fix(tests): remove duplicate scheduling stubs

Assisted-by: Codex:gpt-5 [golangci-lint]

.agents/ai-coding-assistants.md

@@ -35,19 +35,33 @@ All contributions must comply with LocalAI's licensing requirements:
 
 ## Signed-off-by and Developer Certificate of Origin
 
-**AI agents MUST NOT add `Signed-off-by` tags.** Only humans can legally
-certify the Developer Certificate of Origin (DCO). The human submitter
-is responsible for:
+Only humans can certify the Developer Certificate of Origin (DCO). AI
+agents MUST NOT invent or guess a human identity for `Signed-off-by` —
+doing so forges the DCO certification.
 
-- Reviewing all AI-generated code
+However, when a human operator explicitly directs the AI to commit on
+their behalf, the AI is acting as a typing tool — no different from an
+editor macro or `git commit -s`. In that case the AI SHOULD add
+`Signed-off-by:` using the **configured `user.name` / `user.email`** of
+the current git repository (i.e. the operator's own identity). The
+resulting trailer is the operator's signature; they take responsibility
+for it by reviewing and pushing the commit. The AI MUST NOT use any
+other identity and MUST NOT add its own name to the sign-off.
+
+When running `git commit`, prefer `git commit --signoff` (or `-s`) so
+the trailer is emitted by git itself from the configured identity,
+rather than hand-writing it in a heredoc — this guarantees the sign-off
+matches whatever identity the operator is currently using.
+
+The human submitter remains responsible for:
+
+- Reviewing all AI-generated code before it's pushed or merged
 - Ensuring compliance with licensing requirements
-- Adding their own `Signed-off-by` tag (when the project requires DCO)
-  to certify the contribution
 - Taking full responsibility for the contribution
 
-AI agents MUST NOT add `Co-Authored-By` trailers for themselves either.
-A human reviewer owns the contribution; the AI's involvement is recorded
-via `Assisted-by` (see below).
+AI agents MUST NOT add `Co-Authored-By` trailers for themselves. A human
+reviewer owns the contribution; the AI's involvement is recorded via
+`Assisted-by` (see below).
 
 ## Attribution
 
@@ -84,6 +98,12 @@ Assisted-by: Claude:claude-opus-4-7 golangci-lint
 Signed-off-by: Jane Developer <jane@example.com>
 ```
 
+The `Signed-off-by` line uses Jane's own identity because Jane is the
+submitter operating the AI. If Jane asks Claude to create the commit via
+`git commit -s`, git emits that exact trailer from Jane's configured
+identity — no separate human step is needed beyond Jane reviewing the
+diff before pushing.
+
 ## Scope and Responsibility
 
 Using an AI assistant does not reduce the contributor's responsibility.

.docker/llama-cpp-build-target.sh

@@ -4,6 +4,17 @@ set -euo pipefail
 arch=${1:?target architecture is required}
 build_type=${2-}
 
+# SYCL compiles the whole tree with icpx -fsycl, and icpx never finishes
+# ggml-cpu/arch/x86/repack.cpp at -march=sapphirerapids: the job sits on that one
+# translation unit until GitHub kills it at 6h. gcc builds the same file in
+# seconds, so only the SYCL images have to give up the CPU variant matrix.
+case "$build_type" in
+  sycl*)
+    echo llama-cpp-fallback
+    exit 0
+    ;;
+esac
+
 # GPU arm64 base images do not consistently provide the gcc-14 toolchain needed
 # to compile ggml's armv9.2 CPU variants. Keep their portable fallback until the
 # builder images can supply that compiler.

.docker/turboquant-build-target.sh

@@ -4,6 +4,17 @@ set -euo pipefail
 arch=${1:?target architecture is required}
 build_type=${2-}
 
+# SYCL compiles the whole tree with icpx -fsycl, and icpx never finishes
+# ggml-cpu/arch/x86/repack.cpp at -march=sapphirerapids: the job sits on that one
+# translation unit until GitHub kills it at 6h. gcc builds the same file in
+# seconds, so only the SYCL images have to give up the CPU variant matrix.
+case "$build_type" in
+  sycl*)
+    echo turboquant-fallback
+    exit 0
+    ;;
+esac
+
 # GPU arm64 base images do not consistently provide the gcc-14 toolchain needed
 # to compile ggml's armv9.2 CPU variants. Keep their portable fallback until the
 # builder images can supply that compiler.

.github/backend-matrix.yml

@@ -480,6 +480,22 @@ include:
     dockerfile: "./backend/Dockerfile.turboquant"
     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-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-cuda-12-amd64'
+    # bigger-runner: same rationale as -gpu-nvidia-cuda-12-llama-cpp above
+    # (observed 6h5m wall-clock on v4.2.1, just past the 6h job timeout).
+    runs-on: 'bigger-runner'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'cublas'
     cuda-major-version: "12"
     cuda-minor-version: "8"
@@ -1165,6 +1181,21 @@ include:
     dockerfile: "./backend/Dockerfile.turboquant"
     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-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-cuda-13-amd64'
+    # bigger-runner: observed 6h5m wall-clock on v4.2.1 — at the GHA timeout.
+    runs-on: 'bigger-runner'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'cublas'
     cuda-major-version: "13"
     cuda-minor-version: "0"
@@ -1208,6 +1239,20 @@ include:
     backend: "turboquant"
     dockerfile: "./backend/Dockerfile.turboquant"
     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-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-cuda-13-arm64'
+    base-image: "ubuntu:24.04"
+    runs-on: 'ubuntu-24.04-arm'
+    ubuntu-version: '2404'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
   - build-type: 'cublas'
     cuda-major-version: "13"
     cuda-minor-version: "0"
@@ -2477,6 +2522,20 @@ include:
     dockerfile: "./backend/Dockerfile.turboquant"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: 'sycl_f32'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-intel-sycl-f32-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-intel-amd64'
+    runs-on: 'ubuntu-latest'
+    base-image: "intel/oneapi-basekit:2025.3.0-0-devel-ubuntu24.04"
+    skip-drivers: 'false'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'sycl_f32'
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -2519,6 +2578,20 @@ include:
     dockerfile: "./backend/Dockerfile.turboquant"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: 'sycl_f16'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-intel-sycl-f16-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-intel-amd64'
+    runs-on: 'ubuntu-latest'
+    base-image: "intel/oneapi-basekit:2025.3.0-0-devel-ubuntu24.04"
+    skip-drivers: 'false'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'sycl_f16'
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -2985,6 +3058,21 @@ include:
     dockerfile: "./backend/Dockerfile.turboquant"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    platform-tag: 'amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-amd64'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -3015,6 +3103,21 @@ include:
     dockerfile: "./backend/Dockerfile.turboquant"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/arm64'
+    platform-tag: 'arm64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-arm64'
+    runs-on: 'ubuntu-24.04-arm'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -3276,6 +3379,20 @@ include:
     dockerfile: "./backend/Dockerfile.turboquant"
     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-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-l4t-cuda-12-arm64'
+    base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
+    runs-on: 'ubuntu-24.04-arm'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2204'
   - build-type: 'cublas'
     cuda-major-version: "12"
     cuda-minor-version: "0"
@@ -3336,6 +3453,22 @@ include:
     context: "./"
     ubuntu-version: '2404'
   # Stablediffusion-ggml
+  - build-type: 'vulkan'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    platform-tag: 'amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-vulkan-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-vulkan-amd64'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2404'
+  # Stablediffusion-ggml
   - build-type: 'vulkan'
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -3368,6 +3501,22 @@ include:
     context: "./"
     ubuntu-version: '2404'
   # Stablediffusion-ggml
+  - build-type: 'vulkan'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/arm64'
+    platform-tag: 'arm64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-vulkan-buun-llama-cpp'
+    builder-base-image: 'quay.io/go-skynet/ci-cache:base-grpc-vulkan-arm64'
+    runs-on: 'ubuntu-24.04-arm'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "buun-llama-cpp"
+    dockerfile: "./backend/Dockerfile.buun-llama-cpp"
+    context: "./"
+    ubuntu-version: '2404'
+  # Stablediffusion-ggml
   - build-type: 'vulkan'
     cuda-major-version: ""
     cuda-minor-version: ""

.github/workflows/gh-pages.yml

@@ -51,7 +51,16 @@ jobs:
       - name: Setup Go
         uses: actions/setup-go@v5
         with:
-          go-version: '1.22'
+          # Track go.mod rather than a literal. Pinned at 1.22 this installed a
+          # toolchain older than the module's `go 1.26.0`, so the `go run` below
+          # downloaded the real one from proxy.golang.org on every run. That
+          # fetch is not always reachable from the runner and the deploy failed
+          # on five of eight consecutive master pushes with:
+          #   go: download go1.26.0: ... connect: network is unreachable
+          #   ##[error]Command failed: go env GOPATH
+          # Installing the version the module asks for removes the download
+          # instead of depending on it succeeding.
+          go-version-file: go.mod
           cache: false
 
       - name: Setup Hugo

.github/workflows/test-extra.yml

@@ -33,6 +33,7 @@ jobs:
       llama-cpp: ${{ steps.detect.outputs.llama-cpp }}
       ik-llama-cpp: ${{ steps.detect.outputs.ik-llama-cpp }}
       turboquant: ${{ steps.detect.outputs.turboquant }}
+      buun-llama-cpp: ${{ steps.detect.outputs['buun-llama-cpp'] }}
       vllm: ${{ steps.detect.outputs.vllm }}
       sglang: ${{ steps.detect.outputs.sglang }}
       acestep-cpp: ${{ steps.detect.outputs.acestep-cpp }}
@@ -716,6 +717,30 @@ jobs:
       - name: Build turboquant backend image and run gRPC e2e tests
         run: |
           make test-extra-backend-turboquant
+  tests-buun-llama-cpp-grpc:
+    needs: detect-changes
+    if: needs.detect-changes.outputs['buun-llama-cpp'] == 'true' || needs.detect-changes.outputs.run-all == 'true'
+    runs-on: ubuntu-latest
+    timeout-minutes: 90
+    steps:
+      - name: Clone
+        uses: actions/checkout@v6
+        with:
+          submodules: true
+      - name: Setup Go
+        uses: actions/setup-go@v5
+        with:
+          go-version: '1.25.4'
+      # Exercises the buun-llama-cpp (fork-of-a-fork) backend with the
+      # fork-specific TurboQuant/TCQ KV-cache types. BACKEND_TEST_CACHE_TYPE_V
+      # is set to turbo3 so the test round-trips through the fork's KV
+      # allow-list — picking a stock llama.cpp type would only re-test the
+      # shared code path. DFlash speculative decoding is not exercised here
+      # because the one known public target/drafter pair (Qwen3.5-27B) is too
+      # large for CI.
+      - name: Build buun-llama-cpp backend image and run gRPC e2e tests
+        run: |
+          make test-extra-backend-buun-llama-cpp
   # tests-vllm-grpc is currently disabled in CI.
   #
   # The prebuilt vllm CPU wheel is compiled with AVX-512 VNNI/BF16

Makefile

@@ -1,4 +1,5 @@
 # Disable parallel execution for backend builds
+.NOTPARALLEL: backends/buun-llama-cpp
 .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 backends/audio-cpp backends/audio-cpp-darwin
 
 GOCMD=go
@@ -748,6 +749,19 @@ test-extra-backend-bonsai: docker-build-bonsai
 	BACKEND_TEST_MODEL_URL=https://huggingface.co/prism-ml/Bonsai-8B-gguf/resolve/main/Bonsai-8B-Q1_0.gguf \
 	$(MAKE) test-extra-backend
 
+## buun-llama-cpp: exercises the fork-of-a-fork backend (spiritbuun/buun-llama-cpp)
+## with the *TurboQuant/TCQ-specific* KV-cache types (turbo3 for V). Same rationale
+## as turboquant above: picking a standard llama.cpp type would only re-test the
+## shared code path. buun inherits turboquant's turbo2/turbo3/turbo4 and adds
+## turbo2_tcq / turbo3_tcq on top. DFlash speculative decoding is not exercised
+## here because no small DFlash drafter model exists (the known public pair is
+## Qwen3.5-27B, ~54 GB).
+test-extra-backend-buun-llama-cpp: docker-build-buun-llama-cpp
+	BACKEND_IMAGE=local-ai-backend:buun-llama-cpp \
+	BACKEND_TEST_CACHE_TYPE_K=q8_0 \
+	BACKEND_TEST_CACHE_TYPE_V=turbo3 \
+	$(MAKE) test-extra-backend
+
 ## Audio transcription wrapper for the llama-cpp backend.
 ## Drives the new AudioTranscription / AudioTranscriptionStream RPCs against
 ## ggml-org/Qwen3-ASR-0.6B-GGUF (a small ASR model that requires its mmproj
@@ -1284,6 +1298,11 @@ BACKEND_PRIVACY_FILTER = privacy-filter|privacy-filter|.|false|false
 # against apt gRPC/protobuf rather than a prebuilt base-grpc image; the reason
 # is on the audio-cpp block in .github/backend-matrix.yml.
 BACKEND_AUDIO_CPP = audio-cpp|audio-cpp|.|false|false
+# buun-llama-cpp is a fork-of-a-fork (spiritbuun/buun-llama-cpp forks
+# TheTom/llama-cpp-turboquant) that adds DFlash block-diffusion speculative
+# decoding and extra TCQ KV-cache variants on top of TurboQuant. Same thin
+# wrapper pattern as turboquant — reuses backend/cpp/llama-cpp grpc-server.
+BACKEND_BUUN_LLAMA_CPP = buun-llama-cpp|buun-llama-cpp|.|false|false
 
 # Golang backends
 BACKEND_PIPER = piper|golang|.|false|true
@@ -1388,6 +1407,7 @@ $(eval $(call generate-docker-build-target,$(BACKEND_BONSAI)))
 $(eval $(call generate-docker-build-target,$(BACKEND_DS4)))
 $(eval $(call generate-docker-build-target,$(BACKEND_PRIVACY_FILTER)))
 $(eval $(call generate-docker-build-target,$(BACKEND_AUDIO_CPP)))
+$(eval $(call generate-docker-build-target,$(BACKEND_BUUN_LLAMA_CPP)))
 $(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)))
@@ -1456,7 +1476,7 @@ $(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-trellis2cpp docker-build-valkey-store docker-build-audio-cpp
+docker-build-backends: docker-build-llama-cpp docker-build-ik-llama-cpp docker-build-turboquant docker-build-buun-llama-cpp 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 docker-build-audio-cpp
 
 ########################################################
 ### Mock Backend for E2E Tests

backend/Dockerfile.buun-llama-cpp

@@ -0,0 +1,290 @@
+ARG BASE_IMAGE=ubuntu:24.04
+ARG GRPC_BASE_IMAGE=${BASE_IMAGE}
+
+
+# The grpc target does one thing, it builds and installs GRPC.  This is in it's own layer so that it can be effectively cached by CI.
+# You probably don't need to change anything here, and if you do, make sure that CI is adjusted so that the cache continues to work.
+FROM ${GRPC_BASE_IMAGE} AS grpc
+
+# This is a bit of a hack, but it's required in order to be able to effectively cache this layer in CI
+ARG GRPC_MAKEFLAGS="-j4 -Otarget"
+ARG GRPC_VERSION=v1.65.0
+ARG CMAKE_FROM_SOURCE=false
+# CUDA Toolkit 13.x compatibility: CMake 3.31.9+ fixes toolchain detection/arch table issues
+ARG CMAKE_VERSION=3.31.10
+
+ENV MAKEFLAGS=${GRPC_MAKEFLAGS}
+
+WORKDIR /build
+
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        ca-certificates \
+        build-essential curl libssl-dev \
+        git wget && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# Install CMake (the version in 22.04 is too old)
+RUN <<EOT bash
+    if [ "${CMAKE_FROM_SOURCE}" = "true" ]; then
+        curl -L -s https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}.tar.gz -o cmake.tar.gz && tar xvf cmake.tar.gz && cd cmake-${CMAKE_VERSION} && ./configure && make && make install
+    else
+        apt-get update && \
+        apt-get install -y \
+            cmake && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/*
+    fi
+EOT
+
+# We install GRPC to a different prefix here so that we can copy in only the build artifacts later
+# saves several hundred MB on the final docker image size vs copying in the entire GRPC source tree
+# and running make install in the target container
+RUN git clone --recurse-submodules --jobs 4 -b ${GRPC_VERSION} --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+    mkdir -p /build/grpc/cmake/build && \
+    cd /build/grpc/cmake/build && \
+    sed -i "216i\  TESTONLY" "../../third_party/abseil-cpp/absl/container/CMakeLists.txt" && \
+    cmake -DgRPC_INSTALL=ON -DgRPC_BUILD_TESTS=OFF -DCMAKE_INSTALL_PREFIX:PATH=/opt/grpc ../.. && \
+    make && \
+    make install && \
+    rm -rf /build
+
+FROM ${BASE_IMAGE} AS builder
+ARG CMAKE_FROM_SOURCE=false
+ARG CMAKE_VERSION=3.31.10
+# We can target specific CUDA ARCHITECTURES like --build-arg CUDA_DOCKER_ARCH='75;86;89;120'
+ARG CUDA_DOCKER_ARCH
+ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
+ARG CMAKE_ARGS
+ENV CMAKE_ARGS=${CMAKE_ARGS}
+ARG BACKEND=rerankers
+ARG BUILD_TYPE
+ENV BUILD_TYPE=${BUILD_TYPE}
+ARG CUDA_MAJOR_VERSION
+ARG CUDA_MINOR_VERSION
+ARG SKIP_DRIVERS=false
+ENV CUDA_MAJOR_VERSION=${CUDA_MAJOR_VERSION}
+ENV CUDA_MINOR_VERSION=${CUDA_MINOR_VERSION}
+ENV DEBIAN_FRONTEND=noninteractive
+ARG TARGETARCH
+ARG TARGETVARIANT
+ARG GO_VERSION=1.25.4
+ARG UBUNTU_VERSION=2404
+
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        build-essential \
+        ccache git \
+        ca-certificates \
+        make \
+        pkg-config libcurl4-openssl-dev \
+        curl unzip \
+        libssl-dev wget && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# Cuda
+ENV PATH=/usr/local/cuda/bin:${PATH}
+
+# HipBLAS requirements
+ENV PATH=/opt/rocm/bin:${PATH}
+
+
+# Vulkan requirements
+RUN <<EOT bash
+    if [ "${BUILD_TYPE}" = "vulkan" ] && [ "${SKIP_DRIVERS}" = "false" ]; then
+        apt-get update && \
+        apt-get install -y  --no-install-recommends \
+            software-properties-common pciutils wget gpg-agent && \
+        apt-get install -y libglm-dev cmake libxcb-dri3-0 libxcb-present0 libpciaccess0 \
+            libpng-dev libxcb-keysyms1-dev libxcb-dri3-dev libx11-dev g++ gcc \
+            libwayland-dev libxrandr-dev libxcb-randr0-dev libxcb-ewmh-dev \
+            git python-is-python3 bison libx11-xcb-dev liblz4-dev libzstd-dev \
+            ocaml-core ninja-build pkg-config libxml2-dev wayland-protocols python3-jsonschema \
+            clang-format qtbase5-dev qt6-base-dev libxcb-glx0-dev sudo xz-utils
+        if [ "amd64" = "$TARGETARCH" ]; then
+            wget "https://sdk.lunarg.com/sdk/download/1.4.335.0/linux/vulkansdk-linux-x86_64-1.4.335.0.tar.xz" && \
+            tar -xf vulkansdk-linux-x86_64-1.4.335.0.tar.xz && \
+            rm vulkansdk-linux-x86_64-1.4.335.0.tar.xz && \
+            mkdir -p /opt/vulkan-sdk && \
+            mv 1.4.335.0 /opt/vulkan-sdk/ && \
+            cd /opt/vulkan-sdk/1.4.335.0 && \
+            ./vulkansdk --no-deps --maxjobs \
+                vulkan-loader \
+                vulkan-validationlayers \
+                vulkan-extensionlayer \
+                vulkan-tools \
+                shaderc && \
+            cp -rfv /opt/vulkan-sdk/1.4.335.0/x86_64/bin/* /usr/bin/ && \
+            cp -rfv /opt/vulkan-sdk/1.4.335.0/x86_64/lib/* /usr/lib/x86_64-linux-gnu/ && \
+            cp -rfv /opt/vulkan-sdk/1.4.335.0/x86_64/include/* /usr/include/ && \
+            cp -rfv /opt/vulkan-sdk/1.4.335.0/x86_64/share/* /usr/share/ && \
+            rm -rf /opt/vulkan-sdk
+        fi
+        if [ "arm64" = "$TARGETARCH" ]; then
+            mkdir vulkan && cd vulkan && \
+            curl -L -o vulkan-sdk.tar.xz https://github.com/mudler/vulkan-sdk-arm/releases/download/1.4.335.0/vulkansdk-ubuntu-24.04-arm-1.4.335.0.tar.xz && \
+            tar -xvf vulkan-sdk.tar.xz && \
+            rm vulkan-sdk.tar.xz && \
+            cd 1.4.335.0 && \
+            cp -rfv aarch64/bin/* /usr/bin/ && \
+            cp -rfv aarch64/lib/* /usr/lib/aarch64-linux-gnu/ && \
+            cp -rfv aarch64/include/* /usr/include/ && \
+            cp -rfv aarch64/share/* /usr/share/ && \
+            cd ../.. && \
+            rm -rf vulkan
+        fi
+        ldconfig && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/*
+    fi
+EOT
+
+# CuBLAS requirements
+RUN <<EOT bash
+    if ( [ "${BUILD_TYPE}" = "cublas" ] || [ "${BUILD_TYPE}" = "l4t" ] ) && [ "${SKIP_DRIVERS}" = "false" ]; then
+        apt-get update && \
+        apt-get install -y  --no-install-recommends \
+            software-properties-common pciutils
+        if [ "amd64" = "$TARGETARCH" ]; then
+            curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${UBUNTU_VERSION}/x86_64/cuda-keyring_1.1-1_all.deb
+        fi
+        if [ "arm64" = "$TARGETARCH" ]; then
+            if [ "${CUDA_MAJOR_VERSION}" = "13" ]; then
+                curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${UBUNTU_VERSION}/sbsa/cuda-keyring_1.1-1_all.deb
+            else
+                curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${UBUNTU_VERSION}/arm64/cuda-keyring_1.1-1_all.deb
+            fi
+        fi
+        dpkg -i cuda-keyring_1.1-1_all.deb && \
+        rm -f cuda-keyring_1.1-1_all.deb && \
+        apt-get update && \
+        apt-get install -y --no-install-recommends \
+            cuda-nvcc-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcufft-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcurand-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcublas-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcusparse-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcusolver-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION}
+        if [ "${CUDA_MAJOR_VERSION}" = "13" ] && [ "arm64" = "$TARGETARCH" ]; then
+            apt-get install -y --no-install-recommends \
+            libcufile-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcudnn9-cuda-${CUDA_MAJOR_VERSION} cuda-cupti-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libnvjitlink-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION}
+        fi
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/*
+    fi
+EOT
+
+
+# https://github.com/NVIDIA/Isaac-GR00T/issues/343
+RUN <<EOT bash
+    if [ "${BUILD_TYPE}" = "cublas" ] && [ "${TARGETARCH}" = "arm64" ]; then
+        wget https://developer.download.nvidia.com/compute/cudss/0.6.0/local_installers/cudss-local-tegra-repo-ubuntu${UBUNTU_VERSION}-0.6.0_0.6.0-1_arm64.deb && \
+        dpkg -i cudss-local-tegra-repo-ubuntu${UBUNTU_VERSION}-0.6.0_0.6.0-1_arm64.deb && \
+        cp /var/cudss-local-tegra-repo-ubuntu${UBUNTU_VERSION}-0.6.0/cudss-*-keyring.gpg /usr/share/keyrings/ && \
+        apt-get update && apt-get -y install cudss cudss-cuda-${CUDA_MAJOR_VERSION} && \
+        wget https://developer.download.nvidia.com/compute/nvpl/25.5/local_installers/nvpl-local-repo-ubuntu${UBUNTU_VERSION}-25.5_1.0-1_arm64.deb && \
+        dpkg -i nvpl-local-repo-ubuntu${UBUNTU_VERSION}-25.5_1.0-1_arm64.deb && \
+        cp /var/nvpl-local-repo-ubuntu${UBUNTU_VERSION}-25.5/nvpl-*-keyring.gpg /usr/share/keyrings/ && \
+        apt-get update && apt-get install -y nvpl
+    fi
+EOT
+
+# If we are building with clblas support, we need the libraries for the builds
+RUN if [ "${BUILD_TYPE}" = "clblas" ] && [ "${SKIP_DRIVERS}" = "false" ]; then \
+        apt-get update && \
+        apt-get install -y --no-install-recommends \
+            libclblast-dev && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/* \
+    ; fi
+
+RUN if [ "${BUILD_TYPE}" = "hipblas" ] && [ "${SKIP_DRIVERS}" = "false" ]; then \
+        apt-get update && \
+        apt-get install -y --no-install-recommends \
+            hipblas-dev \
+            rocblas-dev && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/* && \
+        # I have no idea why, but the ROCM lib packages don't trigger ldconfig after they install, which results in local-ai and others not being able
+        # to locate the libraries. We run ldconfig ourselves to work around this packaging deficiency
+        ldconfig && \
+        # Log which GPU architectures have rocBLAS kernel support
+        echo "rocBLAS library data architectures:" && \
+        (ls /opt/rocm*/lib/rocblas/library/Kernels* 2>/dev/null || ls /opt/rocm*/lib64/rocblas/library/Kernels* 2>/dev/null) | grep -oP 'gfx[0-9a-z+-]+' | sort -u || \
+        echo "WARNING: No rocBLAS kernel data found" \
+    ; fi
+
+RUN echo "TARGETARCH: $TARGETARCH"
+
+# We need protoc installed, and the version in 22.04 is too old.  We will create one as part installing the GRPC build below
+# but that will also being in a newer version of absl which stablediffusion cannot compile with.  This version of protoc is only
+# here so that we can generate the grpc code for the stablediffusion build
+RUN <<EOT bash
+    if [ "amd64" = "$TARGETARCH" ]; then
+        curl -L -s https://github.com/protocolbuffers/protobuf/releases/download/v27.1/protoc-27.1-linux-x86_64.zip -o protoc.zip && \
+        unzip -j -d /usr/local/bin protoc.zip bin/protoc && \
+        rm protoc.zip
+    fi
+    if [ "arm64" = "$TARGETARCH" ]; then
+        curl -L -s https://github.com/protocolbuffers/protobuf/releases/download/v27.1/protoc-27.1-linux-aarch_64.zip -o protoc.zip && \
+        unzip -j -d /usr/local/bin protoc.zip bin/protoc && \
+        rm protoc.zip
+    fi
+EOT
+
+# Install CMake (the version in 22.04 is too old)
+RUN <<EOT bash
+    if [ "${CMAKE_FROM_SOURCE}" = "true" ]; then
+        curl -L -s https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}.tar.gz -o cmake.tar.gz && tar xvf cmake.tar.gz && cd cmake-${CMAKE_VERSION} && ./configure && make && make install
+    else
+        apt-get update && \
+        apt-get install -y \
+            cmake && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/*
+    fi
+EOT
+
+COPY --from=grpc /opt/grpc /usr/local
+
+
+COPY . /LocalAI
+
+RUN <<'EOT' bash
+set -euxo pipefail
+
+if [[ -n "${CUDA_DOCKER_ARCH:-}" ]]; then
+  CUDA_ARCH_ESC="${CUDA_DOCKER_ARCH//;/\\;}"
+  export CMAKE_ARGS="${CMAKE_ARGS:-} -DCMAKE_CUDA_ARCHITECTURES=${CUDA_ARCH_ESC}"
+  echo "CMAKE_ARGS(env) = ${CMAKE_ARGS}"
+  rm -rf /LocalAI/backend/cpp/buun-llama-cpp-*-build
+fi
+
+cd /LocalAI/backend/cpp/buun-llama-cpp
+
+if [ "${TARGETARCH}" = "arm64" ] || [ "${BUILD_TYPE}" = "hipblas" ]; then
+  make buun-llama-cpp-fallback
+  make buun-llama-cpp-grpc
+  make buun-llama-cpp-rpc-server
+else
+  make buun-llama-cpp-avx
+  make buun-llama-cpp-avx2
+  make buun-llama-cpp-avx512
+  make buun-llama-cpp-fallback
+  make buun-llama-cpp-grpc
+  make buun-llama-cpp-rpc-server
+fi
+EOT
+
+
+# Copy libraries using a script to handle architecture differences
+RUN make -BC /LocalAI/backend/cpp/buun-llama-cpp package
+
+
+FROM scratch
+
+
+# Copy all available binaries (the build process only creates the appropriate ones for the target architecture)
+COPY --from=builder /LocalAI/backend/cpp/buun-llama-cpp/package/. ./

backend/backend.proto

@@ -15,6 +15,7 @@ service Backend {
   rpc PredictStream(PredictOptions) returns (stream Reply) {}
   rpc Embedding(PredictOptions) returns (EmbeddingResult) {}
   rpc GenerateImage(GenerateImageRequest) returns (Result) {}
+  rpc UpscaleImage(UpscaleImageRequest) returns (Result) {}
   rpc GenerateVideo(GenerateVideoRequest) returns (Result) {}
   rpc Generate3D(Generate3DRequest) returns (Result) {}
   rpc AudioTranscription(TranscriptRequest) returns (TranscriptResult) {}
@@ -637,6 +638,12 @@ message GenerateImageRequest {
   string ModelIdentity = 13;
 }
 
+message UpscaleImageRequest {
+  string src = 1;   // input image path
+  string dst = 2;   // output image path
+  int32  scale = 3; // upscale factor (e.g. 2 or 4)
+}
+
 message GenerateVideoRequest {
   string prompt = 1;
   string negative_prompt = 2;  // Negative prompt for video generation

backend/cpp/audio-cpp/Makefile

