger/docs.go
@@ -22,6 +22,77 @@ const docTemplate = `{
     "host": "{{.Host}}",
     "basePath": "{{.BasePath}}",
     "paths": {
+        "/3d/generations": {
+            "post": {
+                "tags": [
+                    "3d"
+                ],
+                "summary": "Creates a 3D asset (binary glTF / GLB) from a conditioning image.",
+                "parameters": [
+                    {
+                        "description": "query params",
+                        "name": "request",
+                        "in": "body",
+                        "required": true,
+                        "schema": {
+                            "$ref": "#/definitions/schema.Model3DRequest"
+                        }
+                    }
+                ],
+                "responses": {
+                    "200": {
+                        "description": "Response",
+                        "schema": {
+                            "$ref": "#/definitions/schema.OpenAIResponse"
+                        }
+                    }
+                }
+            }
+        },
+        "/3d/remesh": {
+            "post": {
+                "consumes": [
+                    "multipart/form-data"
+                ],
+                "produces": [
+                    "model/gltf-binary"
+                ],
+                "tags": [
+                    "3d"
+                ],
+                "summary": "Applies watertight print remeshing to an existing 3D asset.",
+                "parameters": [
+                    {
+                        "type": "string",
+                        "description": "3D model name",
+                        "name": "model",
+                        "in": "formData",
+                        "required": true
+                    },
+                    {
+                        "type": "file",
+                        "description": "Source GLB",
+                        "name": "mesh",
+                        "in": "formData",
+                        "required": true
+                    },
+                    {
+                        "type": "number",
+                        "description": "Detail size as percent of the source bounding-box diagonal (0.35–2.5; default 0.5)",
+                        "name": "detail",
+                        "in": "formData"
+                    }
+                ],
+                "responses": {
+                    "200": {
+                        "description": "Remeshed GLB",
+                        "schema": {
+                            "type": "file"
+                        }
+                    }
+                }
+            }
+        },
         "/api/agent/jobs": {
             "get": {
                 "produces": [
@@ -5662,6 +5733,54 @@ const docTemplate = `{
                 }
             }
         },
+        "schema.Model3DRequest": {
+            "description": "3D asset generation request body. Generation is image-conditioned",
+            "type": "object",
+            "properties": {
+                "background": {
+                    "description": "background handling: auto|keep|black|white",
+                    "type": "string"
+                },
+                "cfg_scale": {
+                    "description": "classifier-free guidance scale (backend default 7.5)",
+                    "type": "number"
+                },
+                "image": {
+                    "description": "conditioning image: URL, base64, or data URI (required)",
+                    "type": "string"
+                },
+                "model": {
+                    "type": "string"
+                },
+                "params": {
+                    "description": "backend-specific generation parameters",
+                    "type": "object",
+                    "additionalProperties": {
+                        "type": "string"
+                    }
+                },
+                "quality": {
+                    "description": "mesh pipeline: auto|coarse|512|1024",
+                    "type": "string"
+                },
+                "response_format": {
+                    "description": "output format (url or b64_json)",
+                    "type": "string"
+                },
+                "seed": {
+                    "description": "random seed; \u003c=0 picks a random seed",
+                    "type": "integer"
+                },
+                "step": {
+                    "description": "flow sampling steps (backend default 12)",
+                    "type": "integer"
+                },
+                "texture_steps": {
+                    "description": "texture flow sampling steps (backend default 12)",
+                    "type": "integer"
+                }
+            }
+        },
         "schema.ModelCapabilities": {
             "type": "object",
             "properties": {
diff --git a/swagger/swagger.json b/swagger/swagger.json
index db9bee118..7b452f7cd 100644
--- a/swagger/swagger.json
+++ b/swagger/swagger.json
@@ -19,6 +19,77 @@
     },
     "basePath": "/",
     "paths": {
+        "/3d/generations": {
+            "post": {
+                "tags": [
+                    "3d"
+                ],
+                "summary": "Creates a 3D asset (binary glTF / GLB) from a conditioning image.",
+                "parameters": [
+                    {
+                        "description": "query params",
+                        "name": "request",
+                        "in": "body",
+                        "required": true,
+                        "schema": {
+                            "$ref": "#/definitions/schema.Model3DRequest"
+                        }
+                    }
+                ],
+                "responses": {
+                    "200": {
+                        "description": "Response",
+                        "schema": {
+                            "$ref": "#/definitions/schema.OpenAIResponse"