@@ -9,7 +9,7 @@
 # recipe is a make target (not a prepare.sh) so 'make purge && make' is a clean
 # rebuild and so the bump bot can see the pin.
 
-AUDIO_CPP_VERSION?=545e29a6f2fde24298cb3b0f07baab4352987ac9
+AUDIO_CPP_VERSION?=5a8312ef7b8aa7cf14e9a24ac568cabd8725d68a
 AUDIO_CPP_REPO?=https://github.com/0xShug0/audio.cpp
 
 CURRENT_MAKEFILE_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))

backend/cpp/buun-llama-cpp/Makefile

@@ -0,0 +1,92 @@
+
+# Pinned to the HEAD of master on https://github.com/spiritbuun/buun-llama-cpp.
+# Auto-bumped nightly by .github/workflows/bump_deps.yaml.
+BUUN_LLAMA_VERSION?=22464d0848b87c5d56b52fdf6af2e5da46bf803e
+LLAMA_REPO?=https://github.com/spiritbuun/buun-llama-cpp
+
+CMAKE_ARGS?=
+BUILD_TYPE?=
+NATIVE?=false
+ONEAPI_VARS?=/opt/intel/oneapi/setvars.sh
+TARGET?=--target grpc-server
+JOBS?=$(shell nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null || echo 1)
+ARCH?=$(shell uname -m)
+
+CURRENT_MAKEFILE_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))
+LLAMA_CPP_DIR := $(CURRENT_MAKEFILE_DIR)/../llama-cpp
+
+GREEN := \033[0;32m
+RESET := \033[0m
+
+# buun-llama-cpp is a llama.cpp fork-of-a-fork (spiritbuun/buun-llama-cpp forked
+# TheTom/llama-cpp-turboquant, which itself forked ggml-org/llama.cpp). Rather
+# than duplicating grpc-server.cpp / CMakeLists.txt / prepare.sh we reuse the
+# ones in backend/cpp/llama-cpp, and only swap which repo+sha the fetch step
+# pulls. Each flavor target copies ../llama-cpp into a sibling
+# ../buun-llama-cpp-<flavor>-build directory, then invokes llama-cpp's own
+# build-llama-cpp-grpc-server with LLAMA_REPO/LLAMA_VERSION overridden to point
+# at the fork.
+PATCHES_DIR := $(CURRENT_MAKEFILE_DIR)/patches
+
+# Each flavor target:
+#   1. copies backend/cpp/llama-cpp/ (grpc-server.cpp + prepare.sh + CMakeLists.txt + Makefile)
+#      into a sibling buun-llama-cpp-<flavor>-build directory;
+#   2. clones the buun fork into buun-llama-cpp-<flavor>-build/llama.cpp via the
+#      copy's own `llama.cpp` target, overriding LLAMA_REPO/LLAMA_VERSION;
+#   3. applies patches from backend/cpp/buun-llama-cpp/patches/ to the cloned
+#      fork sources (for backporting upstream commits the fork hasn't pulled);
+#   4. runs the copy's `grpc-server` target, which produces the binary we copy
+#      up as buun-llama-cpp-<flavor>.
+define buun-llama-cpp-build
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build
+	cp -rf $(LLAMA_CPP_DIR) $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build
+	# Stock llama.cpp patches target upstream and may not apply to this fork.
+	# The buun-specific compatibility series is applied explicitly below.
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build/patches
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build purge
+	# Augment the copied grpc-server.cpp's KV-cache allow-list with the
+	# fork's turbo2/turbo3/turbo4/turbo2_tcq/turbo3_tcq types and wire up the
+	# DFlash-specific option handlers (tree_budget / draft_topk). We patch the
+	# *copy*, never the original under backend/cpp/llama-cpp/, so the stock
+	# llama-cpp build stays compiling against vanilla upstream.
+	bash $(CURRENT_MAKEFILE_DIR)/patch-grpc-server.sh $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build/grpc-server.cpp
+	bash $(LLAMA_CPP_DIR)/disable-score-task.sh $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build/grpc-server.cpp
+	$(info $(GREEN)I buun-llama-cpp build info:$(1)$(RESET))
+	LLAMA_REPO=$(LLAMA_REPO) LLAMA_VERSION=$(BUUN_LLAMA_VERSION) \
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build llama.cpp
+	bash $(CURRENT_MAKEFILE_DIR)/apply-patches.sh $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build/llama.cpp $(PATCHES_DIR)
+	CMAKE_ARGS="$(CMAKE_ARGS) $(2)" TARGET="$(3)" \
+	LLAMA_REPO=$(LLAMA_REPO) LLAMA_VERSION=$(BUUN_LLAMA_VERSION) \
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build grpc-server
+	cp -rfv $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-$(1)-build/grpc-server buun-llama-cpp-$(1)
+endef
+
+buun-llama-cpp-avx2:
+	$(call buun-llama-cpp-build,avx2,-DGGML_AVX=on -DGGML_AVX2=on -DGGML_AVX512=off -DGGML_FMA=on -DGGML_F16C=on,--target grpc-server)
+
+buun-llama-cpp-avx512:
+	$(call buun-llama-cpp-build,avx512,-DGGML_AVX=on -DGGML_AVX2=off -DGGML_AVX512=on -DGGML_FMA=on -DGGML_F16C=on,--target grpc-server)
+
+buun-llama-cpp-avx:
+	$(call buun-llama-cpp-build,avx,-DGGML_AVX=on -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off,--target grpc-server)
+
+buun-llama-cpp-fallback:
+	$(call buun-llama-cpp-build,fallback,-DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off,--target grpc-server)
+
+buun-llama-cpp-grpc:
+	$(call buun-llama-cpp-build,grpc,-DGGML_RPC=ON -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off,--target grpc-server --target rpc-server)
+
+buun-llama-cpp-rpc-server: buun-llama-cpp-grpc
+	cp -rf $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-grpc-build/llama.cpp/build/bin/rpc-server buun-llama-cpp-rpc-server
+
+package:
+	bash package.sh
+
+test:
+	bash test-patch-grpc-server.sh
+
+purge:
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../buun-llama-cpp-*-build
+	rm -rf buun-llama-cpp-* package
+
+clean: purge

backend/cpp/buun-llama-cpp/apply-patches.sh

@@ -0,0 +1,50 @@
+#!/bin/bash
+# Apply the buun-llama-cpp patch series to a cloned buun-llama-cpp checkout.
+#
+# buun-llama-cpp is a fork-of-a-fork that branched off upstream llama.cpp
+# before some API changes the shared backend/cpp/llama-cpp/grpc-server.cpp
+# depends on. We carry those upstream commits as patch files under
+# backend/cpp/buun-llama-cpp/patches/ and apply them here so the reused
+# grpc-server source compiles against the fork unmodified.
+#
+# Drop the corresponding patch from patches/ whenever the fork catches up with
+# upstream — the build will fail fast if a patch stops applying, which is the
+# signal to retire it.
+
+set -euo pipefail
+
+if [[ $# -ne 2 ]]; then
+    echo "usage: $0 <llama.cpp-src-dir> <patches-dir>" >&2
+    exit 2
+fi
+
+SRC_DIR=$1
+PATCHES_DIR=$2
+
+if [[ ! -d "$SRC_DIR" ]]; then
+    echo "source dir does not exist: $SRC_DIR" >&2
+    exit 2
+fi
+
+if [[ ! -d "$PATCHES_DIR" ]]; then
+    echo "no patches dir at $PATCHES_DIR, nothing to apply"
+    exit 0
+fi
+
+shopt -s nullglob
+patches=("$PATCHES_DIR"/*.patch)
+shopt -u nullglob
+
+if [[ ${#patches[@]} -eq 0 ]]; then
+    echo "no .patch files in $PATCHES_DIR, nothing to apply"
+    exit 0
+fi
+
+cd "$SRC_DIR"
+
+for patch in "${patches[@]}"; do
+    echo "==> applying $patch"
+    git apply --verbose "$patch"
+done
+
+echo "all buun-llama-cpp patches applied successfully"

backend/cpp/buun-llama-cpp/package.sh

@@ -0,0 +1,57 @@
+#!/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
+
+cp -avrf $CURDIR/buun-llama-cpp-* $CURDIR/package/
+cp -rfv $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 $CURDIR/package/lib/libpthread.so.0
+else
+    echo "Error: Could not detect architecture"
+    exit 1
+fi
+
+# Package GPU libraries based on BUILD_TYPE
+GPU_LIB_SCRIPT="${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/

backend/cpp/buun-llama-cpp/patch-grpc-server.sh

@@ -0,0 +1,196 @@
+#!/bin/bash
+# Patch the shared backend/cpp/llama-cpp/grpc-server.cpp *copy* used by the
+# buun-llama-cpp build to account for three gaps between upstream and the fork:
+#
+#   1. Augment the kv_cache_types[] allow-list so `LoadModel` accepts the
+#      fork-specific `turbo2` / `turbo3` / `turbo4` cache types plus the buun
+#      additions `turbo2_tcq` / `turbo3_tcq`.
+#
+#   2. Adapt the post-refactor speculative-decoding fields and option handlers
+#      to the fork's legacy flat common_params_speculative layout, while adding
+#      buun-exclusive tree_budget / draft_topk support.
+#      These reference struct fields (common_params.speculative.tree_budget
+#      and .draft_topk) that only exist in buun's common/common.h — adding
+#      them to the shared backend/cpp/llama-cpp/grpc-server.cpp would break
+#      the stock llama-cpp build, so we inject them only into the buun copy.
+#
+#   3. Replace `get_media_marker()` (added upstream in ggml-org/llama.cpp#21962,
+#      server-side random per-instance marker) with the legacy "<__media__>"
+#      literal. The fork branched before that PR, so server-common.cpp has no
+#      get_media_marker symbol. The fork's mtmd_default_marker() still returns
+#      "<__media__>", and Go-side tooling falls back to that sentinel when the
+#      backend does not expose media_marker, so substituting the literal keeps
+#      behavior identical on the buun path.
+#
+# We patch the *copy* sitting in buun-llama-cpp-<flavor>-build/, never the
+# original under backend/cpp/llama-cpp/, so the stock llama-cpp build keeps
+# compiling against vanilla upstream.
+#
+# Idempotent: skips each insertion if its marker is already present (so re-runs
+# of the same build dir don't double-insert).
+
+set -euo pipefail
+
+if [[ $# -ne 1 ]]; then
+    echo "usage: $0 <grpc-server.cpp>" >&2
+    exit 2
+fi
+
+SRC=$1
+
+if [[ ! -f "$SRC" ]]; then
+    echo "grpc-server.cpp not found at $SRC" >&2
+    exit 2
+fi
+
+if grep -q 'GGML_TYPE_TURBO2_TCQ' "$SRC"; then
+    echo "==> $SRC already has buun cache types, skipping KV allow-list patch"
+else
+    echo "==> patching $SRC to allow turbo2/turbo3/turbo4/turbo2_tcq/turbo3_tcq KV-cache types"
+
+    # Insert the five TURBO entries right after the first `    GGML_TYPE_Q5_1,`
+    # line (the kv_cache_types[] allow-list). Using awk because the builder
+    # image does not ship python3, and GNU sed's multi-line `a\` quoting is
+    # awkward.
+    awk '
+        /^    GGML_TYPE_Q5_1,$/ && !done {
+            print
+            print "    // buun-llama-cpp fork extras — added by patch-grpc-server.sh"
+            print "    GGML_TYPE_TURBO2_0,"
+            print "    GGML_TYPE_TURBO3_0,"
+            print "    GGML_TYPE_TURBO4_0,"
+            print "    GGML_TYPE_TURBO2_TCQ,"
+            print "    GGML_TYPE_TURBO3_TCQ,"
+            done = 1
+            next
+        }
+        { print }
+        END {
+            if (!done) {
+                print "patch-grpc-server.sh: anchor `    GGML_TYPE_Q5_1,` not found" > "/dev/stderr"
+                exit 1
+            }
+        }
+    ' "$SRC" > "$SRC.tmp"
+    mv "$SRC.tmp" "$SRC"
+
+    echo "==> KV allow-list patch OK"
+fi
+
+if grep -q 'buun-llama-cpp legacy speculative options' "$SRC"; then
+    echo "==> $SRC already has legacy speculative option handlers, skipping"
+else
+    echo "==> replacing modern speculative option handlers with the fork-compatible set"
+
+    # Replace the whole speculative option section. The fork predates chained
+    # speculative types and the nested draft/ngram families, so retaining any
+    # modern-only handler makes the copied server fail at compile time.
+    awk '
+        /} else if \(!strcmp\(optname, "spec_type"\)/ && !done {
+            print "        // buun-llama-cpp legacy speculative options"
+            print "        } else if (!strcmp(optname, \"spec_type\") || !strcmp(optname, \"speculative_type\")) {"
+            print "            auto type = common_speculative_type_from_name(optval_str.substr(0, optval_str.find(\",\")));"
+            print "            if (type != COMMON_SPECULATIVE_TYPE_COUNT) params.speculative.type = type;"
+            print "        } else if (!strcmp(optname, \"spec_n_max\") || !strcmp(optname, \"draft_max\")) {"
+            print "            if (optval != NULL) { try { params.speculative.n_max = std::stoi(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"spec_n_min\") || !strcmp(optname, \"draft_min\")) {"
+            print "            if (optval != NULL) { try { params.speculative.n_min = std::stoi(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"spec_p_min\") || !strcmp(optname, \"draft_p_min\")) {"
+            print "            if (optval != NULL) { try { params.speculative.p_min = std::stof(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"spec_p_split\")) {"
+            print "            if (optval != NULL) { try { params.speculative.p_split = std::stof(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"spec_ngram_size_n\") || !strcmp(optname, \"ngram_size_n\")) {"
+            print "            if (optval != NULL) { try { params.speculative.ngram_size_n = (uint16_t)std::stoi(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"spec_ngram_size_m\") || !strcmp(optname, \"ngram_size_m\")) {"
+            print "            if (optval != NULL) { try { params.speculative.ngram_size_m = (uint16_t)std::stoi(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"spec_ngram_min_hits\") || !strcmp(optname, \"ngram_min_hits\")) {"
+            print "            if (optval != NULL) { try { params.speculative.ngram_min_hits = (uint16_t)std::stoi(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"draft_gpu_layers\")) {"
+            print "            if (optval != NULL) { try { params.speculative.n_gpu_layers = std::stoi(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"tree_budget\")) {"
+            print "            if (optval != NULL) { try { params.speculative.tree_budget = std::stoi(optval_str); } catch (...) {} }"
+            print "        } else if (!strcmp(optname, \"draft_topk\")) {"
+            print "            if (optval != NULL) { try { params.speculative.draft_topk = std::stoi(optval_str); } catch (...) {} }"
+            skipping = 1
+            next
+        }
+        skipping && /^    }$/ { skipping = 0; done = 1; print; next }
+        !skipping { print }
+        END {
+            if (!done) {
+                print "patch-grpc-server.sh: speculative option section not found" > "/dev/stderr"
+                exit 1
+            }
+        }
+    ' "$SRC" > "$SRC.tmp"
+    mv "$SRC.tmp" "$SRC"
+
+    echo "==> legacy speculative option-handler patch OK"
+fi
+
+# The modern server initializes a vector of speculative types when DraftModel
+# is present. The fork still exposes a single enum value.
+awk '
+    /const bool no_spec_type = params\.speculative\.types\.empty\(\)/ && !done {
+        print "        if (params.speculative.type == COMMON_SPECULATIVE_TYPE_NONE) {"
+        print "            params.speculative.type = COMMON_SPECULATIVE_TYPE_DRAFT;"
+        print "        }"
+        skipping = 1
+        next
+    }
+    skipping && /^        }$/ { skipping = 0; done = 1; next }
+    !skipping { print }
+' "$SRC" > "$SRC.tmp"
+mv "$SRC.tmp" "$SRC"
+
+# Map supported post-refactor fields back to the names used by the pinned fork.
+sed -E \
+    -e 's/params\.speculative\.draft\.mparams\.path/params.speculative.mparams_dft.path/g' \
+    -e 's/params\.speculative\.draft\.n_gpu_layers/params.speculative.n_gpu_layers/g' \
+    -e 's/ctx_server\.impl->model_tgt/ctx_server.impl->model/g' \
+    -e '/params\.cache_idle_slots =/d' \
+    -e '/params\.split_mode = LLAMA_SPLIT_MODE_TENSOR;/d' \
+    -e '/params\.speculative\.draft\.tensor_buft_overrides/d' \
+    "$SRC" > "$SRC.tmp"
+mv "$SRC.tmp" "$SRC"
+
+if ! grep -q '^#define LOCALAI_TURBOQUANT_NO_CHECKPOINT_MIN_STEP' "$SRC"; then
+    sed '0,/^#include/{s/^#include/#define LOCALAI_TURBOQUANT_NO_CHECKPOINT_MIN_STEP 1\n\n#include/}' "$SRC" > "$SRC.tmp"
+    mv "$SRC.tmp" "$SRC"
+fi
+
+if grep -qE 'ctx_server\.get_meta\(\)\.logit_bias_eog|params_base\.sampling\.logit_bias_eog,' "$SRC"; then
+    echo "==> patching $SRC to drop the logit_bias_eog arg from params_from_json_cmpl() callsites (buun still uses the pre-refactor 4-arg signature)"
+    # Upstream llama.cpp refactored params_from_json_cmpl to take a precomputed
+    # logit_bias_eog vector after buun's 2026-04-05 fork-point — simultaneously
+    # adding server_context_meta::logit_bias_eog as the supplier. Buun carries
+    # neither change: its params_from_json_cmpl is still 4-arg, and internally
+    # derives logit_bias_eog from the common_params it's passed. So we just
+    # delete the argument line entirely — the remaining 4 args match buun's
+    # signature and the resulting behavior matches upstream bit-for-bit
+    # (upstream's 5th arg is the same data buun derives internally).
+    #
+    # Guard is broad so this works whether the line has been run through this
+    # block before (leaving params_base.sampling.logit_bias_eog,) or not
+    # (leaving the original ctx_server.get_meta().logit_bias_eog,).