+                        }
+                    }
+                }
+            }
+        },
+        "/3d/remesh": {
+            "post": {
+                "consumes": [
+                    "multipart/form-data"
+                ],
+                "produces": [
+                    "model/gltf-binary"
+                ],
+                "tags": [
+                    "3d"
+                ],
+                "summary": "Applies watertight print remeshing to an existing 3D asset.",
+                "parameters": [
+                    {
+                        "type": "string",
+                        "description": "3D model name",
+                        "name": "model",
+                        "in": "formData",
+                        "required": true
+                    },
+                    {
+                        "type": "file",
+                        "description": "Source GLB",
+                        "name": "mesh",
+                        "in": "formData",
+                        "required": true
+                    },
+                    {
+                        "type": "number",
+                        "description": "Detail size as percent of the source bounding-box diagonal (0.35–2.5; default 0.5)",
+                        "name": "detail",
+                        "in": "formData"
+                    }
+                ],
+                "responses": {
+                    "200": {
+                        "description": "Remeshed GLB",
+                        "schema": {
+                            "type": "file"
+                        }
+                    }
+                }
+            }
+        },
         "/api/agent/jobs": {
             "get": {
                 "produces": [
@@ -5659,6 +5730,54 @@
                 }
             }
         },
+        "schema.Model3DRequest": {
+            "description": "3D asset generation request body. Generation is image-conditioned",
+            "type": "object",
+            "properties": {
+                "background": {
+                    "description": "background handling: auto|keep|black|white",
+                    "type": "string"
+                },
+                "cfg_scale": {
+                    "description": "classifier-free guidance scale (backend default 7.5)",
+                    "type": "number"
+                },
+                "image": {
+                    "description": "conditioning image: URL, base64, or data URI (required)",
+                    "type": "string"
+                },
+                "model": {
+                    "type": "string"
+                },
+                "params": {
+                    "description": "backend-specific generation parameters",
+                    "type": "object",
+                    "additionalProperties": {
+                        "type": "string"
+                    }
+                },
+                "quality": {
+                    "description": "mesh pipeline: auto|coarse|512|1024",
+                    "type": "string"
+                },
+                "response_format": {
+                    "description": "output format (url or b64_json)",
+                    "type": "string"
+                },
+                "seed": {
+                    "description": "random seed; \u003c=0 picks a random seed",
+                    "type": "integer"
+                },
+                "step": {
+                    "description": "flow sampling steps (backend default 12)",
+                    "type": "integer"
+                },
+                "texture_steps": {
+                    "description": "texture flow sampling steps (backend default 12)",
+                    "type": "integer"
+                }
+            }
+        },
         "schema.ModelCapabilities": {
             "type": "object",
             "properties": {
diff --git a/swagger/swagger.yaml b/swagger/swagger.yaml
index 2b92f5233..ebbcf92ca 100644
--- a/swagger/swagger.yaml
+++ b/swagger/swagger.yaml
@@ -1444,6 +1444,41 @@ definitions:
           $ref: '#/definitions/schema.ToolCall'
         type: array
     type: object
+  schema.Model3DRequest:
+    description: 3D asset generation request body. Generation is image-conditioned
+    properties:
+      background:
+        description: 'background handling: auto|keep|black|white'
+        type: string
+      cfg_scale:
+        description: classifier-free guidance scale (backend default 7.5)
+        type: number
+      image:
+        description: 'conditioning image: URL, base64, or data URI (required)'
+        type: string
+      model:
+        type: string
+      params:
+        additionalProperties:
+          type: string
+        description: backend-specific generation parameters
+        type: object
+      quality:
+        description: 'mesh pipeline: auto|coarse|512|1024'
+        type: string
+      response_format:
+        description: output format (url or b64_json)
+        type: string
+      seed:
+        description: random seed; <=0 picks a random seed
+        type: integer
+      step:
+        description: flow sampling steps (backend default 12)
+        type: integer
+      texture_steps:
+        description: texture flow sampling steps (backend default 12)
+        type: integer
+    type: object
   schema.ModelCapabilities:
     properties:
       capabilities:
@@ -2722,6 +2757,53 @@ info:
   title: LocalAI API
   version: 2.0.0
 paths:
+  /3d/generations:
+    post:
+      parameters:
+      - description: query params
+        in: body
+        name: request
+        required: true
+        schema:
+          $ref: '#/definitions/schema.Model3DRequest'
+      responses:
+        "200":
+          description: Response
+          schema:
+            $ref: '#/definitions/schema.OpenAIResponse'
+      summary: Creates a 3D asset (binary glTF / GLB) from a conditioning image.