+    sed -E '/^[[:space:]]+(ctx_server\.get_meta\(\)\.logit_bias_eog|params_base\.sampling\.logit_bias_eog),$/d' "$SRC" > "$SRC.tmp"
+    mv "$SRC.tmp" "$SRC"
+    echo "==> logit_bias_eog arg drop OK"
+else
+    echo "==> $SRC has no logit_bias_eog arg line, skipping"
+fi
+
+if grep -q 'get_media_marker()' "$SRC"; then
+    echo "==> patching $SRC to replace get_media_marker() with legacy \"<__media__>\" literal"
+    # Only one call site today (ModelMetadata), but replace all occurrences to
+    # stay robust if upstream adds more. Use a temp file to avoid relying on
+    # sed -i portability (the builder image uses GNU sed, but keeping this
+    # consistent with the awk block above).
+    sed 's/get_media_marker()/"<__media__>"/g' "$SRC" > "$SRC.tmp"
+    mv "$SRC.tmp" "$SRC"
+    echo "==> get_media_marker() substitution OK"
+else
+    echo "==> $SRC has no get_media_marker() call, skipping media-marker patch"
+fi
+
+echo "==> all patches applied"

backend/cpp/buun-llama-cpp/patches/0001-fattn-atomicAdd-double-shim.patch

@@ -0,0 +1,46 @@
+Subject: [PATCH] ggml-cuda/fattn: provide atomicAdd(double*,double) shim for pre-sm_60
+
+Buun's Q² calibration path in ggml_cuda_turbo_scale_q calls
+  atomicAdd(&d_q_channel_sq_fattn[threadIdx.x], (double)(val * val));
+but native double atomicAdd is only available on compute capability 6.0
+and newer. Compiling against a CUDA arch list that includes older
+architectures (LocalAI's CUDA 12 Docker image builds for the full
+published arch range) fails with:
+
+    fattn.cu(812): error: no instance of overloaded function "atomicAdd"
+      matches the argument list, argument types are: (double *, double)
+
+Add the canonical CUDA-programming-guide shim at the top of fattn.cu so
+pre-sm_60 codegen has a definition to call. On sm_60+ the native CUDA
+intrinsic is used and the shim is elided via __CUDA_ARCH__.
+
+--- a/ggml/src/ggml-cuda/fattn.cu
++++ b/ggml/src/ggml-cuda/fattn.cu
+@@ -7,6 +7,27 @@
+
+ #include <atomic>
+
++// Pre-sm_60 double atomicAdd shim. Native double atomicAdd(double*,double)
++// is only available on CUDA compute capability 6.0+ (see CUDA C Programming
++// Guide, B.15 Atomic Functions). Buun's Q² calibration path below calls
++// atomicAdd with a double*; without this definition, nvcc fails to find a
++// matching overload whenever the compile target list includes pre-sm_60
++// architectures. The standard CAS loop implementation below matches the
++// semantics of the native intrinsic.
++#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 600
++static __device__ double atomicAdd(double * address, double val) {
++    unsigned long long int * address_as_ull = (unsigned long long int *)address;
++    unsigned long long int old = *address_as_ull;
++    unsigned long long int assumed;
++    do {
++        assumed = old;
++        old = atomicCAS(address_as_ull, assumed,
++                        __double_as_longlong(val + __longlong_as_double(assumed)));
++    } while (assumed != old);
++    return __longlong_as_double(old);
++}
++#endif
++
+ // InnerQ: update the fattn-side inverse scale array from host (all devices)
+ void turbo_innerq_update_fattn_scales(const float * scale_inv) {
+     int cur_device;

backend/cpp/buun-llama-cpp/patches/0002-argmax-shfl-xor-sync-add-width.patch

@@ -0,0 +1,32 @@
+Subject: [PATCH] ggml-cuda/argmax: pass WARP_SIZE to the top-K __shfl_xor_sync calls
+
+Two __shfl_xor_sync calls in the top-K intra-warp merge drop the `width`
+argument and rely on the CUDA default (warpSize). Every other call in
+the same file already passes WARP_SIZE explicitly, and the HIP/ROCm
+compatibility shim at ggml/src/ggml-cuda/vendors/hip.h:33 is a 4-arg
+function-like macro — so the 3-arg form fails to preprocess when
+building with hipcc against ROCm:
+
+    argmax.cu:265: error: too few arguments provided to function-like
+      macro invocation
+    note: macro '__shfl_xor_sync' defined here:
+      #define __shfl_xor_sync(mask, var, laneMask, width) \
+              __shfl_xor(var, laneMask, width)
+
+Align the two call sites with the rest of the file by passing WARP_SIZE
+explicitly. On CUDA the generated code is unchanged (warpSize is the
+default); on HIP it now matches the macro's arity.
+
+--- a/ggml/src/ggml-cuda/argmax.cu
++++ b/ggml/src/ggml-cuda/argmax.cu
+@@ -262,8 +262,8 @@
+     // Each step: lane gets partner's min element, if it beats our min, replace and re-heapify
+     for (int offset = WARP_SIZE / 2; offset > 0; offset >>= 1) {
+         for (int i = 0; i < K; i++) {
+-            float partner_val = __shfl_xor_sync(0xFFFFFFFF, heap_val[i], offset);
+-            int partner_idx = __shfl_xor_sync(0xFFFFFFFF, heap_idx[i], offset);
++            float partner_val = __shfl_xor_sync(0xFFFFFFFF, heap_val[i], offset, WARP_SIZE);
++            int partner_idx = __shfl_xor_sync(0xFFFFFFFF, heap_idx[i], offset, WARP_SIZE);
+             if (partner_val > heap_val[0]) {
+                 heap_val[0] = partner_val;
+                 heap_idx[0] = partner_idx;

backend/cpp/buun-llama-cpp/patches/0003-hip-add-memcpy-symbol-aliases.patch

@@ -0,0 +1,24 @@
+Subject: [PATCH] ggml-cuda/vendors/hip: alias cudaMemcpy{To,From}Symbol to hip counterparts
+
+Buun's Q² calibration + TCQ codebook upload paths in fattn.cu use
+cudaMemcpyToSymbol / cudaMemcpyFromSymbol. The HIP-compat header in
+ggml/src/ggml-cuda/vendors/hip.h already aliases the scalar cudaMemcpy
+family (cudaMemcpy, cudaMemcpyAsync, cudaMemcpy2DAsync, …) but is
+missing the symbol variants. Building with hipcc therefore fails with
+15+ "use of undeclared identifier 'cudaMemcpyToSymbol'" errors.
+
+Add the two missing aliases alongside the existing memcpy block. HIP
+provides hipMemcpy{To,From}Symbol with the same signature as CUDA's
+equivalents, so this is a straight name substitution.
+
+--- a/ggml/src/ggml-cuda/vendors/hip.h
++++ b/ggml/src/ggml-cuda/vendors/hip.h
+@@ -85,6 +85,8 @@
+ #define cudaMemcpyDeviceToDevice hipMemcpyDeviceToDevice
+ #define cudaMemcpyDeviceToHost hipMemcpyDeviceToHost
+ #define cudaMemcpyHostToDevice hipMemcpyHostToDevice
++#define cudaMemcpyToSymbol hipMemcpyToSymbol
++#define cudaMemcpyFromSymbol hipMemcpyFromSymbol
+ #define cudaMemcpyKind hipMemcpyKind
+ #define cudaMemset hipMemset
+ #define cudaMemsetAsync hipMemsetAsync

backend/cpp/buun-llama-cpp/patches/0004-fattn-fwht128-shfl-xor-sync-add-width.patch

@@ -0,0 +1,36 @@
+Subject: [PATCH] ggml-cuda/fattn: pass WARP_SIZE to fwht128 __shfl_xor_sync calls
+
+Same issue as the argmax top-K fix: two __shfl_xor_sync call sites in
+the FWHT-128 butterfly kernels (ggml_cuda_fwht128 and fwht128_store_half)
+use the 3-arg CUDA form and omit the `width` argument that the HIP
+function-like macro in vendors/hip.h:33 requires. Hipcc fails with:
+
+    fattn.cu:512: too few arguments provided to function-like macro
+      invocation
+    note: macro '__shfl_xor_sync' defined here:
+      #define __shfl_xor_sync(mask, var, laneMask, width) \
+              __shfl_xor(var, laneMask, width)
+
+Add WARP_SIZE to both calls. CUDA codegen is unchanged (warpSize is the
+default); HIP now matches the macro arity.
+
+--- a/ggml/src/ggml-cuda/fattn.cu
++++ b/ggml/src/ggml-cuda/fattn.cu
+@@ -509,7 +509,7 @@
+     // Intra-warp passes: shuffle xor with stride h, no smem, no sync.
+     #pragma unroll
+     for (int h = 1; h <= 16; h *= 2) {
+-        const float other = __shfl_xor_sync(0xFFFFFFFF, val, h);
++        const float other = __shfl_xor_sync(0xFFFFFFFF, val, h, WARP_SIZE);
+         val = (tid & h) ? (other - val) : (val + other);
+     }
+
+@@ -533,7 +533,7 @@
+ static __device__ __forceinline__ void fwht128_store_half(
+         float val, half * dst_base) {
+     const int tid = threadIdx.x;
+-    const float neighbor = __shfl_xor_sync(0xFFFFFFFF, val, 1);
++    const float neighbor = __shfl_xor_sync(0xFFFFFFFF, val, 1, WARP_SIZE);
+     if ((tid & 1) == 0) {
+         const half2 packed = __floats2half2_rn(val, neighbor);
+         *((half2 *)(dst_base + tid)) = packed;

backend/cpp/buun-llama-cpp/run.sh

@@ -0,0 +1,65 @@
+#!/bin/bash
+set -ex
+
+# Get the absolute current dir where the script is located
+CURDIR=$(dirname "$(realpath $0)")
+
+cd /
+
+echo "CPU info:"
+grep -e "model\sname" /proc/cpuinfo | head -1
+grep -e "flags" /proc/cpuinfo | head -1
+
+BINARY=buun-llama-cpp-fallback
+
+if grep -q -e "\savx\s" /proc/cpuinfo ; then
+	echo "CPU:    AVX    found OK"
+	if [ -e $CURDIR/buun-llama-cpp-avx ]; then
+		BINARY=buun-llama-cpp-avx
+	fi
+fi
+
+if grep -q -e "\savx2\s" /proc/cpuinfo ; then
+	echo "CPU:    AVX2   found OK"
+	if [ -e $CURDIR/buun-llama-cpp-avx2 ]; then
+		BINARY=buun-llama-cpp-avx2
+	fi
+fi
+
+# Check avx 512
+if grep -q -e "\savx512f\s" /proc/cpuinfo ; then
+	echo "CPU:    AVX512F found OK"
+	if [ -e $CURDIR/buun-llama-cpp-avx512 ]; then
+		BINARY=buun-llama-cpp-avx512
+	fi
+fi
+
+if [ -n "$LLAMACPP_GRPC_SERVERS" ]; then
+	if [ -e $CURDIR/buun-llama-cpp-grpc ]; then
+		BINARY=buun-llama-cpp-grpc
+	fi
+fi
+
+# Extend ld library path with the dir where this script is located/lib
+if [ "$(uname)" == "Darwin" ]; then
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
+	# Tell rocBLAS where to find TensileLibrary data (GPU kernel tuning files)
+	if [ -d "$CURDIR/lib/rocblas/library" ]; then
+		export ROCBLAS_TENSILE_LIBPATH=$CURDIR/lib/rocblas/library
+	fi
+fi
+
+# If there is a lib/ld.so, use it
+if [ -f $CURDIR/lib/ld.so ]; then
+	echo "Using lib/ld.so"
+	echo "Using binary: $BINARY"
+	exec $CURDIR/lib/ld.so $CURDIR/$BINARY "$@"
+fi
+
+echo "Using binary: $BINARY"
+exec $CURDIR/$BINARY "$@"
+
+# We should never reach this point, however just in case we do, run fallback
+exec $CURDIR/buun-llama-cpp-fallback "$@"

backend/cpp/buun-llama-cpp/test-patch-grpc-server.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+
+set -euo pipefail
+
+SCRIPT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+SOURCE="$SCRIPT_DIR/../llama-cpp/grpc-server.cpp"
+TMP_DIR=$(mktemp -d)
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+cp "$SOURCE" "$TMP_DIR/grpc-server.cpp"
+bash "$SCRIPT_DIR/patch-grpc-server.sh" "$TMP_DIR/grpc-server.cpp"
+bash "$SCRIPT_DIR/../llama-cpp/disable-score-task.sh" "$TMP_DIR/grpc-server.cpp"
+bash "$SCRIPT_DIR/patch-grpc-server.sh" "$TMP_DIR/grpc-server.cpp"
+
+for unsupported in \
+    'params.cache_idle_slots' \
+    'params.speculative.types' \
+    'params.speculative.draft.' \
+    'common_speculative_types_from_names' \
+    'COMMON_SPECULATIVE_TYPE_DRAFT_SIMPLE' \
+    'ctx_server.impl->model_tgt'; do
+    if grep -Fq "$unsupported" "$TMP_DIR/grpc-server.cpp"; then
+        echo "unsupported buun API remains: $unsupported" >&2
+        exit 1
+    fi
+done
+
+grep -Fq '#define LOCALAI_LLAMA_CPP_NO_SCORE_TASK 1' "$TMP_DIR/grpc-server.cpp"
+grep -Fq '#define LOCALAI_TURBOQUANT_NO_CHECKPOINT_MIN_STEP 1' "$TMP_DIR/grpc-server.cpp"
+grep -Fq 'params.speculative.mparams_dft.path = request->draftmodel();' "$TMP_DIR/grpc-server.cpp"
+grep -Fq 'params.speculative.type = COMMON_SPECULATIVE_TYPE_DRAFT;' "$TMP_DIR/grpc-server.cpp"
+grep -Fq 'ctx_server.impl->model' "$TMP_DIR/grpc-server.cpp"
+
+echo "buun grpc-server compatibility transform passed"

backend/cpp/ik-llama-cpp/Makefile

@@ -1,5 +1,5 @@
 
-IK_LLAMA_VERSION?=0be97a7a5ad113f33e08729261649ccea2cdc5ff
+IK_LLAMA_VERSION?=cb9147fd0d9c08a9a84eee5ac405a73f4e10e3e1
 LLAMA_REPO?=https://github.com/ikawrakow/ik_llama.cpp
 
 CMAKE_ARGS?=

backend/cpp/llama-cpp/run.sh

@@ -12,10 +12,11 @@ grep -e "flags" /proc/cpuinfo | head -1
 
 BINARY=llama-cpp-fallback
 
-# CPU images and x86 GPU images ship a single llama-cpp-cpu-all built with ggml
+# CPU images and most x86 GPU images ship a single llama-cpp-cpu-all built with ggml
 # CPU_ALL_VARIANTS: ggml's backend registry dlopens the best libggml-cpu-*.so for this
 # host, so no shell-side AVX probing. GPU arm64 images still ship llama-cpp-fallback
-# until their builder toolchains support ggml's complete arm variant matrix.
+# until their builder toolchains support ggml's complete arm variant matrix, and so do
+# the SYCL images, whose icpx compiler hangs on the sapphirerapids variant.
 if [ -e "$CURDIR"/llama-cpp-cpu-all ]; then
 	BINARY=llama-cpp-cpu-all
 fi

backend/cpp/turboquant/run.sh

@@ -12,11 +12,12 @@ grep -e "flags" /proc/cpuinfo | head -1
 
 BINARY=turboquant-fallback
 
-# CPU images and x86 GPU images ship a single turboquant-cpu-all built with ggml
+# CPU images and most x86 GPU images ship a single turboquant-cpu-all built with ggml
 # CPU_ALL_VARIANTS: ggml's
 # backend registry dlopens the best libggml-cpu-*.so for this host, so no shell-side
 # probing. GPU arm64 images still ship turboquant-fallback until their builder toolchains
-# support ggml's complete arm variant matrix.
+# support ggml's complete arm variant matrix, and so do the SYCL images, whose icpx
+# compiler hangs on the sapphirerapids variant.
 if [ -e "$CURDIR"/turboquant-cpu-all ]; then
 	BINARY=turboquant-cpu-all
 fi

backend/go/crispasr/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # CrispASR version (release tag)
 CRISPASR_REPO?=https://github.com/CrispStrobe/CrispASR
-CRISPASR_VERSION?=66ac7843e319b588f5410051c575affd19424fb3
+CRISPASR_VERSION?=fcb79282a6bc52e13d858026c42b24fb6e63c97a
 SO_TARGET?=libgocrispasr.so
 
 CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF

backend/go/stablediffusion-ggml/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # stablediffusion.cpp (ggml)
 STABLEDIFFUSION_GGML_REPO?=https://github.com/leejet/stable-diffusion.cpp
-STABLEDIFFUSION_GGML_VERSION?=e31a86ce9110b11a98bd5990c329093244c2d1e3
+STABLEDIFFUSION_GGML_VERSION?=db99efdd6d2a43c7937fd55b3359206c680a75b0
 
 CMAKE_ARGS+=-DGGML_MAX_NAME=128
 

backend/python/diffusers/backend.py

@@ -883,6 +883,34 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
 
         return backend_pb2.Result(message="Media generated", success=True)
 
+    def UpscaleImage(self, request, context):
+        try:
+            if not request.src:
+                return backend_pb2.Result(success=False, message="No source image provided")
+            if not request.dst:
+                return backend_pb2.Result(success=False, message="No destination path provided")
+
+            scale = request.scale if request.scale > 0 else 2
+            image = Image.open(request.src).convert("RGB")
+
+            # If the loaded pipeline supports upscaling (e.g. StableDiffusionUpscalePipeline),
+            # use it; otherwise fall back to high-quality Lanczos resize.
+            if self.pipe is not None and self.PipelineType in ("StableDiffusionUpscalePipeline", "StableDiffusionLatentUpscalePipeline"):
+                print(f"UpscaleImage: using diffusers upscale pipeline ({self.PipelineType})", file=sys.stderr)
+                upscaled = self.pipe(prompt="", image=image).images[0]
+            else:
+                # Fallback: high-quality Lanczos resize
+                print(f"UpscaleImage: no upscale pipeline loaded, using Lanczos resize (scale={scale})", file=sys.stderr)
+                new_w = image.width * scale
+                new_h = image.height * scale
+                upscaled = image.resize((new_w, new_h), Image.LANCZOS)
+
+            upscaled.save(request.dst)
+            return backend_pb2.Result(message="Image upscaled", success=True)
+        except Exception as e:
+            print(f"UpscaleImage error: {e}", file=sys.stderr)
+            return backend_pb2.Result(success=False, message=str(e))
+
     def GenerateVideo(self, request, context):
         try:
             prompt = request.prompt

backend/python/sglang/requirements-cublas12-after.txt

@@ -15,3 +15,12 @@ sglang[all]>=0.5.11
 # load-bearing for flash-attn-4, and this is the narrower change. Raise the
 # bound once 0.46.0 final ships.
 nvidia-modelopt<0.46
+
+# Same failure mode as the nvidia-modelopt bound above, via a different
+# package. sglang -> flashinfer-python -> cuda-tile, unbounded, and the
+# global --prerelease=allow resolves it to 1.6.0rc3, whose build backend
+# imports wheel_stub without declaring it in build-system.requires. With
+# --no-build-isolation nothing installs it and the build dies with
+# "No module named 'wheel_stub'". 1.5.0 is the newest stable release.
+# Raise the bound once 1.6.0 final ships.
+cuda-tile<1.6

backend/python/sglang/requirements-cublas13-after.txt

@@ -15,3 +15,12 @@ sglang[all]>=0.5.11
 # load-bearing for flash-attn-4, and this is the narrower change. Raise the
 # bound once 0.46.0 final ships.
 nvidia-modelopt<0.46
+
+# Same failure mode as the nvidia-modelopt bound above, via a different
+# package. sglang -> flashinfer-python -> cuda-tile, unbounded, and the
+# global --prerelease=allow resolves it to 1.6.0rc3, whose build backend
+# imports wheel_stub without declaring it in build-system.requires. With
+# --no-build-isolation nothing installs it and the build dies with
+# "No module named 'wheel_stub'". 1.5.0 is the newest stable release.
+# Raise the bound once 1.6.0 final ships.
+cuda-tile<1.6

backend/python/sglang/requirements-l4t13-after.txt

@@ -13,3 +13,12 @@
 # FunctionCallParser, ReasoningParser); the [all] extras are optional
 # accelerators not required at import time.
 sglang>=0.5.11
+
+# Same failure mode the cublas profiles carry an nvidia-modelopt bound for,
+# reached through a different package. sglang -> flashinfer-python ->
+# cuda-tile, unbounded, and the global --prerelease=allow resolves it to
+# 1.6.0rc3, whose build backend imports wheel_stub without declaring it in
+# build-system.requires. With --no-build-isolation nothing installs it and
+# the build dies with "No module named 'wheel_stub'". 1.5.0 is the newest
+# stable release. Raise the bound once 1.6.0 final ships.
+cuda-tile<1.6

core/application/application.go

@@ -553,12 +553,17 @@ func (a *Application) start() error {
 	// once at startup and reused across chat sessions that opt in via metadata.
 	if !a.applicationConfig.DisableLocalAIAssistant {
 		holder := mcpTools.NewLocalAIAssistantHolder()
+		var nodeRegistry *nodes.NodeRegistry
+		if a.distributed != nil {
+			nodeRegistry = a.distributed.Registry
+		}
 		assistantClient := localaiInproc.New(
 			a.applicationConfig,
 			a.applicationConfig.SystemState,
 			a.backendLoader,
 			a.modelLoader,
 			a.galleryService,
+			nodeRegistry,
 		)
 		// Wire usage tracking so the assistant's get_usage_stats tool
 		// returns real data; nil values keep the tool returning a clear

core/backend/upscale.go

@@ -0,0 +1,37 @@
+package backend
+
+import (
+	"context"
+
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/pkg/grpc/proto"
+	model "github.com/mudler/LocalAI/pkg/model"
+)
+
+// ImageUpscale loads the model specified in modelConfig and calls UpscaleImage
+// on the backend, writing the result to dst.
+func ImageUpscale(ctx context.Context, src, dst string, scale int, loader *model.ModelLoader, modelConfig config.ModelConfig, appConfig *config.ApplicationConfig) (func() error, error) {
+	opts := ModelOptions(modelConfig, appConfig, model.WithContext(ctx))
+	inferenceModel, err := loader.Load(opts...)
+	if err != nil {
+		recordModelLoadFailure(appConfig, modelConfig.Name, modelConfig.Backend, err, nil)
+		return nil, err
+	}
+
+	fn := func() error {
+		_, err := inferenceModel.UpscaleImage(
+			ctx,
+			&proto.UpscaleImageRequest{
+				Src:   src,
+				Dst:   dst,
+				Scale: int32(scale),
+			},
+		)
+		return err
+	}
+
+	return fn, nil
+}
+
+// ImageUpscaleFunc is a test-friendly indirection.
+var ImageUpscaleFunc = ImageUpscale

core/config/backend_capabilities.go

@@ -44,6 +44,7 @@ const (
 	MethodPredictStream      GRPCMethod = "PredictStream"
 	MethodEmbedding          GRPCMethod = "Embedding"
 	MethodGenerateImage      GRPCMethod = "GenerateImage"
+	MethodUpscaleImage       GRPCMethod = "UpscaleImage"
 	MethodGenerateVideo      GRPCMethod = "GenerateVideo"
 	MethodGenerate3D         GRPCMethod = "Generate3D"
 	MethodAudioTranscription GRPCMethod = "AudioTranscription"
@@ -348,7 +349,7 @@ var BackendCapabilities = map[string]BackendCapability{
 
 	// --- Image/video generation backends ---
 	"diffusers": {
-		GRPCMethods:      []GRPCMethod{MethodGenerateImage, MethodGenerateVideo},
+		GRPCMethods:      []GRPCMethod{MethodGenerateImage, MethodUpscaleImage, MethodGenerateVideo},
 		PossibleUsecases: []string{UsecaseImage, UsecaseVideo},
 		DefaultUsecases:  []string{UsecaseImage},
 		Description:      "HuggingFace diffusers — Stable Diffusion, Flux, video generation",

core/config/meta/constants.go

@@ -38,6 +38,14 @@ var CacheTypeOptions = []FieldOption{
 	{Value: "q4_1", Label: "Q4_1"},
 	{Value: "q5_0", Label: "Q5_0"},
 	{Value: "q5_1", Label: "Q5_1"},
+	// TurboQuant KV-cache types — accepted by the turboquant and
+	// buun-llama-cpp fork backends; stock llama-cpp will reject them at load.