+      tags:
+      - 3d
+  /3d/remesh:
+    post:
+      consumes:
+      - multipart/form-data
+      parameters:
+      - description: 3D model name
+        in: formData
+        name: model
+        required: true
+        type: string
+      - description: Source GLB
+        in: formData
+        name: mesh
+        required: true
+        type: file
+      - description: Detail size as percent of the source bounding-box diagonal (0.35–2.5;
+          default 0.5)
+        in: formData
+        name: detail
+        type: number
+      produces:
+      - model/gltf-binary
+      responses:
+        "200":
+          description: Remeshed GLB
+          schema:
+            type: file
+      summary: Applies watertight print remeshing to an existing 3D asset.
+      tags:
+      - 3d
   /api/agent/jobs:
     get:
       parameters:
diff --git a/tests/integration/valkey_stores_test.go b/tests/integration/valkey_stores_test.go
new file mode 100644
index 000000000..13b6ed379
--- /dev/null
+++ b/tests/integration/valkey_stores_test.go
@@ -0,0 +1,413 @@
+package integration_test
+
+// Integration tests for the valkey-store gRPC backend. They mirror the
+// local-store specs one-for-one and add the two capabilities Valkey provides
+// that local-store cannot: persistence across a backend restart and an
+// identifiable client name.
+//
+// These require a running Valkey Search server (valkey/valkey-bundle:9.1.0,
+// which ships the FT.* module) reachable at $VALKEY_ADDR. When VALKEY_ADDR is
+// unset the whole suite is skipped, so the unit CI never needs the container:
+//
+//	podman run -d --name valkey-store-it -p 6379:6379 valkey/valkey-bundle:9.1.0
+//	make backends/valkey-store
+//	VALKEY_ADDR=localhost:6379 make test-valkey-store
+//
+// Index back-fill is asynchronous, so every Find is preceded by a bounded poll
+// (eventuallyFindable) on the search result rather than a fixed sleep.
+
+import (
+	"context"
+	"fmt"
+	"math"
+	"math/rand/v2"
+	"os"
+	"strings"
+	"sync/atomic"
+	"time"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+	valkey "github.com/valkey-io/valkey-go"
+
+	"github.com/mudler/LocalAI/core/gallery"
+	"github.com/mudler/LocalAI/pkg/grpc"
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+	"github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/store"
+	"github.com/mudler/LocalAI/pkg/system"
+)
+
+// namespaceCounter gives each spec a fresh namespace so the persistent Valkey
+// server does not leak state between tests sharing one instance.
+var namespaceCounter atomic.Int64
+
+func valkeyNormalize(vecs [][]float32) {
+	for i, k := range vecs {
+		norm := float64(0)
+		for _, x := range k {
+			norm += float64(x * x)
+		}
+		norm = math.Sqrt(norm)
+		for j, x := range k {
+			vecs[i][j] = x / float32(norm)
+		}
+	}
+}
+
+// eventuallyFindable polls Find until at least want results come back, absorbing
+// the asynchronous index back-fill without a hard-coded sleep.
+func eventuallyFindable(sc grpc.Backend, query []float32, want int) {
+	EventuallyWithOffset(1, func() int {
+		keys, _, _, err := store.Find(context.Background(), sc, query, want)
+		if err != nil {
+			return -1
+		}
+		return len(keys)
+	}, 15*time.Second, 200*time.Millisecond).Should(BeNumerically(">=", want))
+}
+
+var _ = Describe("Integration tests for the valkey-store backend", Label("stores"), Label("valkey"), func() {
+	Context("Valkey Search get, set, delete and find", func() {
+		var sl *model.ModelLoader
+		var sc grpc.Backend
+		var tmpdir string
+		var namespace string
+		var valkeyAddr string
+		var backendsPath string
+
+		loadStore := func(ns string) grpc.Backend {
+			storeOpts := []model.Option{
+				model.WithBackendString(model.ValkeyStoreBackend),
+				model.WithModel(ns),
+				// Configure the store the LocalAI-native way: the connection
+				// address is threaded to the backend's LoadModel via the model
+				// config options (ModelOptions.Options), exactly as core's
+				// StoreBackend() passes a store's `options:` list. The backend
+				// no longer reads any VALKEY_* env var. $VALKEY_ADDR here is
+				// only the integration harness locating the test server.
+				model.WithLoadGRPCLoadModelOpts(&pb.ModelOptions{
+					Options: []string{"addr:" + valkeyAddr},
+				}),
+			}
+			backend, err := sl.Load(storeOpts...)
+			Expect(err).ToNot(HaveOccurred())
+			Expect(backend).ToNot(BeNil())
+			return backend
+		}
+
+		// initLoader builds a fresh model loader that can discover the
+		// valkey-store gRPC binary. The backend is registered from BACKENDS_PATH
+		// exactly as the real application does at startup
+		// (core/application/startup.go), so sl.Load(WithBackendString(...)) can
+		// resolve the run.sh and spawn the process.
+		initLoader := func() {
+			systemState, err := system.GetSystemState(
+				system.WithModelPath(tmpdir),
+				system.WithBackendPath(backendsPath),
+			)
+			Expect(err).ToNot(HaveOccurred())
+
+			sl = model.NewModelLoader(systemState)
+			Expect(gallery.RegisterBackends(systemState, sl)).To(Succeed())
+		}
+
+		BeforeEach(func() {
+			valkeyAddr = os.Getenv("VALKEY_ADDR")
+			if valkeyAddr == "" {
+				Skip("VALKEY_ADDR is not set; skipping Valkey integration tests")
+			}
+			backendsPath = os.Getenv("BACKENDS_PATH")
+			if backendsPath == "" {
+				Skip("BACKENDS_PATH is not set; build the backend and point BACKENDS_PATH at it (see make test-valkey-store)")
+			}
+
+			var err error
+			tmpdir, err = os.MkdirTemp("", "")
+			Expect(err).ToNot(HaveOccurred())
+
+			namespace = fmt.Sprintf("it-%d-%d", GinkgoRandomSeed(), namespaceCounter.Add(1))
+
+			initLoader()
+			sc = loadStore(namespace)
+		})
+
+		AfterEach(func() {
+			if sl != nil {
+				err := sl.StopAllGRPC()
+				Expect(err).ToNot(HaveOccurred())
+			}
+			if tmpdir != "" {
+				_ = os.RemoveAll(tmpdir)
+			}
+		})
+
+		It("should be able to set a key", func() {
+			err := store.SetSingle(context.Background(), sc, []float32{0.1, 0.2, 0.3}, []byte("test"))
+			Expect(err).ToNot(HaveOccurred())
+		})
+
+		It("should be able to set keys", func() {
+			err := store.SetCols(context.Background(), sc, [][]float32{{0.1, 0.2, 0.3}, {0.4, 0.5, 0.6}}, [][]byte{[]byte("test1"), []byte("test2")})
+			Expect(err).ToNot(HaveOccurred())
+			err = store.SetCols(context.Background(), sc, [][]float32{{0.7, 0.8, 0.9}, {0.10, 0.11, 0.12}}, [][]byte{[]byte("test3"), []byte("test4")})
+			Expect(err).ToNot(HaveOccurred())
+		})
+
+		It("should be able to get a key", func() {
+			err := store.SetSingle(context.Background(), sc, []float32{0.1, 0.2, 0.3}, []byte("test"))
+			Expect(err).ToNot(HaveOccurred())
+
+			val, err := store.GetSingle(context.Background(), sc, []float32{0.1, 0.2, 0.3})
+			Expect(err).ToNot(HaveOccurred())
+			Expect(val).To(Equal([]byte("test")))
+		})
+
+		It("should be able to get keys", func() {
+			err := store.SetCols(context.Background(), sc, [][]float32{{0.1, 0.2, 0.3}, {0.4, 0.5, 0.6}, {0.7, 0.8, 0.9}}, [][]byte{[]byte("test1"), []byte("test2"), []byte("test3")})
+			Expect(err).ToNot(HaveOccurred())
+
+			keys, vals, err := store.GetCols(context.Background(), sc, [][]float32{{0.1, 0.2, 0.3}, {0.4, 0.5, 0.6}, {0.7, 0.8, 0.9}})