+	{Value: "turbo2", Label: "Turbo2 (TurboQuant)"},
+	{Value: "turbo3", Label: "Turbo3 (TurboQuant)"},
+	{Value: "turbo4", Label: "Turbo4 (TurboQuant)"},
+	// Trellis-Coded Quantization variants — buun-llama-cpp only.
+	{Value: "turbo2_tcq", Label: "Turbo2 TCQ (buun-llama-cpp)"},
+	{Value: "turbo3_tcq", Label: "Turbo3 TCQ (buun-llama-cpp)"},
 }
 
 var DiffusersPipelineOptions = []FieldOption{

core/gallery/importers/llama-cpp.go

@@ -38,6 +38,7 @@ func (i *LlamaCPPImporter) AdditionalBackends() []KnownBackendEntry {
 		{Name: "ik-llama-cpp", Modality: "text", Description: "GGUF drop-in replacement for llama-cpp with ik-quants"},
 		{Name: "turboquant", Modality: "text", Description: "GGUF drop-in replacement for llama-cpp with TurboQuant optimizations"},
 		{Name: "vllm-cpp", Modality: "text", Description: "vLLM-style continuous-batching engine (vllm.cpp) consuming GGUF, by the LocalAI team"},
+		{Name: "buun-llama-cpp", Modality: "text", Description: "GGUF drop-in replacement for llama-cpp with DFlash speculative decoding and TurboQuant/TCQ KV-cache quantization"},
 	}
 }
 
@@ -136,7 +137,7 @@ func (i *LlamaCPPImporter) Import(details Details) (gallery.ModelConfig, error)
 	backend := "llama-cpp"
 	if b, ok := preferencesMap["backend"].(string); ok {
 		switch b {
-		case "ik-llama-cpp", "turboquant", "vllm-cpp":
+		case "ik-llama-cpp", "turboquant", "vllm-cpp", "buun-llama-cpp":
 			backend = b
 		}
 	}

core/gallery/importers/llama-cpp_test.go

@@ -203,6 +203,23 @@ var _ = Describe("LlamaCPPImporter", func() {
 			Expect(modelConfig.Files[0].Filename).To(Equal("my-model.gguf"))
 		})
 
+		It("swaps the emitted backend to buun-llama-cpp when preferred", func() {
+			preferences := json.RawMessage(`{"backend": "buun-llama-cpp"}`)
+			details := Details{
+				URI:         "https://example.com/my-model.gguf",
+				Preferences: preferences,
+			}
+
+			modelConfig, err := importer.Import(details)
+
+			Expect(err).ToNot(HaveOccurred())
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("backend: buun-llama-cpp"), fmt.Sprintf("Model config: %+v", modelConfig))
+			Expect(modelConfig.ConfigFile).NotTo(ContainSubstring("backend: llama-cpp\n"), fmt.Sprintf("Model config: %+v", modelConfig))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("model: my-model.gguf"), fmt.Sprintf("Model config: %+v", modelConfig))
+			Expect(len(modelConfig.Files)).To(Equal(1))
+			Expect(modelConfig.Files[0].Filename).To(Equal("my-model.gguf"))
+		})
+
 		It("keeps backend: llama-cpp for unknown backend preferences", func() {
 			// Unknown backend values must not leak into the emitted YAML —
 			// we only honour the curated drop-in replacements.
@@ -551,7 +568,7 @@ var _ = Describe("LlamaCPPImporter", func() {
 	})
 
 	Context("AdditionalBackends", func() {
-		It("advertises ik-llama-cpp, turboquant and vllm-cpp as drop-in replacements", func() {
+		It("advertises all llama-cpp drop-in replacements", func() {
 			entries := importer.AdditionalBackends()
 
 			names := make([]string, 0, len(entries))
@@ -560,7 +577,7 @@ var _ = Describe("LlamaCPPImporter", func() {
 				names = append(names, e.Name)
 				byName[e.Name] = e
 			}
-			Expect(names).To(ConsistOf("ik-llama-cpp", "turboquant", "vllm-cpp"))
+			Expect(names).To(ConsistOf("ik-llama-cpp", "turboquant", "vllm-cpp", "buun-llama-cpp"))
 
 			for _, name := range names {
 				e := byName[name]

core/http/auth/features.go

@@ -39,6 +39,8 @@ var RouteFeatureRegistry = []RouteFeature{
 	{"POST", "/images/generations", FeatureImages},
 	{"POST", "/v1/images/inpainting", FeatureImages},
 	{"POST", "/images/inpainting", FeatureImages},
+	{"POST", "/v1/images/upscale", FeatureImages},
+	{"POST", "/images/upscale", FeatureImages},
 
 	// Audio transcription
 	{"POST", "/v1/audio/transcriptions", FeatureAudioTranscription},
@@ -116,6 +118,10 @@ var RouteFeatureRegistry = []RouteFeature{
 	// Rerank
 	{"POST", "/v1/rerank", FeatureRerank},
 
+	// Moderation
+	{"POST", "/v1/moderations", FeatureModeration},
+	{"POST", "/moderations", FeatureModeration},
+
 	// Stores
 	{"POST", "/stores/set", FeatureStores},
 	{"POST", "/stores/delete", FeatureStores},
@@ -191,6 +197,7 @@ func APIFeatureMetas() []FeatureMeta {
 		{FeatureEmbeddings, "Embeddings", true},
 		{FeatureSound, "Sound Generation", true},
 		{FeatureRealtime, "Realtime", true},
+		{FeatureModeration, "Moderation", true},
 		{FeatureRerank, "Rerank", true},
 		{FeatureTokenize, "Tokenize", true},
 		{FeatureMCP, "MCP", true},

core/http/auth/features_moderation_test.go

@@ -0,0 +1,24 @@
+package auth_test
+
+import (
+	. "github.com/mudler/LocalAI/core/http/auth"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("Moderation feature registration", func() {
+	It("registers both moderation routes as default-on API features", func() {
+		Expect(APIFeatures).To(ContainElement(FeatureModeration))
+
+		patterns := []string{}
+		for _, route := range RouteFeatureRegistry {
+			if route.Feature == FeatureModeration {
+				patterns = append(patterns, route.Pattern)
+			}
+		}
+		Expect(patterns).To(ConsistOf("/v1/moderations", "/moderations"))
+
+		metas := APIFeatureMetas()
+		Expect(metas).To(ContainElement(FeatureMeta{Key: FeatureModeration, Label: "Moderation", DefaultValue: true}))
+	})
+})

core/http/auth/helpers_test.go

@@ -59,10 +59,14 @@ func ok(c echo.Context) error {
 func newAuthTestApp(db *gorm.DB, appConfig *config.ApplicationConfig) *echo.Echo {
 	e := echo.New()
 	e.Use(auth.Middleware(db, appConfig))
+	if db != nil {
+		e.Use(auth.RequireRouteFeature(db))
+	}
 
 	// API routes (require auth)
 	e.GET("/v1/models", ok)
 	e.POST("/v1/chat/completions", ok)
+	e.POST("/v1/moderations", ok)
 	e.GET("/api/settings", ok)
 	e.POST("/api/settings", ok)
 
@@ -81,10 +85,14 @@ func newAuthTestApp(db *gorm.DB, appConfig *config.ApplicationConfig) *echo.Echo
 func newAdminTestApp(db *gorm.DB, appConfig *config.ApplicationConfig) *echo.Echo {
 	e := echo.New()
 	e.Use(auth.Middleware(db, appConfig))
+	if db != nil {
+		e.Use(auth.RequireRouteFeature(db))
+	}
 
 	// Regular routes
 	e.GET("/v1/models", ok)
 	e.POST("/v1/chat/completions", ok)
+	e.POST("/v1/moderations", ok)
 
 	// Admin-only routes
 	adminMw := auth.RequireAdmin()

core/http/auth/middleware_test.go

@@ -91,6 +91,19 @@ var _ = Describe("Auth Middleware", func() {
 			Expect(rec.Code).To(Equal(http.StatusOK))
 		})
 
+		It("allows authenticated users to call moderation by default", func() {
+			sessionID := createTestSession(db, user.ID)
+			rec := doRequest(app, http.MethodPost, "/v1/moderations", withSessionCookie(sessionID))
+			Expect(rec.Code).To(Equal(http.StatusOK))
+		})
+
+		It("blocks moderation when the user's feature is disabled", func() {
+			Expect(auth.UpdateUserPermissions(db, user.ID, auth.PermissionMap{auth.FeatureModeration: false})).To(Succeed())
+			sessionID := createTestSession(db, user.ID)
+			rec := doRequest(app, http.MethodPost, "/v1/moderations", withSessionCookie(sessionID))
+			Expect(rec.Code).To(Equal(http.StatusForbidden))
+		})
+
 		It("allows requests with valid session as Bearer token", func() {
 			sessionID := createTestSession(db, user.ID)
 			rec := doRequest(app, http.MethodGet, "/v1/models", withBearerToken(sessionID))
@@ -156,6 +169,11 @@ var _ = Describe("Auth Middleware", func() {
 			Expect(rec.Code).To(Equal(http.StatusUnauthorized))
 		})
 
+		It("returns 401 for unauthenticated moderation requests", func() {
+			rec := doRequest(app, http.MethodPost, "/v1/moderations")
+			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))

core/http/auth/permissions.go

@@ -51,6 +51,7 @@ const (
 	FeatureEmbeddings          = "embeddings"
 	FeatureSound               = "sound"
 	FeatureRealtime            = "realtime"
+	FeatureModeration          = "moderation"
 	FeatureRerank              = "rerank"
 	FeatureTokenize            = "tokenize"
 	FeatureMCP                 = "mcp"
@@ -75,7 +76,7 @@ var APIFeatures = []string{
 	FeatureChat, FeatureImages, FeatureAudioSpeech, FeatureAudioTranscription,
 	FeatureAudioDiarization, FeatureAudioClassification,
 	FeatureVAD, FeatureDetection, FeatureVideo, Feature3D, FeatureEmbeddings, FeatureSound,
-	FeatureRealtime, FeatureRerank, FeatureTokenize, FeatureMCP, FeatureStores,
+	FeatureRealtime, FeatureModeration, FeatureRerank, FeatureTokenize, FeatureMCP, FeatureStores,
 	FeatureFaceRecognition, FeatureVoiceRecognition, FeatureAudioTransform,
 	FeaturePIIFilter,
 }

core/http/endpoints/localai/api_instructions.go

@@ -30,6 +30,12 @@ var instructionDefs = []instructionDef{
 		Tags:        []string{"inference", "embeddings"},
 		Intro:       "Set \"stream\": true for SSE streaming. Supports tool/function calling when the model config has function templates configured.",
 	},
+	{
+		Name:        "moderation",
+		Description: "OpenAI-compatible text moderation using a local completion model",
+		Tags:        []string{"moderation"},
+		Intro:       "POST /v1/moderations accepts a text string or array plus a LocalAI completion model. LocalAI constrains the model to the OpenAI moderation category schema and returns one result per input. Multimodal moderation inputs are not yet supported.",
+	},
 	{
 		Name:        "audio",
 		Description: "Text-to-speech, voice activity detection, transcription, speaker diarization, sound classification, and sound generation",

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(18))
+			Expect(instructions).To(HaveLen(19))
 
 			// Verify each instruction has required fields and correct URL format
 			for _, s := range instructions {
@@ -69,6 +69,7 @@ var _ = Describe("API Instructions Endpoints", func() {
 
 			Expect(names).To(ContainElements(
 				"chat-inference",
+				"moderation",
 				"config-management",
 				"model-management",
 				"monitoring",

core/http/endpoints/localai/system.go

@@ -12,7 +12,7 @@ import (
 // @Tags monitoring
 // @Success 200 {object} schema.SystemInformationResponse "Response"
 // @Router /system [get]
-func SystemInformations(ml *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+func SystemInformations(cl *config.ModelConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		availableBackends := []string{}
 		loadedModels := ml.ListLoadedModels()
@@ -25,7 +25,14 @@ func SystemInformations(ml *model.ModelLoader, appConfig *config.ApplicationConf
 
 		sysmodels := []schema.SysInfoModel{}
 		for _, m := range loadedModels {
-			sysmodels = append(sysmodels, schema.SysInfoModel{ID: m.ID})
+			entry := schema.SysInfoModel{ID: m.ID}
+			// The loader tracks only the ID. Which engine is serving a model is
+			// the first thing an operator wants beside its name, and it is one
+			// config lookup away.
+			if cfg, ok := cl.GetModelConfig(m.ID); ok {
+				entry.Backend = cfg.Backend
+			}
+			sysmodels = append(sysmodels, entry)
 		}
 		return c.JSON(200,
 			schema.SystemInformationResponse{

core/http/endpoints/localai/traces.go

@@ -3,6 +3,7 @@ package localai
 import (
 	"net/http"
 	"strconv"
+	"time"
 
 	"github.com/labstack/echo/v4"
 	"github.com/mudler/LocalAI/core/http/middleware"
@@ -85,6 +86,35 @@ func GetAPITracesEndpoint() echo.HandlerFunc {
 	}
 }
 
+// GetAPITracesSummaryEndpoint returns counted totals over a recent window
+// @Summary Summarize recent API traces
+// @Description Returns request, failure and latency totals over a recent window, plus a bucketed series for sparklines. Exists so callers wanting three numbers do not have to fetch the whole trace list and count it themselves.
+// @Tags monitoring
+// @Produce json
+// @Param hours query int false "Window in hours (default 24, max 168)"
+// @Success 200 {object} middleware.TraceSummary "Counted trace totals"
+// @Router /api/traces/summary [get]
+func GetAPITracesSummaryEndpoint() echo.HandlerFunc {
+	return func(c echo.Context) error {
+		hours := 24
+		if raw := c.QueryParam("hours"); raw != "" {
+			if v, err := strconv.Atoi(raw); err == nil && v > 0 {
+				hours = v
+			}
+		}
+		// A week is plenty for a dashboard, and the trace buffer is bounded
+		// anyway; an unbounded window would just scan the whole buffer.
+		if hours > 168 {
+			hours = 168
+		}
+		return c.JSON(http.StatusOK, middleware.GetTracesSummary(time.Duration(hours)*time.Hour, traceSummaryBuckets))
+	}
+}
+
+// Enough columns for a sparkline to show a shape, few enough that each one
+// still holds a meaningful count on a quiet installation.
+const traceSummaryBuckets = 12
+
 // GetAPITraceEndpoint returns a single API trace with its full payload
 // @Summary Get one API trace
 // @Description Returns a single captured API exchange, including the request and response bodies omitted from the list response

core/http/endpoints/mcp/localai_assistant_test.go

@@ -84,6 +84,22 @@ func (stubClient) ListNodes(_ context.Context) ([]localaitools.Node, error) {
 	return []localaitools.Node{}, nil
 }
 
+func (stubClient) ListScheduling(_ context.Context) ([]localaitools.ModelSchedulingConfig, error) {
+	return []localaitools.ModelSchedulingConfig{}, nil
+}
+
+func (stubClient) GetScheduling(_ context.Context, _ string) (*localaitools.ModelSchedulingConfig, error) {
+	return &localaitools.ModelSchedulingConfig{}, nil
+}
+
+func (stubClient) SetScheduling(_ context.Context, _ localaitools.SetSchedulingRequest) (*localaitools.ModelSchedulingConfig, error) {
+	return &localaitools.ModelSchedulingConfig{}, nil
+}
+
+func (stubClient) DeleteScheduling(_ context.Context, _ string) error {
+	return nil
+}
+
 func (stubClient) SetNodeVRAMBudget(_ context.Context, _, _ string) error {
 	return nil
 }

core/http/endpoints/openai/moderations.go

@@ -0,0 +1,190 @@
+package openai
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"math"
+	"net/http"
+	"strings"
+
+	"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/LocalAI/core/templates"
+	"github.com/mudler/LocalAI/pkg/functions"
+	"github.com/mudler/LocalAI/pkg/model"
+)
+
+var moderationCategories = []string{
+	"harassment",
+	"harassment/threatening",
+	"hate",
+	"hate/threatening",
+	"illicit",
+	"illicit/violent",
+	"self-harm",
+	"self-harm/intent",
+	"self-harm/instructions",
+	"sexual",
+	"sexual/minors",
+	"violence",
+	"violence/graphic",
+}
+
+type moderationGenerator func(context.Context, string, *config.ModelConfig) (string, backend.TokenUsage, error)
+
+type generatedModeration struct {
+	Categories     map[string]bool    `json:"categories"`
+	CategoryScores map[string]float64 `json:"category_scores"`
+}
+
+// ModerationEndpoint implements the text input subset of OpenAI's moderation
+// API using any LocalAI completion model and constrained JSON generation.
+// @Summary Classify text for potentially harmful content.
+// @Tags moderation
+// @Param request body schema.ModerationRequest true "query params"
+// @Success 200 {object} schema.ModerationResponse "Response"
+// @Router /v1/moderations [post]
+func ModerationEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator *templates.Evaluator, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+	return moderationEndpoint(func(ctx context.Context, input string, cfg *config.ModelConfig) (string, backend.TokenUsage, error) {
+		prompt := moderationPrompt(input)
+		var messages schema.Messages
+		if cfg.TemplateConfig.UseTokenizerTemplate {
+			messages = schema.Messages{{Role: "user", Content: prompt}}
+			prompt = ""
+		} else if evaluator != nil {
+			if rendered, err := evaluator.EvaluateTemplateForPrompt(templates.CompletionPromptTemplate, *cfg, templates.PromptTemplateData{Input: prompt, SystemPrompt: cfg.SystemPrompt}); err == nil {
+				prompt = rendered
+			}
+		}
+
+		predict, err := backend.ModelInferenceFunc(ctx, prompt, messages, nil, nil, nil, ml, cfg, cl, appConfig, nil, "", "", nil, nil, nil, nil)
+		if err != nil {
+			return "", backend.TokenUsage{}, err
+		}
+		response, err := predict()
+		return response.Response, response.Usage, err
+	})
+}
+
+func moderationEndpoint(generate moderationGenerator) echo.HandlerFunc {
+	return func(c echo.Context) error {
+		input, ok := c.Get(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST).(*schema.ModerationRequest)
+		if !ok || input == nil {
+			return echo.NewHTTPError(http.StatusBadRequest, "invalid moderation request")
+		}
+		if len(input.Input) == 0 {
+			return echo.NewHTTPError(http.StatusBadRequest, "input must contain at least one text string")
+		}
+		if generate == nil {
+			return echo.NewHTTPError(http.StatusInternalServerError, "moderation generator is unavailable")
+		}
+
+		modelConfig, ok := c.Get(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG).(*config.ModelConfig)
+		if !ok || modelConfig == nil {
+			return echo.NewHTTPError(http.StatusBadRequest, "moderation model configuration is unavailable")
+		}
+
+		grammar, err := moderationGrammar()
+		if err != nil {
+			return echo.NewHTTPError(http.StatusInternalServerError, "failed to build moderation grammar").SetInternal(err)
+		}
+		cfg := *modelConfig
+		cfg.Grammar = grammar
+		maxTokens := 512
+		cfg.Maxtokens = &maxTokens
+
+		results := make([]schema.ModerationResult, 0, len(input.Input))
+		promptTokens, completionTokens := 0, 0
+		for _, text := range input.Input {
+			if strings.TrimSpace(text) == "" {
+				return echo.NewHTTPError(http.StatusBadRequest, "input strings must not be empty")
+			}
+			raw, usage, err := generate(c.Request().Context(), text, &cfg)
+			if err != nil {
+				return echo.NewHTTPError(http.StatusInternalServerError, "moderation inference failed").SetInternal(err)
+			}
+			result, err := parseModerationResult(raw)
+			if err != nil {
+				return echo.NewHTTPError(http.StatusInternalServerError, "invalid moderation result: "+err.Error()).SetInternal(err)
+			}
+			results = append(results, result)
+			promptTokens += usage.Prompt
+			completionTokens += usage.Completion
+		}
+
+		middleware.StampUsage(c, input.Model, promptTokens, completionTokens)
+		return c.JSON(http.StatusOK, schema.ModerationResponse{
+			ID:      "modr-" + uuid.NewString(),
+			Model:   input.Model,
+			Results: results,
+		})
+	}
+}
+
+func moderationPrompt(input string) string {
+	encoded, _ := json.Marshal(input)
+	return `Classify the supplied text for safety moderation. Return only the JSON object required by the grammar.
+For each category, set the boolean to true only when the text matches that category. Give each category a confidence score from 0 to 1.
+The categories are harassment, harassment/threatening, hate, hate/threatening, illicit, illicit/violent, self-harm, self-harm/intent, self-harm/instructions, sexual, sexual/minors, violence, and violence/graphic.
+Text to classify: ` + string(encoded)
+}
+
+func moderationGrammar() (string, error) {
+	boolProperties := map[string]any{}
+	scoreProperties := map[string]any{}
+	for _, category := range moderationCategories {
+		boolProperties[category] = map[string]any{"type": "boolean"}
+		scoreProperties[category] = map[string]any{"type": "number"}
+	}
+	structure := functions.JSONFunctionStructure{AnyOf: []functions.Item{{
+		Type: "object",
+		Properties: map[string]any{
+			"categories": map[string]any{
+				"type":                 "object",
+				"properties":           boolProperties,
+				"required":             moderationCategories,
+				"additionalProperties": false,
+			},
+			"category_scores": map[string]any{
+				"type":                 "object",
+				"properties":           scoreProperties,
+				"required":             moderationCategories,
+				"additionalProperties": false,
+			},
+		},
+	}}}
+	return structure.Grammar()
+}
+
+func parseModerationResult(raw string) (schema.ModerationResult, error) {
+	var generated generatedModeration
+	if err := json.Unmarshal([]byte(strings.TrimSpace(raw)), &generated); err != nil {
+		return schema.ModerationResult{}, err
+	}
+
+	result := schema.ModerationResult{
+		Categories:                make(map[string]bool, len(moderationCategories)),
+		CategoryScores:            make(map[string]float64, len(moderationCategories)),
+		CategoryAppliedInputTypes: make(map[string][]string, len(moderationCategories)),
+	}
+	for _, category := range moderationCategories {
+		flagged, exists := generated.Categories[category]
+		if !exists {
+			return schema.ModerationResult{}, fmt.Errorf("missing category %q", category)
+		}
+		score, exists := generated.CategoryScores[category]
+		if !exists || math.IsNaN(score) || math.IsInf(score, 0) || score < 0 || score > 1 {
+			return schema.ModerationResult{}, fmt.Errorf("category %q has an invalid score", category)
+		}
+		result.Categories[category] = flagged
+		result.CategoryScores[category] = score
+		result.CategoryAppliedInputTypes[category] = []string{"text"}
+		result.Flagged = result.Flagged || flagged
+	}
+	return result, nil
+}

core/http/endpoints/openai/moderations_test.go

@@ -0,0 +1,105 @@
+package openai
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+
+	"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/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("Moderations endpoint", func() {
+	It("classifies each text input and returns the OpenAI response shape", func() {
+		inputs := []string{}
+		generate := func(_ context.Context, input string, cfg *config.ModelConfig) (string, backend.TokenUsage, error) {
+			inputs = append(inputs, input)
+			Expect(cfg.Grammar).To(ContainSubstring("harassment"))
+			return `{
+				"categories":{"harassment":true,"harassment/threatening":false,"hate":false,"hate/threatening":false,"illicit":false,"illicit/violent":false,"self-harm":false,"self-harm/intent":false,"self-harm/instructions":false,"sexual":false,"sexual/minors":false,"violence":false,"violence/graphic":false},
+				"category_scores":{"harassment":0.9,"harassment/threatening":0.1,"hate":0,"hate/threatening":0,"illicit":0,"illicit/violent":0,"self-harm":0,"self-harm/intent":0,"self-harm/instructions":0,"sexual":0,"sexual/minors":0,"violence":0,"violence/graphic":0}
+			}`, backend.TokenUsage{Prompt: 12, Completion: 8}, nil
+		}
+
+		e := echo.New()
+		req := httptest.NewRequest(http.MethodPost, "/v1/moderations", strings.NewReader(`{"model":"guard","input":["first","second"]}`))
+		req.Header.Set(echo.HeaderContentType, echo.MIMEApplicationJSON)
+		rec := httptest.NewRecorder()
+		ctx := e.NewContext(req, rec)
+		ctx.Set(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST, &schema.ModerationRequest{
+			BasicModelRequest: schema.BasicModelRequest{Model: "guard"},
+			Input:             schema.ModerationInput{"first", "second"},
+		})
+		modelConfig := &config.ModelConfig{Name: "guard"}
+		modelConfig.Model = "guard.gguf"
+		ctx.Set(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG, modelConfig)
+
+		Expect(moderationEndpoint(generate)(ctx)).To(Succeed())
+		Expect(rec.Code).To(Equal(http.StatusOK))
+		Expect(inputs).To(Equal([]string{"first", "second"}))
+
+		var response schema.ModerationResponse
+		Expect(json.Unmarshal(rec.Body.Bytes(), &response)).To(Succeed())
+		Expect(response.ID).To(HavePrefix("modr-"))
+		Expect(response.Model).To(Equal("guard"))
+		Expect(response.Results).To(HaveLen(2))
+		Expect(response.Results[0].Flagged).To(BeTrue())
+		Expect(response.Results[0].Categories["harassment"]).To(BeTrue())
+		Expect(response.Results[0].CategoryAppliedInputTypes["harassment"]).To(Equal([]string{"text"}))
+	})
+
+	It("rejects an empty input list", func() {
+		e := echo.New()
+		ctx := e.NewContext(httptest.NewRequest(http.MethodPost, "/v1/moderations", nil), httptest.NewRecorder())
+		ctx.Set(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST, &schema.ModerationRequest{
+			BasicModelRequest: schema.BasicModelRequest{Model: "guard"},
+		})
+		ctx.Set(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG, &config.ModelConfig{Name: "guard"})
+
+		err := moderationEndpoint(nil)(ctx)
+		Expect(err).To(MatchError(ContainSubstring("input must contain at least one text string")))
+		Expect(err.(*echo.HTTPError).Code).To(Equal(http.StatusBadRequest))
+	})
+
+	It("surfaces malformed classifier output without returning a partial result", func() {
+		generate := func(context.Context, string, *config.ModelConfig) (string, backend.TokenUsage, error) {
+			return "not-json", backend.TokenUsage{}, nil
+		}
+		e := echo.New()
+		ctx := e.NewContext(httptest.NewRequest(http.MethodPost, "/v1/moderations", nil), httptest.NewRecorder())
+		ctx.Set(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST, &schema.ModerationRequest{
+			BasicModelRequest: schema.BasicModelRequest{Model: "guard"},
+			Input:             schema.ModerationInput{"text"},
+		})
+		ctx.Set(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG, &config.ModelConfig{Name: "guard"})
+
+		err := moderationEndpoint(generate)(ctx)
+		Expect(err).To(MatchError(ContainSubstring("invalid moderation result")))
+		Expect(err.(*echo.HTTPError).Code).To(Equal(http.StatusInternalServerError))
+	})
+})
+
+var _ = Describe("Moderation input", func() {
+	DescribeTable("accepts OpenAI text input forms",
+		func(body string, expected schema.ModerationInput) {
+			var req schema.ModerationRequest
+			Expect(json.Unmarshal([]byte(body), &req)).To(Succeed())
+			Expect(req.Input).To(Equal(expected))
+		},
+		Entry("single text", `{"input":"hello"}`, schema.ModerationInput{"hello"}),
+		Entry("text array", `{"input":["hello","world"]}`, schema.ModerationInput{"hello", "world"}),
+	)
+
+	It("rejects multimodal input in the text-only MVP", func() {
+		var req schema.ModerationRequest
+		err := json.Unmarshal([]byte(`{"input":[{"type":"image_url","image_url":{"url":"https://example.com/a.png"}}]}`), &req)
+		Expect(err).To(MatchError(ContainSubstring("text string or array of text strings")))
+	})
+})

core/http/endpoints/openai/upscale.go

@@ -0,0 +1,134 @@
+package openai
+
+import (
+	"fmt"
+	"io"
+	"net/http"
+	"net/url"
+	"os"
+	"path/filepath"
+	"strconv"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/labstack/echo/v4"
+	"github.com/mudler/xlog"
+
+	"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"
+)
+
+// UpscaleEndpoint handles POST /v1/images/upscale
+//
+// @Summary      Image upscaling
+// @Description  Upscale an image using a specified model (e.g. stable-diffusion-x4-upscaler). Accepts multipart/form-data.
+// @Tags         images
+// @Accept       multipart/form-data
+// @Produce      application/json
+// @Param        model   formData  string  true   "Upscaler model identifier (e.g. stable-diffusion-x4-upscaler)"
+// @Param        image   formData  file    true   "Input image file"
+// @Param        scale   formData  int     false  "Upscale factor: 2 or 4 (default 2)"
+// @Success      200 {object} schema.OpenAIResponse
+// @Failure      400 {object} map[string]string
+// @Failure      500 {object} map[string]string
+// @Router       /v1/images/upscale [post]
+func UpscaleEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) echo.HandlerFunc {
+	return func(c echo.Context) error {
+		modelName := c.FormValue("model")
+		scaleStr := c.FormValue("scale")
+
+		if modelName == "" {
+			xlog.Error("Upscale Endpoint - missing model")
+			return echo.NewHTTPError(http.StatusBadRequest, "missing model")
+		}
+
+		scale := 2
+		if scaleStr != "" {
+			v, err := strconv.Atoi(scaleStr)
+			if err != nil || (v != 2 && v != 4) {
+				return echo.NewHTTPError(http.StatusBadRequest, "scale must be 2 or 4")
+			}
+			scale = v
+		}
+
+		// Read uploaded image
+		imageFile, err := c.FormFile("image")
+		if err != nil {
+			xlog.Error("Upscale Endpoint - missing image file", "error", err)
+			return echo.NewHTTPError(http.StatusBadRequest, "missing image file")
+		}
+
+		imgSrc, err := imageFile.Open()
+		if err != nil {
+			return err
+		}
+		defer imgSrc.Close()
+		imgBytes, err := io.ReadAll(imgSrc)
+		if err != nil {
+			return err
+		}
+
+		// Get model config from middleware context
+		cfg, ok := c.Get(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG).(*config.ModelConfig)
+		if !ok || cfg == nil {
+			xlog.Error("Upscale Endpoint - model config not found in context")
+			return echo.ErrBadRequest
+		}
+
+		tmpDir := filepath.Join(appConfig.GeneratedContentDir, "images")
+		if err := os.MkdirAll(tmpDir, 0750); err != nil {
+			return echo.NewHTTPError(http.StatusInternalServerError, "failed to prepare storage")
+		}
+
+		// Write input image to a temp file
+		srcTmp, err := os.CreateTemp(tmpDir, "upscale_src_")
+		if err != nil {
+			return err
+		}
+		if _, err := srcTmp.Write(imgBytes); err != nil {
+			_ = srcTmp.Close()
+			_ = os.Remove(srcTmp.Name())
+			return err
+		}
+		if err := srcTmp.Close(); err != nil {
+			xlog.Warn("Upscale Endpoint - failed to close src temp file", "error", err)
+		}
+		srcPath := srcTmp.Name()
+		defer os.Remove(srcPath)
+
+		// Prepare output file path
+		id := uuid.New().String()
+		dstPath := filepath.Join(tmpDir, fmt.Sprintf("upscale_%s.png", id))
+
+		fn, err := backend.ImageUpscaleFunc(c.Request().Context(), srcPath, dstPath, scale, ml, *cfg, appConfig)
+		if err != nil {
+			return err
+		}
+		if err := fn(); err != nil {
+			_ = os.Remove(dstPath)
+			return err
+		}
+
+		baseURL := middleware.BaseURL(c)
+		imgURL, err := url.JoinPath(baseURL, "generated-images", filepath.Base(dstPath))
+		if err != nil {
+			_ = os.Remove(dstPath)
+			return err
+		}
+
+		created := int(time.Now().Unix())
+		resp := &schema.OpenAIResponse{
+			ID:      id,
+			Created: created,
+			Data:    []schema.Item{{URL: imgURL}},
+			Usage: &schema.OpenAIUsage{
+				InputTokensDetails: &schema.InputTokensDetails{},
+			},
+		}
+
+		return c.JSON(http.StatusOK, resp)
+	}
+}

core/http/endpoints/openai/upscale_test.go

@@ -0,0 +1,89 @@
+package openai
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+
+	"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"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("Image upscaling", func() {
+	var (
+		appConfig *config.ApplicationConfig
+		tmpDir    string
+	)
+
+	BeforeEach(func() {
+		var err error
+		tmpDir, err = os.MkdirTemp("", "upscale")
+		Expect(err).ToNot(HaveOccurred())
+		appConfig = config.NewApplicationConfig(config.WithGeneratedContentDir(tmpDir))
+	})
+
+	AfterEach(func() {
+		Expect(os.RemoveAll(tmpDir)).To(Succeed())
+	})
+
+	It("stores the result in the directory served by /generated-images", func() {
+		original := backend.ImageUpscaleFunc
+		backend.ImageUpscaleFunc = func(_ context.Context, _, dst string, scale int, _ *model.ModelLoader, _ config.ModelConfig, _ *config.ApplicationConfig) (func() error, error) {
+			Expect(scale).To(Equal(4))
+			return func() error {
+				return os.WriteFile(dst, []byte("PNGDATA"), 0o644)
+			}, nil
+		}
+		DeferCleanup(func() { backend.ImageUpscaleFunc = original })
+
+		req, _ := makeMultipartRequest(
+			map[string]string{"model": "stable-diffusion-x4-upscaler", "scale": "4"},
+			map[string][]byte{"image": []byte("IMAGEDATA")},
+		)
+		rec := httptest.NewRecorder()
+		ctx := echo.New().NewContext(req, rec)
+		ctx.Set(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG, &config.ModelConfig{Backend: "diffusers"})
+
+		Expect(UpscaleEndpoint(nil, nil, appConfig)(ctx)).To(Succeed())
+		Expect(rec.Code).To(Equal(http.StatusOK))
+
+		var response schema.OpenAIResponse
+		Expect(json.Unmarshal(rec.Body.Bytes(), &response)).To(Succeed())
+		Expect(response.Data).To(HaveLen(1))
+		Expect(response.Data[0].URL).To(ContainSubstring("/generated-images/upscale_"))
+
+		filename := filepath.Base(response.Data[0].URL)
+		contents, err := os.ReadFile(filepath.Join(tmpDir, "images", filename))
+		Expect(err).ToNot(HaveOccurred())
+		Expect(contents).To(Equal([]byte("PNGDATA")))
+	})
+
+	It("rejects unsupported scale factors", func() {
+		req, _ := makeMultipartRequest(
+			map[string]string{"model": "stable-diffusion-x4-upscaler", "scale": "3"},
+			map[string][]byte{"image": []byte("IMAGEDATA")},
+		)
+		rec := httptest.NewRecorder()
+		ctx := echo.New().NewContext(req, rec)
+		ctx.Set(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG, &config.ModelConfig{Backend: "diffusers"})
+
+		err := UpscaleEndpoint(nil, nil, appConfig)(ctx)
+		var httpErr *echo.HTTPError
+		Expect(err).To(MatchError(ContainSubstring("scale must be 2 or 4")))
+		Expect(err).To(BeAssignableToTypeOf(httpErr))
+		httpErr = err.(*echo.HTTPError)
+		Expect(httpErr.Code).To(Equal(http.StatusBadRequest))
+		Expect(httpErr.Message).To(Equal("scale must be 2 or 4"))
+		Expect(bytes.TrimSpace(rec.Body.Bytes())).To(BeEmpty())
+	})
+})

core/http/middleware/trace_summary.go

@@ -0,0 +1,109 @@
+// SPDX-License-Identifier: MIT
+
+package middleware
+
+import (
+	"math"
+	"slices"
+	"time"
+)
+
+// TraceSummary is the counted view of the trace buffer.
+//
+// It exists so a caller that wants "how many, how many failed, how slow" does
+// not have to fetch every exchange and count them in the browser. The Operate
+// overview needs exactly those three numbers, and the trace list is capped in
+// the thousands, so shipping it across the wire to produce a single integer is
+// waste that grows with the buffer.
+type TraceSummary struct {
+	Total       int           `json:"total"`
+	Errors      int           `json:"errors"`
+	P95Millis   int64         `json:"p95_ms"`
+	WindowHours int           `json:"window_hours"`
+	Buckets     []TraceBucket `json:"buckets"`
+}
+
+// TraceBucket is one column of a sparkline: oldest first, so the series reads
+// left to right the way a chart is drawn.
+type TraceBucket struct {
+	Start  time.Time `json:"start"`
+	Count  int       `json:"count"`
+	Errors int       `json:"errors"`
+}
+
+// GetTracesSummary counts the buffered exchanges over the given window.
+func GetTracesSummary(window time.Duration, buckets int) TraceSummary {
+	return summarize(GetTraces(), window, buckets)
+}
+
+func summarize(traces []APIExchange, window time.Duration, buckets int) TraceSummary {
+	if buckets < 1 {
+		buckets = 1
+	}
+	now := time.Now()
+	cutoff := now.Add(-window)
+
+	summary := TraceSummary{
+		WindowHours: int(window.Hours()),
+		// Never nil: a nil slice serialises as null and breaks .map() on the
+		// other side, which is a silent runtime error rather than an empty chart.
+		Buckets: make([]TraceBucket, buckets),
+	}
+
+	bucketWidth := window / time.Duration(buckets)
+	for i := range summary.Buckets {
+		summary.Buckets[i].Start = cutoff.Add(time.Duration(i) * bucketWidth)
+	}
+
+	durations := make([]time.Duration, 0, len(traces))
+	for _, t := range traces {
+		if t.Timestamp.Before(cutoff) {
+			continue
+		}
+		summary.Total++
+		failed := isFailure(t)
+		if failed {
+			summary.Errors++
+		}
+		durations = append(durations, t.Duration)
+
+		// Clamp rather than skip: a request timestamped a hair in the future
+		// (clock skew, or arriving mid-call) still belongs in the newest column.
+		idx := int(t.Timestamp.Sub(cutoff) / bucketWidth)
+		if idx >= buckets {
+			idx = buckets - 1
+		}
+		if idx < 0 {
+			idx = 0
+		}
+		summary.Buckets[idx].Count++
+		if failed {
+			summary.Buckets[idx].Errors++
+		}
+	}
+
+	summary.P95Millis = percentileMillis(durations, 0.95)
+	return summary
+}
+
+// A 4xx is the caller getting it wrong, which is not the installation being
+// unhealthy. Only 5xx and a transport-level error count against the runtime.
+func isFailure(t APIExchange) bool {
+	return t.Error != "" || t.Response.Status >= 500
+}
+
+func percentileMillis(durations []time.Duration, p float64) int64 {
+	if len(durations) == 0 {
+		return 0
+	}
+	slices.Sort(durations)
+	// Nearest-rank: the smallest value at or above the pth percentile.
+	rank := int(math.Ceil(p*float64(len(durations)))) - 1
+	if rank < 0 {
+		rank = 0
+	}
+	if rank >= len(durations) {
+		rank = len(durations) - 1
+	}
+	return durations[rank].Milliseconds()
+}

core/http/middleware/trace_summary_test.go

@@ -0,0 +1,79 @@
+// SPDX-License-Identifier: MIT
+
+package middleware
+
+import (
+	"time"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("API trace summary", func() {
+	exchange := func(age time.Duration, status int, dur time.Duration) APIExchange {
+		return APIExchange{
+			Timestamp: time.Now().Add(-age),
+			Duration:  dur,
+			Response:  APIExchangeResponse{Status: status},
+		}
+	}
+
+	It("counts only what falls inside the window", func() {
+		traces := []APIExchange{
+			exchange(1*time.Hour, 200, 10*time.Millisecond),
+			exchange(2*time.Hour, 200, 10*time.Millisecond),
+			// Older than the window: must not be counted at all.
+			exchange(48*time.Hour, 500, 10*time.Millisecond),
+		}
+		s := summarize(traces, 24*time.Hour, 6)
+		Expect(s.Total).To(Equal(2))
+		Expect(s.Errors).To(BeZero())
+	})
+
+	It("treats 5xx and a transport error as failures, but not 4xx", func() {
+		traces := []APIExchange{
+			exchange(time.Minute, 500, time.Millisecond),
+			exchange(time.Minute, 503, time.Millisecond),
+			// A client sending a bad request is not the server failing.
+			exchange(time.Minute, 404, time.Millisecond),
+			exchange(time.Minute, 200, time.Millisecond),
+		}
+		traces[3].Error = "connection reset"
+
+		s := summarize(traces, 24*time.Hour, 6)
+		Expect(s.Total).To(Equal(4))
+		Expect(s.Errors).To(Equal(3))
+	})
+
+	It("reports p95 as a real percentile rather than the slowest request", func() {
+		traces := make([]APIExchange, 0, 100)
+		for i := 1; i <= 100; i++ {
+			traces = append(traces, exchange(time.Minute, 200, time.Duration(i)*time.Millisecond))
+		}
+		s := summarize(traces, 24*time.Hour, 6)
+		// 95th of 1..100ms, not the 100ms max.
+		Expect(s.P95Millis).To(BeNumerically("~", 95, 1))
+	})
+
+	It("buckets oldest-first so a sparkline reads left to right", func() {
+		traces := []APIExchange{
+			exchange(30*time.Minute, 200, time.Millisecond),
+			exchange(30*time.Minute, 200, time.Millisecond),
+			exchange(5*time.Hour, 200, time.Millisecond),
+		}
+		s := summarize(traces, 6*time.Hour, 6)
+		Expect(s.Buckets).To(HaveLen(6))
+		Expect(s.Buckets[0].Count).To(Equal(1), "the 5h-old request lands in the first bucket")
+		Expect(s.Buckets[5].Count).To(Equal(2), "the recent pair lands in the last")
+	})
+
+	It("returns an empty, non-nil summary when nothing has been traced", func() {
+		s := summarize(nil, 24*time.Hour, 6)
+		Expect(s.Total).To(BeZero())
+		Expect(s.Errors).To(BeZero())
+		Expect(s.P95Millis).To(BeZero())
+		// A nil slice serialises as null and breaks .map() in the browser.
+		Expect(s.Buckets).NotTo(BeNil())
+		Expect(s.Buckets).To(HaveLen(6))
+	})
+})

core/http/react-ui/e2e/activity-page.spec.js

@@ -43,6 +43,40 @@ test('lists live operations and cancels one from a labelled button', async ({ pa
   expect(cancelledPath).toBe('/api/operations/job-gemma/cancel')
 })
 
+test('pauses a model download without invoking destructive cancel', async ({ page }) => {
+  await stub(page, {
+    operations: [{
+      id: 'gemma-3-27b-it',
+      name: 'gemma-3-27b-it',
+      jobID: 'job-gemma',
+      progress: 22,
+      taskType: 'installation',
+      isBackend: false,
+      isQueued: false,
+      isDeletion: false,
+      cancellable: true,
+      phase: 'downloading',
+    }],
+  })
+
+  const requests = []
+  await page.route('**/api/operations/job-gemma/pause', (route) => {
+    requests.push(new URL(route.request().url()).pathname)
+    return route.fulfill({ contentType: 'application/json', body: '{}' })
+  })
+  await page.route('**/api/operations/job-gemma/cancel', (route) => {
+    requests.push(new URL(route.request().url()).pathname)
+    return route.fulfill({ contentType: 'application/json', body: '{}' })
+  })
+
+  await page.goto('/app/activity')
+
+  const card = page.locator('.operation-card').filter({ hasText: 'gemma-3-27b-it' })
+  await card.locator('.operation-card__pause').click()
+
+  await expect.poll(() => requests).toEqual(['/api/operations/job-gemma/pause'])
+})
+
 test('separates an unacknowledged failure from the record', async ({ page }) => {
   await stub(page, {
     operations: [{

core/http/react-ui/e2e/admin-console.spec.js

@@ -5,7 +5,9 @@ test.describe('Admin console', () => {
     await page.goto('/app/backends')
     const rail = page.locator('.console-rail')
     await expect(rail).toBeVisible()
-    for (const group of ['Inference', 'Cluster', 'Observability', 'Access', 'System']) {
+    // Four groups since the overview landed: Inference folded into Runtime
+    // (both are "the runtime right now"), Access and System into Administration.
+    for (const group of ['Runtime', 'Cluster', 'Observability', 'Administration']) {
       await expect(rail.locator('.console-group-title', { hasText: group })).toBeVisible()
     }
   })

core/http/react-ui/e2e/backends-notice.spec.js

@@ -0,0 +1,23 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// A notice is a hairline with a coloured left edge, not a filled panel. A tint
+// makes every notice shout at the weight of an error, which is how notices stop
+// being read — and it is the same treatment the Operate overview uses for the
+// rows that want a decision.
+
+test('the backends notice is an edge, not a filled card', async ({ page }) => {
+  // The upgrade banner is the notice worth pinning, so make one exist.
+  await page.route('**/api/backends/upgrades', route => route.fulfill({
+    json: { 'llama-cpp': { backend_name: 'llama-cpp', installed_version: '0.9.4', available_version: '0.9.7' } },
+  }))
+  await page.goto('/app/backends')
+  const notice = page.locator('.bk-notice', { hasText: /update/i }).first()
+  await expect(notice).toBeVisible()
+  const s = await notice.evaluate(el => {
+    const cs = getComputedStyle(el)
+    return { bg: cs.backgroundColor, left: parseFloat(cs.borderLeftWidth), top: parseFloat(cs.borderTopWidth) }
+  })
+  expect(s.bg).toMatch(/rgba\(0, 0, 0, 0\)|transparent/)
+  expect(s.left).toBeGreaterThanOrEqual(3)
+  expect(s.top).toBeLessThanOrEqual(1)
+})

core/http/react-ui/e2e/chat-transcript.spec.js

@@ -0,0 +1,55 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// Chat reads as a transcript rather than a bubble thread (mock 04).
+
+const CHAT = {
+  chats: [{
+    id: 'c1', name: 'Transcript', model: 'mock-model',
+    history: [
+      { role: 'user', content: 'Which backends do I have?' },
+      { role: 'assistant', content: 'Seven are installed.' },
+    ],
+  }],
+  activeChatId: 'c1',
+}
+
+test.describe('Chat transcript', () => {
+  test.beforeEach(async ({ page }) => {
+    await page.addInitScript(chat => {
+      localStorage.setItem('localai_chats_data', JSON.stringify(chat))
+    }, CHAT)
+    await page.goto('/app/chat')
+  })
+
+  test('neither role is a filled, rounded bubble', async ({ page }) => {
+    const user = page.locator('.chat-message-user .chat-message-content').first()
+    await expect(user).toBeVisible()
+    const cs = await user.evaluate(el => {
+      const s = getComputedStyle(el)
+      return { radius: s.borderTopLeftRadius, shadow: s.boxShadow }
+    })
+    // A rounded filled bubble carries the speaker in shape and side; a
+    // transcript carries it in words, which survives being read aloud.