+			Expect(err).ToNot(HaveOccurred())
+			Expect(keys).To(HaveLen(3))
+			Expect(vals).To(HaveLen(3))
+			for i, k := range keys {
+				v := vals[i]
+				switch {
+				case k[0] == 0.1 && k[1] == 0.2 && k[2] == 0.3:
+					Expect(v).To(Equal([]byte("test1")))
+				case k[0] == 0.4 && k[1] == 0.5 && k[2] == 0.6:
+					Expect(v).To(Equal([]byte("test2")))
+				default:
+					Expect(k).To(Equal([]float32{0.7, 0.8, 0.9}))
+					Expect(v).To(Equal([]byte("test3")))
+				}
+			}
+
+			keys, vals, err = store.GetCols(context.Background(), sc, [][]float32{{0.7, 0.8, 0.9}, {0.1, 0.2, 0.3}})
+			Expect(err).ToNot(HaveOccurred())
+			Expect(keys).To(HaveLen(2))
+			Expect(vals).To(HaveLen(2))
+		})
+
+		It("should be able to delete a key", func() {
+			err := store.SetSingle(context.Background(), sc, []float32{0.1, 0.2, 0.3}, []byte("test"))
+			Expect(err).ToNot(HaveOccurred())
+
+			err = store.DeleteSingle(context.Background(), sc, []float32{0.1, 0.2, 0.3})
+			Expect(err).ToNot(HaveOccurred())
+
+			val, _ := store.GetSingle(context.Background(), sc, []float32{0.1, 0.2, 0.3})
+			Expect(val).To(BeNil())
+		})
+
+		It("should be able to delete keys", func() {
+			err := store.SetCols(context.Background(), sc, [][]float32{{0.1, 0.2, 0.3}, {0.4, 0.5, 0.6}, {0.7, 0.8, 0.9}}, [][]byte{[]byte("test1"), []byte("test2"), []byte("test3")})
+			Expect(err).ToNot(HaveOccurred())
+
+			err = store.DeleteCols(context.Background(), sc, [][]float32{{0.1, 0.2, 0.3}, {0.7, 0.8, 0.9}})
+			Expect(err).ToNot(HaveOccurred())
+
+			keys, vals, err := store.GetCols(context.Background(), sc, [][]float32{{0.4, 0.5, 0.6}})
+			Expect(err).ToNot(HaveOccurred())
+			Expect(keys).To(HaveLen(1))
+			Expect(vals).To(HaveLen(1))
+			Expect(keys[0]).To(Equal([]float32{0.4, 0.5, 0.6}))
+			Expect(vals[0]).To(Equal([]byte("test2")))
+
+			keys, vals, err = store.GetCols(context.Background(), sc, [][]float32{{0.1, 0.2, 0.3}, {0.7, 0.8, 0.9}})
+			Expect(err).ToNot(HaveOccurred())
+			Expect(keys).To(HaveLen(0))
+			Expect(vals).To(HaveLen(0))
+		})
+
+		It("should be able to find similar keys", func() {
+			err := store.SetCols(context.Background(), sc, [][]float32{{0.5, 0.5, 0.5}, {0.6, 0.6, -0.6}, {0.7, -0.7, -0.7}}, [][]byte{[]byte("test1"), []byte("test2"), []byte("test3")})
+			Expect(err).ToNot(HaveOccurred())
+
+			eventuallyFindable(sc, []float32{0.1, 0.3, 0.5}, 2)
+
+			keys, vals, sims, err := store.Find(context.Background(), sc, []float32{0.1, 0.3, 0.5}, 2)
+			Expect(err).ToNot(HaveOccurred())
+			Expect(keys).To(HaveLen(2))
+			Expect(vals).To(HaveLen(2))
+			Expect(sims).To(HaveLen(2))
+			Expect(keys[0]).To(Equal([]float32{0.5, 0.5, 0.5}))
+			Expect(vals[0]).To(Equal([]byte("test1")))
+			Expect(keys[1]).To(Equal([]float32{0.6, 0.6, -0.6}))
+		})
+
+		It("should be able to find similar normalized keys", func() {
+			keys := [][]float32{{0.1, 0.3, 0.5}, {0.5, 0.5, 0.5}, {0.6, 0.6, -0.6}, {0.7, -0.7, -0.7}}
+			vals := [][]byte{[]byte("test0"), []byte("test1"), []byte("test2"), []byte("test3")}