+    expect(cs.radius).toBe('0px')
+    expect(cs.shadow).toBe('none')
+  })
+
+  test('both turns run full width in one column, not left and right', async ({ page }) => {
+    const user = page.locator('.chat-message-user').first()
+    const assistant = page.locator('.chat-message-assistant').first()
+    const [u, a] = [await user.boundingBox(), await assistant.boundingBox()]
+    expect(Math.abs(u.x - a.x)).toBeLessThan(2)
+  })
+
+  test('every turn says who is speaking', async ({ page }) => {
+    await expect(page.locator('.chat-message-user .chat-message-model')).toHaveText('You')
+    await expect(page.locator('.chat-message-assistant .chat-message-model').first())
+      .toHaveText('mock-model')
+  })
+
+  test('turns are separated by a rule', async ({ page }) => {
+    const border = await page.locator('.chat-message').first()
+      .evaluate(el => getComputedStyle(el).borderBottomStyle)
+    expect(border).toBe('solid')
+  })
+})

core/http/react-ui/e2e/chrome-audit.spec.js

@@ -0,0 +1,50 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// A standing guard against the two defects an earlier automated edit left
+// scattered through the pages: icons stripped of their fa-* class (which render
+// nothing at all), and controls left with the user agent's own chrome, which is
+// a pale grey button on a dark ground.
+const ROUTES = [
+  '/app', '/app/chat', '/app/models', '/app/studio', '/app/talk',
+  '/app/agents', '/app/skills', '/app/collections', '/app/agent-jobs',
+  '/app/fine-tune', '/app/quantize', '/app/face', '/app/voice',
+  '/app/manage', '/app/backends', '/app/activity', '/app/operate',
+  '/app/settings', '/app/traces', '/app/usage', '/app/nodes', '/app/p2p',
+  '/app/voice-library', '/app/voice-library/new', '/app/account',
+]
+
+test('no page renders a dead icon or a default-chrome control', async ({ page }) => {
+  // One test walks every route, so its budget has to scale with the list rather
+  // than sit on Playwright's per-test default of 30s. At 25 routes that default
+  // allows ~1.2s per navigation, which holds on a developer machine and does
+  // not on a loaded CI runner: the suite went red on the commit that added this
+  // spec, timing out mid-loop at waitForTimeout rather than at any single goto,
+  // which is what cumulative slowness looks like as opposed to one hung route.
+  // Six seconds a route absorbs a slow runner and still fails promptly if a
+  // route really does hang.
+  test.setTimeout(ROUTES.length * 6_000)
+
+  const findings = []
+  for (const route of ROUTES) {
+    await page.goto(route)
+    await page.waitForTimeout(400)
+    const found = await page.evaluate(() => {
+      const out = []
+      for (const el of document.querySelectorAll('button, a')) {
+        if (el.getBoundingClientRect().width === 0) continue
+        const cs = getComputedStyle(el)
+        if (cs.borderTopStyle === 'outset' || cs.backgroundColor === 'rgb(239, 239, 239)') {
+          out.push(`default-chrome: "${(el.textContent || '').trim().slice(0, 24)}" [${el.className}]`)
+        }
+      }
+      for (const i of document.querySelectorAll('i')) {
+        if (!/\bfa-/.test((i.className || '').toString())) {
+          out.push(`dead-icon: [${i.className}]`)
+        }
+      }
+      return [...new Set(out)]
+    })
+    for (const f of found) findings.push(`${route} — ${f}`)
+  }
+  expect(findings).toEqual([])
+})

core/http/react-ui/e2e/console-narrow.spec.js

@@ -0,0 +1,101 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// Small-screen behaviour of the Operate console and the dashboard stat cards.
+//
+// Both defects here are about a narrow viewport but neither is only a narrow
+// viewport problem: the stat cards were being laid out by the wrong rule at
+// every width, and the rail's height was never bounded.
+
+test.describe('Operate console on a narrow screen', () => {
+  test('expanding the rail leaves the page still on screen', async ({ page }) => {
+    await page.setViewportSize({ width: 390, height: 800 })
+    await page.goto('/app/manage')
+
+    const toggle = page.locator('.console-rail-toggle')
+    await expect(toggle).toBeVisible()
+    await toggle.click()
+    await expect(page.locator('.console-rail-groups')).toBeVisible()
+
+    // Thirteen destinations in one column is taller than a phone. If opening
+    // the menu pushes the page's own heading past the fold, the menu has
+    // replaced the page instead of annotating it.
+    // Manage titles itself with .view-bar__title rather than .page-title.
+    const heading = page.locator('.page-title, .view-bar__title').first()
+    const box = await heading.boundingBox()
+    expect(box).not.toBeNull()
+    expect(box.y).toBeLessThan(800)
+  })
+
+  test('the rail scrolls internally rather than growing without bound', async ({ page }) => {
+    await page.setViewportSize({ width: 390, height: 800 })
+    await page.goto('/app/manage')
+    await page.locator('.console-rail-toggle').click()
+
+    const groups = page.locator('.console-rail-groups')
+    await expect(groups).toBeVisible()
+    const height = await groups.evaluate(el => el.getBoundingClientRect().height)
+    expect(height).toBeLessThan(800)
+  })
+})
+
+test.describe('Headline figures', () => {
+  // Host used shadowed StatCards; it now shares the Operate overview's hairline
+  // figure strip, so the guard is that its labels stay legible, not that it
+  // keeps a card gap.
+  for (const width of [768, 1024]) {
+    test(`Host figure labels are not clipped at ${width}px`, async ({ page }) => {
+      await page.setViewportSize({ width, height: 1000 })
+      await page.goto('/app/manage')
+      const labels = page.locator('.stat-strip__label')
+      await expect(labels.first()).toBeVisible()
+      const clipped = await labels.evaluateAll(els =>
+        els.filter(el => el.scrollWidth > el.clientWidth + 1).map(el => el.textContent))
+      expect(clipped).toEqual([])
+    })
+  }
+
+  test('a Host figure routes into the thing it counts', async ({ page }) => {
+    await page.setViewportSize({ width: 1280, height: 1000 })
+    await page.goto('/app/manage')
+    const cell = page.locator('.stat-strip__cell').first()
+    await expect(cell).toBeVisible()
+    // A count is worth more when it is also the way to what it counted.
+    await expect(cell).toHaveJSProperty('tagName', 'BUTTON')
+  })
+
+  test('the figure strip keeps its height inside the flex column', async ({ page }) => {
+    // .page--app is a flex column whose split view takes flex:1, so a child
+    // with no intrinsic minimum gets shrunk to nothing. This strip did exactly
+    // that and rendered 2px tall with four invisible cells.
+    await page.setViewportSize({ width: 1440, height: 900 })
+    await page.goto('/app/manage')
+    const strip = page.locator('.manage-summary')
+    await expect(strip).toBeVisible()
+    const h = await strip.evaluate(el => el.getBoundingClientRect().height)
+    expect(h).toBeGreaterThan(40)
+  })
+})
+
+test.describe('Headline figure contrast', () => {
+  test('every figure is legible against the cell it sits on', async ({ page }) => {
+    // A <button> does not inherit colour, so a value with no tone rule fell
+    // back to the UA's `buttontext` — pure black on the dark ground, invisible.
+    await page.setViewportSize({ width: 1440, height: 950 })
+    await page.goto('/app/manage')
+    const bad = await page.locator('.stat-strip__value').evaluateAll(els => els
+      .map(el => ({ text: el.textContent, color: getComputedStyle(el).color }))
+      .filter(v => v.color === 'rgb(0, 0, 0)'))
+    expect(bad).toEqual([])
+  })
+
+  test('the strip keeps its top margin against the shared shorthand', async ({ page }) => {
+    // `.stat-strip` declares `margin: 0 0 ...` later in the file, which was
+    // silently resetting this element's top margin and leaving it flush
+    // against the resources panel above it.
+    await page.setViewportSize({ width: 1440, height: 950 })
+    await page.goto('/app/manage')
+    const top = await page.locator('.manage-summary')
+      .evaluate(el => parseFloat(getComputedStyle(el).marginTop))
+    expect(top).toBeGreaterThan(12)
+  })
+})

core/http/react-ui/e2e/home-lanes.spec.js

@@ -0,0 +1,103 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// Home's resident-model list and the app footer.
+
+const SYS_INFO = {
+  backends: ['llama-cpp'],
+  loaded_models: [
+    { id: 'qwen3-8b-instruct', backend: 'llama-cpp' },
+    { id: 'parakeet-tdt-0.6b' },
+  ],
+}
+
+async function mockLoaded(page) {
+  await page.route('**/system', route => route.fulfill({ json: SYS_INFO }))
+  await page.route('**/v1/models', route =>
+    route.fulfill({ json: { data: [{ id: 'qwen3-8b-instruct' }, { id: 'parakeet-tdt-0.6b' }] } }))
+}
+
+test.describe('Home resident models', () => {
+  test('resident models read as lanes, not status chips', async ({ page }) => {
+    await mockLoaded(page)
+    await page.goto('/app')
+    const lanes = page.locator('.home-loaded .lane')
+    await expect(lanes).toHaveCount(2)
+    // Model ids are identifiers, so they are set in mono like every other
+    // identifier in the app.
+    const family = await lanes.first().locator('.lane__name').evaluate(
+      el => getComputedStyle(el).fontFamily.toLowerCase())
+    expect(family).toMatch(/mono|consol|menlo/)
+  })
+
+  test('each lane keeps its stop control', async ({ page }) => {
+    await mockLoaded(page)
+    await page.goto('/app')
+    const lane = page.locator('.home-loaded .lane').first()
+    await expect(lane.getByRole('button', { name: /stop/i })).toBeVisible()
+  })
+
+  test('the header reports how many are resident as a figure', async ({ page }) => {
+    await mockLoaded(page)
+    await page.goto('/app')
+    const stat = page.locator('[data-testid="home-stat-loaded"]')
+    await expect(stat).toBeVisible()
+    await expect(stat).toContainText('2')
+    // Digits that sit in a column need to line up.
+    const numeric = await stat.locator('.home-stat__value').evaluate(
+      el => getComputedStyle(el).fontVariantNumeric)
+    expect(numeric).toContain('tabular-nums')
+  })
+
+  test('a resident model names the engine serving it', async ({ page }) => {
+    await mockLoaded(page)
+    await page.goto('/app')
+    // Lanes are sorted by id, so target by content rather than position.
+    const qwen = page.locator('.home-loaded .lane', { hasText: 'qwen3-8b-instruct' })
+    await expect(qwen).toContainText('llama-cpp')
+  })
+
+  test('a model without a config shows no engine rather than a guess', async ({ page }) => {
+    await mockLoaded(page)
+    await page.goto('/app')
+    // parakeet has no backend in the payload; the column stays blank.
+    const parakeet = page.locator('.home-loaded .lane', { hasText: 'parakeet-tdt-0.6b' })
+    await expect(parakeet).not.toContainText('llama-cpp')
+  })
+
+  test('jump-back-in offers the three places worth returning to', async ({ page }) => {
+    await mockLoaded(page)
+    await page.goto('/app')
+    const lanes = page.locator('.lanes--jump .lane')
+    await expect(lanes).toHaveCount(3)
+    await expect(lanes.first()).toContainText('Discover')
+  })
+
+  test('nothing resident still says so', async ({ page }) => {
+    await page.route('**/system', route =>
+      route.fulfill({ json: { backends: ['llama-cpp'], loaded_models: [] } }))
+    await page.route('**/v1/models', route => route.fulfill({ json: { data: [{ id: 'a-model' }] } }))
+    await page.goto('/app')
+    await expect(page.locator('.home-loaded-empty')).toBeVisible()
+    await expect(page.locator('.home-loaded .lane')).toHaveCount(0)
+  })
+})
+
+test.describe('App footer', () => {
+  test('is one line, not three stacked rows', async ({ page }) => {
+    await page.goto('/app')
+    const footer = page.locator('.app-footer')
+    await expect(footer).toBeVisible()
+    // Three centred rows of chrome cost more vertical space than the content
+    // they sit under is usually worth.
+    const height = await footer.evaluate(el => el.getBoundingClientRect().height)
+    expect(height).toBeLessThan(56)
+  })
+
+  test('keeps every link it had', async ({ page }) => {
+    await page.goto('/app')
+    const footer = page.locator('.app-footer')
+    for (const name of [/github/i, /documentation/i, /author/i]) {
+      await expect(footer.getByRole('link', { name })).toBeVisible()
+    }
+  })
+})

core/http/react-ui/e2e/models-gallery.spec.js

@@ -1786,6 +1786,19 @@ test.describe("Models Gallery - Discover split view", () => {
     );
   });
 
+  test("a build that fits at some context sizes warns rather than erroring", async ({
+    page,
+  }) => {
+    await railItem(page, "llama-model").click();
+    const verdict = page.locator(".discover__chart-verdict");
+    await expect(verdict).toBeVisible();
+    // A model that fits at 32k but not 64k is a trade-off, and #11288 keeps a
+    // test on such a build still being installable. Only "fits nowhere" earns
+    // the error tone; anything short of that warns.
+    await expect(verdict).toHaveClass(/discover__chart-verdict--warn/);
+    await expect(verdict).not.toHaveClass(/discover__chart-verdict--bad/);
+  });
+
   test("a host with no GPU gets no chart rather than an unanchored one", async ({
     page,
   }) => {

core/http/react-ui/e2e/models-recommended-panel.spec.js

@@ -56,81 +56,23 @@ async function gotoModels(page) {
 }
 
 test.describe("Models gallery - recommended panel prominence", () => {
-  test("first visit with nothing installed shows the panel expanded", async ({ page }) => {
+  test("it is a section in the flow, not a dismissable card", async ({ page }) => {
     await mockGallery(page, 0);
     await gotoModels(page);
-
-    await expect(toggle(page)).toHaveAttribute("aria-expanded", "true");
-    await expect(grid(page)).toBeVisible();
-    await expect(grid(page).getByText("tiny-chat")).toBeVisible();
+    await expect(panel(page)).toBeVisible();
+    // No close button and no collapse: this is the one thing the page has to
+    // say about the machine it runs on, not an interruption to be shut.
+    await expect(panel(page).locator("button[aria-expanded]")).toHaveCount(0);
+    await expect(panel(page).getByRole("button", { name: /dismiss|close/i })).toHaveCount(0);
+    // And no card chrome, so it sits in the pane rather than on top of it.
+    const border = await panel(page).evaluate((el) => getComputedStyle(el).borderTopWidth);
+    expect(parseFloat(border)).toBe(0);
   });
 
-  test("a user with models installed gets it collapsed by default", async ({ page }) => {
-    await mockGallery(page, 12);
-    await gotoModels(page);
 
-    await expect(toggle(page)).toHaveAttribute("aria-expanded", "false");
-    await expect(grid(page)).toBeHidden();
-    // Collapsed is a summary, not a removal: the heading stays on the page.
-    await expect(panel(page).getByText("Recommended for your hardware")).toBeVisible();
-    await expect(panel(page).getByText("2 models suggested")).toBeVisible();
-  });
 
-  test("the collapsed summary expands again on activation", async ({ page }) => {
-    await mockGallery(page, 12);
-    await gotoModels(page);
 
-    await expect(grid(page)).toBeHidden();
-    await toggle(page).click();
 
-    await expect(toggle(page)).toHaveAttribute("aria-expanded", "true");
-    await expect(grid(page)).toBeVisible();
-    await expect(page.evaluate((k) => localStorage.getItem(k), COLLAPSE_KEY)).resolves.toBe("0");
-  });
-
-  test("the collapse choice persists across a reload", async ({ page }) => {
-    await mockGallery(page, 0);
-    await gotoModels(page);
-    await expect(grid(page)).toBeVisible();
-
-    await toggle(page).click();
-    await expect(grid(page)).toBeHidden();
-
-    await page.reload();
-    await expect(panel(page)).toBeVisible({ timeout: 20_000 });
-    await expect(toggle(page)).toHaveAttribute("aria-expanded", "false");
-    await expect(grid(page)).toBeHidden();
-  });
-
-  test("dismissing it persists across a reload", async ({ page }) => {
-    await mockGallery(page, 0);
-    await gotoModels(page);
-
-    await panel(page).getByRole("button", { name: "Dismiss recommendations" }).click();
-    await expect(panel(page)).toHaveCount(0);
-    await expect(page.evaluate((k) => localStorage.getItem(k), DISMISS_KEY)).resolves.toBe("1");
-
-    await page.reload();
-    // The rail having entries is the marker that the page finished rendering
-    // without the panel. It used to be the table, which no longer exists.
-    await expect(
-      page.locator('[data-testid="discover-rail-item"]').first(),
-    ).toBeVisible({ timeout: 20_000 });
-    await expect(panel(page)).toHaveCount(0);
-  });
-
-  test("the toggle is keyboard operable and exposes its state", async ({ page }) => {
-    await mockGallery(page, 12);
-    await gotoModels(page);
-
-    await toggle(page).focus();
-    await expect(toggle(page)).toBeFocused();
-    await page.keyboard.press("Enter");
-    await expect(toggle(page)).toHaveAttribute("aria-expanded", "true");
-    // aria-controls must resolve to the region it actually shows and hides.
-    await expect(toggle(page)).toHaveAttribute("aria-controls", "rec-models-content");
-    await expect(grid(page)).toBeVisible();
-  });
 
   test("recommendations render and their install buttons still work", async ({ page }) => {
     await mockGallery(page, 0);
@@ -141,11 +83,23 @@ test.describe("Models gallery - recommended panel prominence", () => {
     });
     await gotoModels(page);
 
-    const card = grid(page).locator(".rec-models-item", { hasText: "tiny-chat" });
-    await expect(card).toBeVisible();
-    await expect(card.getByText("512.0 MB")).toBeVisible();
-    await card.getByRole("button", { name: "Install" }).click();
+    // Ranked candidates read in fit order, so these are lanes now rather than
+    // a grid of equal cards.
+    const row = grid(page).locator(".lane", { hasText: "tiny-chat" });
+    await expect(row).toBeVisible();
+    await expect(row.getByText("512.0 MB")).toBeVisible();
+    await row.getByRole("button", { name: "Install" }).click();
 
     await expect.poll(() => installed).toBe("tiny-chat");
   });
+
+  test("the best fit is called out, the rest are alternatives", async ({ page }) => {
+    await mockGallery(page, 0);
+    await gotoModels(page);
+    const rows = grid(page).locator(".lane");
+    await expect(rows.first().locator(".lane__tag--evidence")).toHaveText("Best fit");
+    // One opinion per page: the others are alternatives, not runners-up worth
+    // their own colour.
+    await expect(grid(page).locator(".lane__tag--evidence")).toHaveCount(1);
+  });
 });

core/http/react-ui/e2e/operate-overview.spec.js

@@ -0,0 +1,172 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// Operate overview (src/pages/OperateOverview.jsx).
+//
+// The page exists to answer "is anything wrong" without visiting four other
+// pages, so the tests are written against that behaviour rather than against
+// the markup: what does it say when nothing is wrong, and does each source of
+// trouble actually surface.
+
+const OVERVIEW = '[data-testid="operate-overview"]'
+const CLEAR = '[data-testid="operate-attention-clear"]'
+const ITEM = '[data-testid="operate-attention-item"]'
+
+const NO_UPGRADES = {}
+const ONE_UPGRADE = {
+  'llama-cpp': {
+    backend_name: 'llama-cpp',
+    installed_version: '0.9.4',
+    available_version: '0.9.7',
+  },
+}
+
+// A quiet installation: nothing running, nothing stale, every node healthy.
+async function mockQuiet(page, { upgrades = NO_UPGRADES, operations = [] } = {}) {
+  await page.route('**/api/backends/upgrades', route =>
+    route.fulfill({ json: upgrades }))
+  await page.route('**/api/operations', route =>
+    route.fulfill({ json: operations }))
+  await page.route('**/api/nodes', route =>
+    route.fulfill({ json: [{ id: 'node-a', status: 'healthy', healthy: true }] }))
+}
+
+test.describe('Operate overview', () => {
+  test('Operate opens the overview, not whichever page happens to be first', async ({ page }) => {
+    await mockQuiet(page)
+    await page.goto('/app')
+    await page.locator('.sidebar-nav a.nav-item', { hasText: 'Operate' }).click()
+    // Today this lands on /app/backends purely because Backends is the first
+    // entry in operateConsole.groups — an ordering accident, not a decision.
+    await expect(page).toHaveURL(/\/app\/operate$/)
+    await expect(page.locator(OVERVIEW)).toBeVisible()
+  })
+
+  test('says so plainly when nothing needs attention', async ({ page }) => {
+    await mockQuiet(page)
+    await page.goto('/app/operate')
+    await expect(page.locator(CLEAR)).toBeVisible()
+    // The empty state is one line, not a panel full of reassuring green.
+    await expect(page.locator(ITEM)).toHaveCount(0)
+  })
+
+  test('a stale backend becomes an attention item naming the version jump', async ({ page }) => {
+    await mockQuiet(page, { upgrades: ONE_UPGRADE })
+    await page.goto('/app/operate')
+    const item = page.locator(ITEM, { hasText: 'llama-cpp' })
+    await expect(item).toBeVisible()
+    await expect(item).toContainText('0.9.4')
+    await expect(item).toContainText('0.9.7')
+    await expect(page.locator(CLEAR)).toHaveCount(0)
+  })
+
+  test('a failed operation becomes an attention item', async ({ page }) => {
+    await mockQuiet(page, {
+      operations: [{ id: 'op-1', name: 'qwen3-8b', type: 'install', error: 'no space left on device' }],
+    })
+    await page.goto('/app/operate')
+    await expect(page.locator(ITEM, { hasText: 'qwen3-8b' })).toBeVisible()
+  })
+
+  test('the rail reports backend updates alongside the label', async ({ page }) => {
+    await mockQuiet(page, { upgrades: ONE_UPGRADE })
+    await page.goto('/app/operate')
+    const backends = page.locator('.console-rail a.nav-item[href="/app/backends"]')
+    await expect(backends).toBeVisible()
+    await expect(backends.locator('.nav-signal')).toContainText('1')
+  })
+
+  test('the rail groups Runtime, Cluster, Observability and Administration', async ({ page }) => {
+    await mockQuiet(page)
+    await page.goto('/app/operate')
+    const rail = page.locator('.console-rail')
+    for (const group of ['Runtime', 'Cluster', 'Observability', 'Administration']) {
+      await expect(rail.locator('.console-group-title', { hasText: group })).toBeVisible()
+    }
+    // Six headings for thirteen items was the defect; the old pairs are gone.
+    for (const gone of ['Inference', 'Access']) {
+      await expect(rail.locator('.console-group-title', { hasText: new RegExp(`^${gone}$`) })).toHaveCount(0)
+    }
+  })
+
+  test('regrouping does not change what a non-distributed host can see', async ({ page }) => {
+    await page.route('**/api/features', route =>
+      route.fulfill({ json: { distributed: false, agents: true, mcp: true } }))
+    await mockQuiet(page)
+    await page.goto('/app/operate')
+    const rail = page.locator('.console-rail')
+    await expect(rail.locator('a.nav-item[href="/app/backends"]')).toBeVisible()
+    // Gating is the thing most likely to break silently when items move group.
+    await expect(rail.locator('a.nav-item[href="/app/nodes"]')).toHaveCount(0)
+    await expect(rail.locator('a.nav-item[href="/app/scheduling"]')).toHaveCount(0)
+  })
+
+  test('the sidebar keeps its operations badge', async ({ page }) => {
+    // Regression guard: this change edits the same config the badge reads, and
+    // the badge deliberately lives on the always-visible sidebar entry rather
+    // than the collapsible rail.
+    await mockQuiet(page, {
+      operations: [{ id: 'op-1', name: 'qwen3-8b', type: 'install', progress: 40 }],
+    })
+    await page.goto('/app')
+    await expect(page.locator('.sidebar-nav .nav-badge')).toBeVisible()
+  })
+
+  test('does not poll the summary away from Operate', async ({ page }) => {
+    let upgradeCalls = 0
+    await page.route('**/api/backends/upgrades', route => {
+      upgradeCalls += 1
+      route.fulfill({ json: NO_UPGRADES })
+    })
+    await page.route('**/api/operations', route => route.fulfill({ json: [] }))
+    await page.goto('/app/chat')
+    await expect(page.locator('.sidebar')).toBeVisible()
+    await page.waitForTimeout(1500)
+    // Nobody asked for this data outside Operate; a dashboard-shaped poll on
+    // every page is exactly what OperationsContext exists to avoid.