+			valkeyNormalize(keys)
+
+			err := store.SetCols(context.Background(), sc, keys, vals)
+			Expect(err).ToNot(HaveOccurred())
+
+			eventuallyFindable(sc, keys[0], 3)
+
+			ks, _, sims, err := store.Find(context.Background(), sc, keys[0], 3)
+			Expect(err).ToNot(HaveOccurred())
+			Expect(ks).To(HaveLen(3))
+			Expect(sims).To(HaveLen(3))
+			Expect(ks[0]).To(Equal(keys[0]))
+			Expect(sims[0]).To(BeNumerically("~", 1, 0.0001))
+		})
+
+		It("produces the correct cosine similarities for orthogonal and opposite unit vectors", func() {
+			keys := [][]float32{{1.0, 0.0, 0.0}, {0.0, 1.0, 0.0}, {0.0, 0.0, 1.0}, {-1.0, 0.0, 0.0}}
+			vals := [][]byte{[]byte("x"), []byte("y"), []byte("z"), []byte("-z")}
+
+			err := store.SetCols(context.Background(), sc, keys, vals)
+			Expect(err).ToNot(HaveOccurred())
+
+			eventuallyFindable(sc, keys[0], 4)
+
+			_, _, sims, err := store.Find(context.Background(), sc, keys[0], 4)
+			Expect(err).ToNot(HaveOccurred())
+			Expect(sims).To(HaveLen(4))
+			Expect(sims[0]).To(BeNumerically("~", 1, 0.0001))
+			Expect(sims[1]).To(BeNumerically("~", 0, 0.0001))
+			Expect(sims[2]).To(BeNumerically("~", 0, 0.0001))
+			Expect(sims[3]).To(BeNumerically("~", -1, 0.0001))
+		})
+
+		It("produces the correct cosine similarities for orthogonal and opposite vectors", func() {
+			keys := [][]float32{{1.0, 0.0, 1.0}, {0.0, 2.0, 0.0}, {0.0, 0.0, -1.0}, {-1.0, 0.0, -1.0}}
+			vals := [][]byte{[]byte("x"), []byte("y"), []byte("z"), []byte("-z")}
+
+			err := store.SetCols(context.Background(), sc, keys, vals)
+			Expect(err).ToNot(HaveOccurred())
+
+			eventuallyFindable(sc, keys[0], 4)
+
+			_, _, sims, err := store.Find(context.Background(), sc, keys[0], 4)
+			Expect(err).ToNot(HaveOccurred())
+			Expect(sims[0]).To(BeNumerically("~", 1, 0.1))
+			Expect(sims[1]).To(BeNumerically("~", 0, 0.1))
+			Expect(sims[2]).To(BeNumerically("~", -0.7, 0.1))
+			Expect(sims[3]).To(BeNumerically("~", -1, 0.1))
+		})
+
+		expectTriangleEq := func(keys [][]float32, vals [][]byte) {
+			eventuallyFindable(sc, keys[0], len(keys))
+			sims := map[string]map[string]float32{}
+			for i, k := range keys {
+				_, valsk, simsk, err := store.Find(context.Background(), sc, k, len(keys))
+				Expect(err).ToNot(HaveOccurred())
+				for j, v := range valsk {
+					p := string(vals[i])
+					q := string(v)
+					if sims[p] == nil {
+						sims[p] = map[string]float32{}
+					}
+					sims[p][q] = simsk[j]
+				}
+			}
+			for _, simsu := range sims {
+				for w, simw := range simsu {
+					uws := math.Acos(clampUnit(float64(simw)))
+					for v := range simsu {
+						uvws := math.Acos(clampUnit(float64(simsu[v]))) + math.Acos(clampUnit(float64(sims[v][w])))
+						Expect(uws).To(BeNumerically("<=", uvws+0.0001))
+					}
+				}
+			}
+		}
+
+		It("obeys the triangle inequality for normalized values", func() {
+			keys := [][]float32{
+				{1.0, 0.0, 0.0}, {0.0, 1.0, 0.0}, {0.0, 0.0, 1.0},
+				{-1.0, 0.0, 0.0}, {0.0, -1.0, 0.0}, {0.0, 0.0, -1.0},
+				{2.0, 3.0, 4.0}, {9.0, 7.0, 1.0}, {0.0, -1.2, 2.3},
+			}
+			vals := [][]byte{
+				[]byte("x"), []byte("y"), []byte("z"),
+				[]byte("-x"), []byte("-y"), []byte("-z"),
+				[]byte("u"), []byte("v"), []byte("w"),
+			}
+			valkeyNormalize(keys[6:])
+
+			err := store.SetCols(context.Background(), sc, keys, vals)
+			Expect(err).ToNot(HaveOccurred())
+			expectTriangleEq(keys, vals)
+		})
+
+		It("obeys the triangle inequality for random 768-d vectors", func() {
+			rnd := rand.New(rand.NewPCG(151, 0))
+			keys := make([][]float32, 20)
+			vals := make([][]byte, 20)
+			for i := range keys {
+				k := make([]float32, 768)
+				for j := range k {
+					k[j] = rnd.Float32()
+				}
+				keys[i] = k
+			}
+			c := byte('a')
+			for i := range vals {
+				vals[i] = []byte{c}
+				c++
+			}
+
+			err := store.SetCols(context.Background(), sc, keys, vals)
+			Expect(err).ToNot(HaveOccurred())
+			expectTriangleEq(keys, vals)
+		})
+
+		It("persists data across a backend restart", func() {
+			// This is the capability local-store lacks: after the backend
+			// process is torn down and a fresh one is spawned (same Valkey
+			// server, same namespace), the data is still there.
+			err := store.SetCols(context.Background(), sc,
+				[][]float32{{0.1, 0.2, 0.3}, {0.4, 0.5, 0.6}},
+				[][]byte{[]byte("persisted1"), []byte("persisted2")})
+			Expect(err).ToNot(HaveOccurred())
+
+			Expect(sl.StopAllGRPC()).ToNot(HaveOccurred())
+
+			// Fresh loader/process, same namespace → same index/prefix.
+			initLoader()
+			sc = loadStore(namespace)
+
+			val, err := store.GetSingle(context.Background(), sc, []float32{0.1, 0.2, 0.3})
+			Expect(err).ToNot(HaveOccurred())
+			Expect(val).To(Equal([]byte("persisted1")))
+
+			eventuallyFindable(sc, []float32{0.1, 0.2, 0.3}, 2)
+			keys, _, _, err := store.Find(context.Background(), sc, []float32{0.1, 0.2, 0.3}, 2)
+			Expect(err).ToNot(HaveOccurred())
+			Expect(keys).To(HaveLen(2))
+		})
+
+		It("identifies its connection with the mandatory client name", func() {
+			// Seed a key so the backend has certainly opened a connection.
+			err := store.SetSingle(context.Background(), sc, []float32{0.1, 0.2, 0.3}, []byte("test"))
+			Expect(err).ToNot(HaveOccurred())
+
+			client, err := valkey.NewClient(valkey.ClientOption{
+				InitAddress:  []string{valkeyAddr},
+				DisableCache: true,
+			})
+			Expect(err).ToNot(HaveOccurred())
+			defer client.Close()
+
+			list, err := client.Do(context.Background(), client.B().ClientList().Build()).ToString()
+			Expect(err).ToNot(HaveOccurred())
+			Expect(strings.Contains(list, "name=localai-valkey-store")).To(BeTrue(), "expected a connection named localai-valkey-store in CLIENT LIST")
+		})
+	})
+})
+
+// clampUnit keeps a similarity inside [-1, 1] before math.Acos, guarding against
+// float rounding that would otherwise push |x| slightly over 1 and yield NaN.
+func clampUnit(x float64) float64 {
+	if x > 1 {
+		return 1
+	}
+	if x < -1 {
+		return -1
+	}
+	return x
+}