+    expect(upgradeCalls).toBe(0)
+  })
+})
+
+test.describe('Operate overview headline', () => {
+  const SUMMARY = {
+    total: 18402, errors: 37, p95_ms: 842, window_hours: 24,
+    buckets: Array.from({ length: 12 }, (_, i) => ({ count: 100 + i * 10, errors: i })),
+  }
+
+  test('reports counted totals rather than fetching the trace list', async ({ page }) => {
+    let listCalls = 0
+    await page.route('**/api/traces?**', route => { listCalls += 1; route.fulfill({ json: [] }) })
+    await page.route('**/api/traces/summary', route => route.fulfill({ json: SUMMARY }))
+    await mockQuiet(page)
+    await page.goto('/app/operate')
+    const headline = page.locator('.operate-headline')
+    await expect(headline).toBeVisible()
+    await expect(headline).toContainText('18,402')
+    await expect(headline).toContainText('37')
+    await expect(headline).toContainText('842')
+    // The whole point of the endpoint: three numbers, not the buffer.
+    expect(listCalls).toBe(0)
+  })
+
+  test('a quiet installation keeps the grid and says why it is empty', async ({ page }) => {
+    // Hiding the grid removed the page's structure exactly when someone was
+    // most likely to be looking at it, and "0 failed" is information.
+    await page.route('**/api/traces/summary', route =>
+      route.fulfill({ json: { total: 0, errors: 0, p95_ms: 0, window_hours: 24, buckets: [] } }))
+    await mockQuiet(page)
+    await page.goto('/app/operate')
+    await expect(page.locator('.operate-headline')).toBeVisible()
+    await expect(page.locator('.operate-headline__cell')).toHaveCount(4)
+    await expect(page.locator('.operate-headline__note')).toBeVisible()
+  })
+
+  test('the sections state counts rather than listing their destinations', async ({ page }) => {
+    await page.route('**/api/traces/summary', route =>
+      route.fulfill({ json: { total: 18402, errors: 37, p95_ms: 842, window_hours: 24, buckets: [] } }))
+    await mockQuiet(page)
+    await page.goto('/app/operate')
+    const runtime = page.locator('.lanes--sections .lane').first()
+    await expect(runtime).toContainText('backends')
+    await expect(runtime).toContainText('running')
+  })
+})

core/http/react-ui/e2e/page-render-smoke.spec.js

@@ -19,6 +19,7 @@ const PAGES = [
   ['/app/account', 'Account'],
   ['/app/studio', 'Studio'],
   ['/app/manage', 'Manage'],
+  ['/app/operate', 'Operate overview'],
   ['/app/backends', 'Backends'],
   ['/app/activity', 'Activity'],
   ['/app/settings', 'Settings'],

core/http/react-ui/e2e/studio-overview.spec.js

@@ -0,0 +1,155 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// Studio overview (src/pages/StudioOverview.jsx).
+//
+// Studio was a tab strip over six generators that opened on Images and told you
+// nothing about what this machine could actually run. The tests are about that:
+// what the strip reports before you click, and the difference between a
+// modality that is switched off and one that merely has no model.
+
+const OVERVIEW = '[data-testid="studio-overview"]'
+const MODALITY = '[data-testid="studio-modality"]'
+const tabFor = (page, key) => page.locator(`.studio-tab[data-tab="${key}"]`)
+
+const model = (id, ...capabilities) => ({ id, capabilities })
+
+// Images and speech covered, video and sound not. 3D and transform are feature
+// flags rather than models, so they are controlled separately.
+const SOME_MODELS = {
+  data: [
+    model('flux.1-schnell', 'FLAG_IMAGE'),
+    model('kokoro-82m', 'FLAG_TTS'),
+    model('qwen3-8b', 'FLAG_CHAT'),
+  ],
+}
+
+async function mockCapabilities(page, payload = SOME_MODELS) {
+  await page.route('**/api/models/capabilities', route => route.fulfill({ json: payload }))
+}
+
+test.describe('Studio overview', () => {
+  test('Studio opens on the overview rather than dropping into Images', async ({ page }) => {
+    await mockCapabilities(page)
+    await page.goto('/app/studio')
+    await expect(page.locator(OVERVIEW)).toBeVisible()
+  })
+
+  test('a generator path opens that generator', async ({ page }) => {
+    await mockCapabilities(page)
+    await page.goto('/app/studio/images')
+    await expect(page.locator(OVERVIEW)).toHaveCount(0)
+    await expect(page.locator('.media-layout')).toBeVisible()
+  })
+
+  test('an unrecognised tab falls back to the overview, not to Images', async ({ page }) => {
+    await mockCapabilities(page)
+    await page.goto('/app/studio/nonsense')
+    await expect(page.locator(OVERVIEW)).toBeVisible()
+  })
+
+  test('the tab strip reports which modalities have a model', async ({ page }) => {
+    await mockCapabilities(page)
+    await page.goto('/app/studio')
+    // Filled: something installed advertises the capability.
+    await expect(tabFor(page, 'images').locator('.studio-tab__dot--on')).toBeVisible()
+    await expect(tabFor(page, 'tts').locator('.studio-tab__dot--on')).toBeVisible()
+    // Hollow: the modality is available, nothing serves it yet.
+    await expect(tabFor(page, 'video').locator('.studio-tab__dot--off')).toBeVisible()
+    await expect(tabFor(page, 'sound').locator('.studio-tab__dot--off')).toBeVisible()
+  })
+
+  test('a modality with no model offers a way to install one', async ({ page }) => {
+    await mockCapabilities(page)
+    await page.goto('/app/studio')
+    const video = page.locator(`${MODALITY}[data-modality="video"]`)
+    await expect(video).toBeVisible()
+    // The point of the lane: not a dead tab, a route to fixing it.
+    await expect(video.locator('a[href*="/app/models"]')).toBeVisible()
+  })
+
+  test('a modality with a model names it instead of offering an install', async ({ page }) => {
+    await mockCapabilities(page)
+    await page.goto('/app/studio')
+    const images = page.locator(`${MODALITY}[data-modality="images"]`)
+    await expect(images).toContainText('flux.1-schnell')
+    await expect(images.locator('a[href*="/app/models"]')).toHaveCount(0)
+  })
+
+  test('a disabled feature gets no tab and no lane at all', async ({ page }) => {
+    // Switched off is a different thing from "no model installed", and
+    // conflating them is how someone ends up staring at a control that cannot
+    // work. 3d is a permission rather than an /api/features entry, and
+    // hasFeature() short-circuits to true for admins and for auth-off
+    // installations, so withholding it needs a real non-admin session.
+    await page.route('**/api/auth/status', route => route.fulfill({
+      json: {
+        authEnabled: true,
+        user: { name: 'someone', role: 'user', permissions: { images: true, video: true, tts: true, sound: true } },
+      },
+    }))
+    await mockCapabilities(page)
+    await page.goto('/app/studio')
+    await expect(page.locator(OVERVIEW)).toBeVisible()
+    await expect(tabFor(page, 'threed')).toHaveCount(0)
+    await expect(page.locator(`${MODALITY}[data-modality="threed"]`)).toHaveCount(0)
+  })
+
+  test('asks the capabilities endpoint once, not once per modality', async ({ page }) => {
+    let calls = 0
+    await page.route('**/api/models/capabilities', route => {
+      calls += 1
+      route.fulfill({ json: SOME_MODELS })
+    })
+    await page.goto('/app/studio')
+    await expect(page.locator(OVERVIEW)).toBeVisible()
+    await page.waitForTimeout(500)
+    // useModels() fetches the whole list and filters in the browser, so one
+    // hook per modality would be six identical requests on every mount.
+    expect(calls).toBe(1)
+  })
+
+  test('an installation with no models at all still renders every modality', async ({ page }) => {
+    await mockCapabilities(page, { data: [] })
+    await page.goto('/app/studio')
+    await expect(page.locator(OVERVIEW)).toBeVisible()
+    await expect(page.locator(MODALITY).first()).toBeVisible()
+    await expect(page.locator('.studio-tab__dot--on')).toHaveCount(0)
+  })
+
+  test('recent outputs surface what was generated earlier', async ({ page }) => {
+    await mockCapabilities(page)
+    // History is localStorage, written by each generator. The overview is the
+    // first place it is read across modalities rather than within one.
+    await page.addInitScript(() => {
+      localStorage.setItem('localai_image_history', JSON.stringify([
+        { id: 'i1', createdAt: Date.now(), model: 'flux.1-schnell', prompt: 'a brass orrery', elapsedMs: 6100 },
+      ]))
+    })
+    await page.goto('/app/studio')
+    const shelf = page.locator('[data-testid="studio-recent"]')
+    await expect(shelf).toBeVisible()
+    await expect(shelf).toContainText('flux.1-schnell')
+  })
+
+  test('no history means no empty shelf', async ({ page }) => {
+    await mockCapabilities(page)
+    await page.goto('/app/studio')
+    await expect(page.locator('[data-testid="studio-recent"]')).toHaveCount(0)
+  })
+
+  test('the overview is reachable back from a generator tab', async ({ page }) => {
+    await mockCapabilities(page)
+    await page.goto('/app/studio/images')
+    await tabFor(page, 'overview').click()
+    await expect(page.locator(OVERVIEW)).toBeVisible()
+  })
+
+  test('a legacy ?tab= link is redirected to its path', async ({ page }) => {
+    // Bookmarks and older docs still use the query form; they must keep working
+    // and must land on the canonical URL rather than a second spelling of it.
+    await mockCapabilities(page)
+    await page.goto('/app/studio?tab=images')
+    await expect(page).toHaveURL(/\/app\/studio\/images$/)
+    await expect(page.locator('.media-layout')).toBeVisible()
+  })
+})

core/http/react-ui/e2e/studio-transform.spec.js

@@ -2,7 +2,7 @@ import { test, expect } from './coverage-fixtures.js'
 
 test.describe('Studio - Transform', () => {
   test('Studio exposes a Transform tab that renders Audio Transform', async ({ page }) => {
-    await page.goto('/app/studio?tab=transform')
+    await page.goto('/app/studio/transform')
     await expect(page.locator('.studio-tab', { hasText: 'Transform' })).toBeVisible()
     await expect(page.locator('h1.page-title', { hasText: 'Audio Transform' })).toBeVisible({ timeout: 15_000 })
   })

core/http/react-ui/e2e/studio-workbench.spec.js

@@ -0,0 +1,49 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// The generator workbenches (mock 5b/5c): the control column and the record of
+// what the form actually sent.
+
+test.describe('Studio workbench', () => {
+  test('the control column is a hairline field stack, not a shadowed card', async ({ page }) => {
+    await page.goto('/app/studio/images')
+    const controls = page.locator('.media-controls')
+    await expect(controls).toBeVisible()
+    const style = await controls.evaluate(el => {
+      const cs = getComputedStyle(el)
+      return { shadow: cs.boxShadow, radius: cs.borderTopLeftRadius }
+    })
+    expect(style.shadow).toBe('none')
+    expect(style.radius).toBe('0px')
+  })
+
+  test('fields are separated by a rule and labelled in caps', async ({ page }) => {
+    await page.goto('/app/studio/images')
+    const label = page.locator('.media-controls .form-label').first()
+    await expect(label).toBeVisible()
+    const cs = await label.evaluate(el => getComputedStyle(el).textTransform)
+    expect(cs).toBe('uppercase')
+  })
+
+  test('no request is shown before one has been made', async ({ page }) => {
+    // A panel describing a request nobody sent is a tutorial, not a record.
+    await page.goto('/app/studio/images')
+    await expect(page.locator('.request-panel')).toHaveCount(0)
+  })
+
+  test('generating records the request that was actually sent', async ({ page }) => {
+    await page.route('**/api/models/capabilities', route =>
+      route.fulfill({ json: { data: [{ id: 'flux-mock', capabilities: ['FLAG_IMAGE'] }] } }))
+    await page.route('**/v1/images/generations', route =>
+      route.fulfill({ json: { data: [{ url: 'https://example.invalid/a.png' }] } }))
+
+    await page.goto('/app/studio/images')
+    await page.locator('.media-controls textarea').first().fill('a brass orrery')
+    await page.getByRole('button', { name: /generate/i }).click()
+
+    const panel = page.locator('.request-panel')
+    await expect(panel).toBeVisible()
+    await expect(panel).toContainText('/v1/images/generations')
+    await expect(panel).toContainText('a brass orrery')
+    await expect(panel.getByRole('button', { name: /curl/i })).toBeVisible()
+  })
+})

core/http/react-ui/e2e/theme-default.spec.js

@@ -0,0 +1,15 @@
+import { test, expect } from './coverage-fixtures.js'
+
+test.describe('Theme default', () => {
+  test('a fresh install opens dark even when the OS prefers light', async ({ page }) => {
+    await page.emulateMedia({ colorScheme: 'light' })
+    await page.goto('/app')
+    await expect(page.locator('html')).toHaveAttribute('data-theme', 'dark')
+  })
+
+  test('a stored choice still wins', async ({ page }) => {
+    await page.addInitScript(() => localStorage.setItem('localai-theme', 'light'))
+    await page.goto('/app')
+    await expect(page.locator('html')).toHaveAttribute('data-theme', 'light')
+  })
+})

core/http/react-ui/e2e/threed-gen.spec.js

@@ -276,9 +276,11 @@ test.describe('3D generation', () => {
       })
     })
 
-    await page.goto('/app/studio?tab=threed')
+    await page.goto('/app/studio/threed')
     await expect(page.getByRole('button', { name: '3D', exact: true })).toHaveCount(0)
-    await expect(page.locator('.studio-tab', { hasText: 'Images' })).toHaveClass(/studio-tab-active/)
+    // Falls back to the overview rather than Images. Landing on Images was
+    // never a decision, only the first entry in the tab array.
+    await expect(page.locator('.studio-tab[data-tab="overview"]')).toHaveClass(/studio-tab-active/)
 
     await page.goto('/app/3d')
     await expect(page).toHaveURL(/\/app\/?$/)

core/http/react-ui/e2e/traces-latency.spec.js

@@ -0,0 +1,42 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// Traces rows carry latency as a shape, not only as a number buried in the
+// expanded detail (mock 6d).
+
+const TRACES = [
+  { id: '1', timestamp: new Date().toISOString(), duration: 4_200_000_000,
+    request: { method: 'POST', path: '/v1/chat/completions' }, response: { status: 500 }, error: 'context length exceeded' },
+  { id: '2', timestamp: new Date().toISOString(), duration: 980_000_000,
+    request: { method: 'POST', path: '/v1/chat/completions' }, response: { status: 200 } },
+  { id: '3', timestamp: new Date().toISOString(), duration: 186_000_000,
+    request: { method: 'POST', path: '/v1/embeddings' }, response: { status: 200 } },
+]
+
+test.describe('Traces latency', () => {
+  test.beforeEach(async ({ page }) => {
+    await page.route('**/api/traces**', route => route.fulfill({ json: TRACES }))
+    await page.goto('/app/traces')
+  })
+
+  test('every row shows a latency bar and figure', async ({ page }) => {
+    const cells = page.locator('.lat')
+    await expect(cells).toHaveCount(3)
+    await expect(cells.first()).toContainText('4.20s')
+  })
+
+  test('the bar is scaled against the slowest request in view', async ({ page }) => {
+    // Wait for the rows: goto alone does not guarantee the fetch has painted.
+    await expect(page.locator('.lat__bar i')).toHaveCount(3)
+    const widths = await page.locator('.lat__bar i').evaluateAll(
+      els => els.map(el => parseFloat(el.style.width)))
+    // 4.2s is the slowest, so it is full; 186ms is a sliver of it.
+    expect(widths[0]).toBe(100)
+    expect(widths[2]).toBeLessThan(20)
+  })
+
+  test('a slow request is marked, not just long', async ({ page }) => {
+    await expect(page.locator('.lat')).toHaveCount(3)
+    // Colour carries the threshold; the figure carries the value.
+    await expect(page.locator('.lat__bar--slow')).toHaveCount(1)
+  })
+})

core/http/react-ui/e2e/voice-library-empty.spec.js

@@ -0,0 +1,33 @@
+import { test, expect } from './coverage-fixtures.js'
+
+// The empty voice library must offer its action, visibly, inside the panel.
+
+test.describe('Voice library empty state', () => {
+  test.beforeEach(async ({ page }) => {
+    await page.route('**/api/voice-profiles', route => route.fulfill({ json: [] }))
+    await page.goto('/app/voice-library')
+  })
+
+  test('the create action is visible and a normal size', async ({ page }) => {
+    const action = page.locator('.empty-state__actions a.btn').first()
+    await expect(action).toBeVisible()
+    const box = await action.boundingBox()
+    // It had been carrying the panel's own min-height:430px, which made it an
+    // invisible box that pushed itself out of view.
+    expect(box.height).toBeLessThan(80)
+  })
+
+  test('the action sits inside the panel, not past its edge', async ({ page }) => {
+    const panel = page.locator('.empty-state').first()
+    const action = page.locator('.empty-state__actions a.btn').first()
+    const [p, a] = [await panel.boundingBox(), await action.boundingBox()]
+    expect(a.y + a.height).toBeLessThanOrEqual(p.y + p.height + 1)
+  })
+
+  test('the panel renders its icon', async ({ page }) => {
+    const icon = page.locator('.empty-state-icon').first()
+    await expect(icon).toBeVisible()
+    const box = await icon.boundingBox()
+    expect(box.width).toBeGreaterThan(0)
+  })
+})

core/http/react-ui/inline-style-baseline.txt

@@ -1 +1 @@
-624
+538

core/http/react-ui/public/locales/de/admin.json

@@ -12,6 +12,8 @@
     "timeLeft": "{{value}} left",
     "cancel": "Cancel",
     "cancelLabel": "Cancel {{name}}",
+    "pause": "Pause",
+    "pauseLabel": "Pause {{name}} and keep downloaded data",
     "retry": "Retry",
     "retryLabel": "Retry {{name}}",
     "nodeCount": "{{count}} nodes",
@@ -143,5 +145,36 @@
   "explorer": {
     "title": "Explorer",
     "subtitle": "Dateien und Konfiguration durchsuchen"
+  },
+  "operate": {
+    "overview": {
+      "title": "Overview",
+      "subtitle": "Everything running on this installation, and anything that wants a decision.",
+      "attention": {
+        "heading": "Needs attention",
+        "clear": "Nothing needs attention. Backends are current, no operation has failed, and every node is healthy.",
+        "backendUpdate": "Update available: {{from}} → {{to}}"
+      },
+      "sections": {
+        "heading": "Sections",
+        "runtime": "Runtime",
+        "runtimeSummary": "{{backends}} backends · {{models}} models · {{updates}} updates · {{running}} running",
+        "cluster": "Cluster",
+        "clusterSummary": "{{nodes}} nodes",
+        "observability": "Observability",
+        "observabilitySummary": "Usage and traces",
+        "administration": "Administration",
+        "administrationSummary": "Users, middleware and settings · {{memory}} memory in use",
+        "clusterSingle": "Single node",
+        "observabilityCounted": "{{requests}} requests · {{errors}} failed · p95 {{p95}} ms"
+      },
+      "headline": {
+        "requests": "Requests · {{hours}}h",
+        "errors": "Failed requests",
+        "p95": "p95 latency",
+        "quiet": "No requests served in this window yet.",
+        "host": "Host memory"
+      }
+    }
   }
 }

core/http/react-ui/public/locales/de/chat.json

@@ -118,5 +118,8 @@
     "newChat": "Neuer Chat",
     "clearAll": "Alle löschen",
     "deleteAllTitle": "Alle Unterhaltungen löschen"
+  },
+  "message": {
+    "you": "You"
   }
 }

core/http/react-ui/public/locales/de/home.json

@@ -17,7 +17,9 @@
     "modelsLoaded_other": "{{count}} models loaded",
     "noModelsLoaded": "No models loaded",
     "nodes_one": "{{count}} node",
-    "nodes_other": "{{count}} nodes"
+    "nodes_other": "{{count}} nodes",
+    "loadedLabel": "Loaded",
+    "nodesLabel": "Nodes"
   },
   "assistant": {
     "title": "LocalAI per Chat verwalten",
@@ -47,7 +49,8 @@
     "count_one": "{{count}} Modell geladen",
     "count_other": "{{count}} Modelle geladen",
     "stop": "Modell stoppen",
-    "stopAll": "Alle stoppen"
+    "stopAll": "Alle stoppen",
+    "serving": "Serving"
   },
   "stopDialog": {
     "title": "Modell stoppen",
@@ -88,5 +91,14 @@
     "browse": "Browse the API",
     "hide": "Hide endpoints",
     "dismiss": "Dismiss"
+  },
+  "jump": {
+    "heading": "Jump back in",
+    "discover": "Discover",
+    "discoverSummary": "Browse the gallery and install models",
+    "create": "Create",
+    "createSummary": "Open a chat, image or voice session",
+    "operate": "Operate",
+    "operateSummary": "{{models}} models configured · nodes, activity and traces"
   }
 }

core/http/react-ui/public/locales/de/media.json

@@ -5,6 +5,32 @@
       "video": "Video",
       "tts": "TTS",
       "sound": "Audio",
+      "transform": "Transform",
+      "overview": "Overview"
+    },
+    "overview": {
+      "eyebrow": "{{ready}} of {{total}} modalities ready",
+      "title": "Studio",
+      "subtitle": "Generate images, video, 3D, speech and sound with the models on this machine.",
+      "canMake": "What you can make",
+      "running": "Running now",
+      "recent": "Recent outputs",
+      "noModel": "No model installed",
+      "install": "Install a model",
+      "ready": "Ready",
+      "seconds": "{{seconds}}s",
+      "describe": {
+        "images": "Text to image, image to image, reference images",
+        "video": "Text to video and image to video",
+        "threed": "Image to mesh reconstruction",
+        "tts": "Text to speech using your voice library",
+        "sound": "Music and sound effects from a prompt",
+        "transform": "Separation, enhancement and voice conversion"
+      }
+    },
+    "groups": {
+      "create": "Create",
+      "voice": "Voice",
       "transform": "Transform"
     }
   },
@@ -157,5 +183,10 @@
     "clearMessage": "Alle Verlaufseinträge entfernen? Diese Aktion kann nicht rückgängig gemacht werden.",
     "clearConfirm": "Löschen",
     "cleared": "Verlauf gelöscht"
+  },
+  "request": {
+    "heading": "Request",
+    "copyCurl": "Copy as curl",
+    "copied": "Copied"
   }
 }

core/http/react-ui/public/locales/de/models.json

@@ -10,7 +10,9 @@
     "installStarted": "{{model}} wird installiert…",
     "installFailed": "Installation fehlgeschlagen: {{message}}",
     "dismiss": "Empfehlungen ausblenden",
-    "summary": "{{n}} Modelle vorgeschlagen"
+    "summary": "{{n}} Modelle vorgeschlagen",
+    "bestFit": "Best fit",
+    "alternative": "Also fits"
   },
   "stats": {
     "available": "Verfügbar",

core/http/react-ui/public/locales/de/nav.json

@@ -24,7 +24,9 @@
     "observability": "Observability",
     "access": "Access",
     "system": "System",
-    "activity": "Activity"
+    "activity": "Activity",
+    "runtime": "Laufzeit",
+    "administration": "Verwaltung"
   },
   "items": {
     "home": "Start",
@@ -57,7 +59,8 @@
     "settings": "Einstellungen",
     "api": "API",
     "middleware": "Middleware",
-    "activity": "Aktivität"
+    "activity": "Aktivität",
+    "overview": "Übersicht"
   },
   "footer": {
     "github": "GitHub",

core/http/react-ui/public/locales/en/admin.json

@@ -12,6 +12,8 @@
     "timeLeft": "{{value}} left",
     "cancel": "Cancel",
     "cancelLabel": "Cancel {{name}}",
+    "pause": "Pause",
+    "pauseLabel": "Pause {{name}} and keep downloaded data",
     "retry": "Retry",
     "retryLabel": "Retry {{name}}",
     "nodeCount": "{{count}} nodes",
@@ -166,5 +168,36 @@
   "explorer": {
     "title": "Explorer",
     "subtitle": "Browse files and configuration"
+  },
+  "operate": {
+    "overview": {
+      "title": "Overview",
+      "subtitle": "Everything running on this installation, and anything that wants a decision.",
+      "attention": {
+        "heading": "Needs attention",
+        "clear": "Nothing needs attention. Backends are current, no operation has failed, and every node is healthy.",
+        "backendUpdate": "Update available: {{from}} → {{to}}"
+      },
+      "sections": {
+        "heading": "Sections",
+        "runtime": "Runtime",
+        "runtimeSummary": "{{backends}} backends · {{models}} models · {{updates}} updates · {{running}} running",
+        "cluster": "Cluster",
+        "clusterSummary": "{{nodes}} nodes",
+        "observability": "Observability",
+        "observabilitySummary": "Usage and traces",
+        "administration": "Administration",
+        "administrationSummary": "Users, middleware and settings · {{memory}} memory in use",
+        "clusterSingle": "Single node",
+        "observabilityCounted": "{{requests}} requests · {{errors}} failed · p95 {{p95}} ms"
+      },
+      "headline": {
+        "requests": "Requests · {{hours}}h",
+        "errors": "Failed requests",
+        "p95": "p95 latency",
+        "quiet": "No requests served in this window yet.",
+        "host": "Host memory"
+      }
+    }
   }
 }

core/http/react-ui/public/locales/en/chat.json

@@ -124,5 +124,8 @@
     "newChat": "New chat",
     "clearAll": "Clear all",
     "deleteAllTitle": "Delete all conversations"
+  },
+  "message": {
+    "you": "You"
   }
 }

core/http/react-ui/public/locales/en/home.json

@@ -17,7 +17,9 @@
     "modelsLoaded_other": "{{count}} models loaded",
     "noModelsLoaded": "No models loaded",
     "nodes_one": "{{count}} node",
-    "nodes_other": "{{count}} nodes"
+    "nodes_other": "{{count}} nodes",
+    "loadedLabel": "Loaded",
+    "nodesLabel": "Nodes"
   },
   "assistant": {
     "title": "Manage LocalAI by chatting",
@@ -47,7 +49,8 @@
     "count_one": "{{count}} model loaded",
     "count_other": "{{count}} models loaded",
     "stop": "Stop model",
-    "stopAll": "Stop all"
+    "stopAll": "Stop all",
+    "serving": "Serving"
   },
   "stopDialog": {
     "title": "Stop Model",
@@ -103,5 +106,14 @@
     "browse": "Browse the API",
     "hide": "Hide endpoints",
     "dismiss": "Dismiss"
+  },
+  "jump": {
+    "heading": "Jump back in",
+    "discover": "Discover",
+    "discoverSummary": "Browse the gallery and install models",
+    "create": "Create",
+    "createSummary": "Open a chat, image or voice session",
+    "operate": "Operate",
+    "operateSummary": "{{models}} models configured · nodes, activity and traces"
   }
 }

core/http/react-ui/public/locales/en/media.json

@@ -6,7 +6,33 @@
       "tts": "TTS",
       "sound": "Sound",
       "transform": "Transform",
-      "threed": "3D"
+      "threed": "3D",
+      "overview": "Overview"
+    },
+    "overview": {
+      "eyebrow": "{{ready}} of {{total}} modalities ready",
+      "title": "Studio",
+      "subtitle": "Generate images, video, 3D, speech and sound with the models on this machine.",
+      "canMake": "What you can make",
+      "running": "Running now",
+      "recent": "Recent outputs",
+      "noModel": "No model installed",
+      "install": "Install a model",
+      "ready": "Ready",
+      "seconds": "{{seconds}}s",
+      "describe": {
+        "images": "Text to image, image to image, reference images",
+        "video": "Text to video and image to video",
+        "threed": "Image to mesh reconstruction",
+        "tts": "Text to speech using your voice library",
+        "sound": "Music and sound effects from a prompt",
+        "transform": "Separation, enhancement and voice conversion"
+      }
+    },
+    "groups": {
+      "create": "Create",
+      "voice": "Voice",
+      "transform": "Transform"
     }
   },
   "image": {
@@ -426,5 +452,10 @@
     "clearMessage": "Remove all history entries? This cannot be undone.",
     "clearConfirm": "Clear",
     "cleared": "History cleared"
+  },
+  "request": {
+    "heading": "Request",
+    "copyCurl": "Copy as curl",
+    "copied": "Copied"
   }
 }

core/http/react-ui/public/locales/en/models.json

@@ -11,7 +11,9 @@
     "installStarted": "Installing {{model}}…",
     "installFailed": "Install failed: {{message}}",
     "dismiss": "Dismiss recommendations",
-    "summary": "{{n}} models suggested"
+    "summary": "{{n}} models suggested",
+    "bestFit": "Best fit",
+    "alternative": "Also fits"
   },
   "stats": {
     "available": "Available",

core/http/react-ui/public/locales/en/nav.json

@@ -24,7 +24,9 @@
     "observability": "Observability",
     "access": "Access",
     "system": "System",
-    "activity": "Activity"
+    "activity": "Activity",
+    "runtime": "Runtime",
+    "administration": "Administration"
   },
   "items": {
     "home": "Home",
@@ -58,7 +60,8 @@
     "system": "System",
     "settings": "Settings",
     "api": "API",
-    "activity": "Activity"
+    "activity": "Activity",
+    "overview": "Overview"
   },
   "footer": {
     "github": "GitHub",

core/http/react-ui/public/locales/es/admin.json

@@ -12,6 +12,8 @@
     "timeLeft": "{{value}} left",
     "cancel": "Cancel",
     "cancelLabel": "Cancel {{name}}",
+    "pause": "Pause",
+    "pauseLabel": "Pause {{name}} and keep downloaded data",
     "retry": "Retry",
     "retryLabel": "Retry {{name}}",
     "nodeCount": "{{count}} nodes",
@@ -143,5 +145,36 @@
   "explorer": {
     "title": "Explorador",
     "subtitle": "Explora archivos y configuración"
+  },
+  "operate": {
+    "overview": {
+      "title": "Overview",
+      "subtitle": "Everything running on this installation, and anything that wants a decision.",
+      "attention": {
+        "heading": "Needs attention",
+        "clear": "Nothing needs attention. Backends are current, no operation has failed, and every node is healthy.",
+        "backendUpdate": "Update available: {{from}} → {{to}}"
+      },
+      "sections": {
+        "heading": "Sections",
+        "runtime": "Runtime",
+        "runtimeSummary": "{{backends}} backends · {{models}} models · {{updates}} updates · {{running}} running",
+        "cluster": "Cluster",
+        "clusterSummary": "{{nodes}} nodes",
+        "observability": "Observability",
+        "observabilitySummary": "Usage and traces",
+        "administration": "Administration",
+        "administrationSummary": "Users, middleware and settings · {{memory}} memory in use",
+        "clusterSingle": "Single node",
+        "observabilityCounted": "{{requests}} requests · {{errors}} failed · p95 {{p95}} ms"
+      },
+      "headline": {
+        "requests": "Requests · {{hours}}h",
+        "errors": "Failed requests",
+        "p95": "p95 latency",
+        "quiet": "No requests served in this window yet.",
+        "host": "Host memory"
+      }
+    }
   }
 }

core/http/react-ui/public/locales/es/chat.json

@@ -118,5 +118,8 @@
     "newChat": "Nuevo chat",
     "clearAll": "Borrar todo",
     "deleteAllTitle": "Eliminar todas las conversaciones"
+  },
+  "message": {
+    "you": "You"
   }
 }

core/http/react-ui/public/locales/es/home.json

@@ -17,7 +17,9 @@
     "modelsLoaded_other": "{{count}} models loaded",
     "noModelsLoaded": "No models loaded",
     "nodes_one": "{{count}} node",
-    "nodes_other": "{{count}} nodes"
+    "nodes_other": "{{count}} nodes",
+    "loadedLabel": "Loaded",
+    "nodesLabel": "Nodes"
   },
   "assistant": {
     "title": "Administra LocalAI chateando",
@@ -47,7 +49,8 @@
     "count_one": "{{count}} modelo cargado",
     "count_other": "{{count}} modelos cargados",
     "stop": "Detener modelo",
-    "stopAll": "Detener todos"
+    "stopAll": "Detener todos",
+    "serving": "Serving"
   },
   "stopDialog": {
     "title": "Detener modelo",
@@ -88,5 +91,14 @@
     "browse": "Browse the API",
     "hide": "Hide endpoints",
     "dismiss": "Dismiss"
+  },
+  "jump": {
+    "heading": "Jump back in",
+    "discover": "Discover",
+    "discoverSummary": "Browse the gallery and install models",
+    "create": "Create",
+    "createSummary": "Open a chat, image or voice session",
+    "operate": "Operate",
+    "operateSummary": "{{models}} models configured · nodes, activity and traces"
   }
 }

core/http/react-ui/public/locales/es/media.json

@@ -5,6 +5,32 @@
       "video": "Video",
       "tts": "TTS",
       "sound": "Sonido",
+      "transform": "Transform",
+      "overview": "Overview"
+    },
+    "overview": {
+      "eyebrow": "{{ready}} of {{total}} modalities ready",
+      "title": "Studio",
+      "subtitle": "Generate images, video, 3D, speech and sound with the models on this machine.",
+      "canMake": "What you can make",
+      "running": "Running now",
+      "recent": "Recent outputs",
+      "noModel": "No model installed",
+      "install": "Install a model",
+      "ready": "Ready",
+      "seconds": "{{seconds}}s",
+      "describe": {
+        "images": "Text to image, image to image, reference images",
+        "video": "Text to video and image to video",
+        "threed": "Image to mesh reconstruction",
+        "tts": "Text to speech using your voice library",
+        "sound": "Music and sound effects from a prompt",
+        "transform": "Separation, enhancement and voice conversion"
+      }
+    },
+    "groups": {
+      "create": "Create",
+      "voice": "Voice",
       "transform": "Transform"
     }
   },
@@ -157,5 +183,10 @@
     "clearMessage": "¿Eliminar todas las entradas del historial? Esto no se puede deshacer.",
     "clearConfirm": "Borrar",
     "cleared": "Historial borrado"
+  },
+  "request": {
+    "heading": "Request",
+    "copyCurl": "Copy as curl",
+    "copied": "Copied"
   }
 }

core/http/react-ui/public/locales/es/models.json

@@ -10,7 +10,9 @@
     "installStarted": "Instalando {{model}}…",
     "installFailed": "Error al instalar: {{message}}",
     "dismiss": "Descartar recomendaciones",
-    "summary": "{{n}} modelos sugeridos"
+    "summary": "{{n}} modelos sugeridos",
+    "bestFit": "Best fit",
+    "alternative": "Also fits"
   },
   "stats": {
     "available": "Disponibles",

core/http/react-ui/public/locales/es/nav.json

@@ -24,7 +24,9 @@
     "observability": "Observability",
     "access": "Access",
     "system": "System",
-    "activity": "Activity"
+    "activity": "Activity",
+    "runtime": "Runtime",
+    "administration": "Administración"
   },
   "items": {
     "home": "Inicio",
@@ -57,7 +59,8 @@
     "settings": "Configuración",
     "api": "API",
     "middleware": "Middleware",
-    "activity": "Actividad"
+    "activity": "Actividad",
+    "overview": "Resumen"
   },
   "footer": {
     "github": "GitHub",

core/http/react-ui/public/locales/id/admin.json

@@ -12,6 +12,8 @@
     "timeLeft": "{{value}} left",
     "cancel": "Cancel",
     "cancelLabel": "Cancel {{name}}",
+    "pause": "Pause",
+    "pauseLabel": "Pause {{name}} and keep downloaded data",
     "retry": "Retry",
     "retryLabel": "Retry {{name}}",
     "nodeCount": "{{count}} nodes",
@@ -166,5 +168,36 @@
   "explorer": {
     "title": "Penjelajah",
     "subtitle": "Jelajahi file dan konfigurasi"
+  },
+  "operate": {
+    "overview": {
+      "title": "Overview",
+      "subtitle": "Everything running on this installation, and anything that wants a decision.",
+      "attention": {
+        "heading": "Needs attention",
+        "clear": "Nothing needs attention. Backends are current, no operation has failed, and every node is healthy.",
+        "backendUpdate": "Update available: {{from}} → {{to}}"
+      },
+      "sections": {
+        "heading": "Sections",
+        "runtime": "Runtime",
+        "runtimeSummary": "{{backends}} backends · {{models}} models · {{updates}} updates · {{running}} running",
+        "cluster": "Cluster",
+        "clusterSummary": "{{nodes}} nodes",
+        "observability": "Observability",
+        "observabilitySummary": "Usage and traces",
+        "administration": "Administration",
+        "administrationSummary": "Users, middleware and settings · {{memory}} memory in use",
+        "clusterSingle": "Single node",
+        "observabilityCounted": "{{requests}} requests · {{errors}} failed · p95 {{p95}} ms"
+      },
+      "headline": {
+        "requests": "Requests · {{hours}}h",
+        "errors": "Failed requests",
+        "p95": "p95 latency",
+        "quiet": "No requests served in this window yet.",
+        "host": "Host memory"
+      }
+    }
   }
 }

core/http/react-ui/public/locales/id/chat.json

@@ -118,5 +118,8 @@
     "newChat": "Obrolan baru",
     "clearAll": "Hapus semua",
     "deleteAllTitle": "Hapus semua percakapan"
+  },
+  "message": {
+    "you": "You"
   }
 }

core/http/react-ui/public/locales/id/home.json

@@ -17,7 +17,9 @@
     "modelsLoaded_other": "{{count}} model dimuat",
     "noModelsLoaded": "Tidak ada model yang dimuat",
     "nodes_one": "{{count}} node",
-    "nodes_other": "{{count}} nodes"
+    "nodes_other": "{{count}} nodes",
+    "loadedLabel": "Loaded",
+    "nodesLabel": "Nodes"
   },
   "assistant": {
     "title": "Kelola LocalAI melalui obrolan",
@@ -47,7 +49,8 @@
     "count_one": "{{count}} model dimuat",
     "count_other": "{{count}} model dimuat",
     "stop": "Hentikan model",
-    "stopAll": "Hentikan semua"
+    "stopAll": "Hentikan semua",
+    "serving": "Serving"
   },
   "stopDialog": {
     "title": "Hentikan Model",
@@ -88,5 +91,14 @@
     "browse": "Jelajahi API",
     "hide": "Sembunyikan endpoint",
     "dismiss": "Abaikan"
+  },
+  "jump": {
+    "heading": "Jump back in",
+    "discover": "Discover",
+    "discoverSummary": "Browse the gallery and install models",
+    "create": "Create",
+    "createSummary": "Open a chat, image or voice session",
+    "operate": "Operate",
+    "operateSummary": "{{models}} models configured · nodes, activity and traces"
   }
 }

core/http/react-ui/public/locales/id/media.json

@@ -5,7 +5,33 @@
       "video": "Video",
       "tts": "TTS",
       "sound": "Suara",
-      "transform": "Transformasi"
+      "transform": "Transformasi",
+      "overview": "Overview"
+    },
+    "overview": {
+      "eyebrow": "{{ready}} of {{total}} modalities ready",
+      "title": "Studio",
+      "subtitle": "Generate images, video, 3D, speech and sound with the models on this machine.",
+      "canMake": "What you can make",
+      "running": "Running now",
+      "recent": "Recent outputs",
+      "noModel": "No model installed",
+      "install": "Install a model",
+      "ready": "Ready",
+      "seconds": "{{seconds}}s",
+      "describe": {
+        "images": "Text to image, image to image, reference images",
+        "video": "Text to video and image to video",
+        "threed": "Image to mesh reconstruction",
+        "tts": "Text to speech using your voice library",
+        "sound": "Music and sound effects from a prompt",
+        "transform": "Separation, enhancement and voice conversion"
+      }
+    },
+    "groups": {
+      "create": "Create",
+      "voice": "Voice",
+      "transform": "Transform"
     }
   },
   "image": {
@@ -204,5 +230,10 @@
     "clearMessage": "Hapus semua entri riwayat? Tindakan ini tidak dapat dibatalkan.",
     "clearConfirm": "Hapus",
     "cleared": "Riwayat dihapus"
+  },
+  "request": {
+    "heading": "Request",
+    "copyCurl": "Copy as curl",
+    "copied": "Copied"
   }
 }

core/http/react-ui/public/locales/id/models.json

@@ -11,7 +11,9 @@
     "installStarted": "Menginstal {{model}}…",
     "installFailed": "Instalasi gagal: {{message}}",
     "dismiss": "Tutup rekomendasi",
-    "summary": "{{n}} model disarankan"
+    "summary": "{{n}} model disarankan",
+    "bestFit": "Best fit",
+    "alternative": "Also fits"
   },
   "stats": {
     "available": "Tersedia",

core/http/react-ui/public/locales/id/nav.json

@@ -24,7 +24,9 @@
     "observability": "Observabilitas",
     "access": "Akses",
     "system": "Sistem",
-    "activity": "Activity"
+    "activity": "Activity",
+    "runtime": "Runtime",
+    "administration": "Administrasi"
   },
   "items": {
     "home": "Beranda",
@@ -57,7 +59,8 @@
     "system": "Sistem",
     "settings": "Pengaturan",
     "api": "API",
-    "activity": "Aktivitas"
+    "activity": "Aktivitas",
+    "overview": "Ikhtisar"
   },
   "footer": {
     "github": "GitHub",

core/http/react-ui/public/locales/it/admin.json

@@ -12,6 +12,8 @@
     "timeLeft": "{{value}} left",
     "cancel": "Cancel",
     "cancelLabel": "Cancel {{name}}",
+    "pause": "Pause",
+    "pauseLabel": "Pause {{name}} and keep downloaded data",
     "retry": "Retry",
     "retryLabel": "Retry {{name}}",
     "nodeCount": "{{count}} nodes",
@@ -143,5 +145,36 @@
   "explorer": {
     "title": "Esplora risorse",
     "subtitle": "Sfoglia file e configurazioni"
+  },
+  "operate": {
+    "overview": {
+      "title": "Overview",
+      "subtitle": "Everything running on this installation, and anything that wants a decision.",
+      "attention": {
+        "heading": "Needs attention",
+        "clear": "Nothing needs attention. Backends are current, no operation has failed, and every node is healthy.",
+        "backendUpdate": "Update available: {{from}} → {{to}}"
+      },
+      "sections": {
+        "heading": "Sections",
+        "runtime": "Runtime",
+        "runtimeSummary": "{{backends}} backends · {{models}} models · {{updates}} updates · {{running}} running",
+        "cluster": "Cluster",
+        "clusterSummary": "{{nodes}} nodes",
+        "observability": "Observability",
+        "observabilitySummary": "Usage and traces",
+        "administration": "Administration",
+        "administrationSummary": "Users, middleware and settings · {{memory}} memory in use",
+        "clusterSingle": "Single node",
+        "observabilityCounted": "{{requests}} requests · {{errors}} failed · p95 {{p95}} ms"
+      },
+      "headline": {
+        "requests": "Requests · {{hours}}h",
+        "errors": "Failed requests",
+        "p95": "p95 latency",
+        "quiet": "No requests served in this window yet.",
+        "host": "Host memory"
+      }
+    }
   }
 }

core/http/react-ui/public/locales/it/chat.json

@@ -118,5 +118,8 @@
     "newChat": "Nuova chat",
     "clearAll": "Cancella tutto",
     "deleteAllTitle": "Elimina tutte le conversazioni"
+  },
+  "message": {
+    "you": "You"
   }
 }

core/http/react-ui/public/locales/it/home.json

@@ -17,7 +17,9 @@
     "modelsLoaded_other": "{{count}} modelli caricati",
     "noModelsLoaded": "Nessun modello caricato",
     "nodes_one": "{{count}} nodo",
-    "nodes_other": "{{count}} nodi"
+    "nodes_other": "{{count}} nodi",
+    "loadedLabel": "Loaded",
+    "nodesLabel": "Nodes"
   },
   "assistant": {
     "title": "Gestisci LocalAI chattando",
@@ -47,7 +49,8 @@
     "count_one": "{{count}} modello caricato",
     "count_other": "{{count}} modelli caricati",
     "stop": "Ferma modello",
-    "stopAll": "Ferma tutti"
+    "stopAll": "Ferma tutti",
+    "serving": "Serving"
   },
   "stopDialog": {
     "title": "Ferma modello",
@@ -88,5 +91,14 @@
     "browse": "Esplora le API",
     "hide": "Nascondi gli endpoint",
     "dismiss": "Ignora"
+  },
+  "jump": {
+    "heading": "Jump back in",
+    "discover": "Discover",
+    "discoverSummary": "Browse the gallery and install models",
+    "create": "Create",
+    "createSummary": "Open a chat, image or voice session",
+    "operate": "Operate",
+    "operateSummary": "{{models}} models configured · nodes, activity and traces"
   }
 }

core/http/react-ui/public/locales/it/media.json

@@ -5,6 +5,32 @@
       "video": "Video",
       "tts": "TTS",
       "sound": "Audio",
+      "transform": "Transform",
+      "overview": "Overview"
+    },
+    "overview": {
+      "eyebrow": "{{ready}} of {{total}} modalities ready",
+      "title": "Studio",
+      "subtitle": "Generate images, video, 3D, speech and sound with the models on this machine.",
+      "canMake": "What you can make",
+      "running": "Running now",
+      "recent": "Recent outputs",
+      "noModel": "No model installed",
+      "install": "Install a model",
+      "ready": "Ready",
+      "seconds": "{{seconds}}s",
+      "describe": {
+        "images": "Text to image, image to image, reference images",
+        "video": "Text to video and image to video",
+        "threed": "Image to mesh reconstruction",
+        "tts": "Text to speech using your voice library",
+        "sound": "Music and sound effects from a prompt",
+        "transform": "Separation, enhancement and voice conversion"
+      }
+    },
+    "groups": {
+      "create": "Create",
+      "voice": "Voice",
       "transform": "Transform"
     }
   },
@@ -157,5 +183,10 @@
     "clearMessage": "Rimuovere tutte le voci della cronologia? Questa azione non può essere annullata.",
     "clearConfirm": "Cancella",
     "cleared": "Cronologia cancellata"
+  },
+  "request": {
+    "heading": "Request",
+    "copyCurl": "Copy as curl",
+    "copied": "Copied"
   }
 }

core/http/react-ui/public/locales/it/models.json

@@ -10,7 +10,9 @@
     "installStarted": "Installazione di {{model}}…",
     "installFailed": "Installazione non riuscita: {{message}}",
     "dismiss": "Nascondi i consigli",
-    "summary": "{{n}} modelli suggeriti"
+    "summary": "{{n}} modelli suggeriti",
+    "bestFit": "Best fit",
+    "alternative": "Also fits"
   },
   "stats": {
     "available": "Disponibili",

core/http/react-ui/public/locales/it/nav.json

@@ -24,7 +24,9 @@
     "observability": "Observability",
     "access": "Access",
     "system": "System",
-    "activity": "Activity"
+    "activity": "Activity",
+    "runtime": "Runtime",
+    "administration": "Amministrazione"
   },
   "items": {
     "home": "Home",
@@ -57,7 +59,8 @@
     "settings": "Impostazioni",
     "api": "API",
     "middleware": "Middleware",
-    "activity": "Attività"
+    "activity": "Attività",
+    "overview": "Panoramica"
   },
   "footer": {
     "github": "GitHub",