fix(ollama): accept float-encoded integer options (fixes #9837) (#9849)

fix(ollama): accept float-encoded integer options (num_ctx, top_k, ...)

Home Assistant's Ollama integration encodes integer options as JSON
floats (e.g. `"num_ctx": 8192.0`). Stdlib `json.Unmarshal` refuses to
decode a number with fractional notation into an `int` field, so the
entire request was rejected with HTTP 400 before reaching the backend:

  Unmarshal type error: expected=int, got=number 8192.0,
  field=options.num_ctx

Add a custom `UnmarshalJSON` on `OllamaOptions` that routes the int
fields (`top_k`, `num_predict`, `seed`, `repeat_last_n`, `num_ctx`)
through `*json.Number`, then converts via `Int64()` with a `Float64()`
fallback. Public field types are unchanged, so endpoint code is
untouched. Float fields and `stop` continue to parse via the default
path.

Fixes #9837

Assisted-by: Claude Code:claude-opus-4-7

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Co-authored-by: Ettore Di Giacinto <mudler@localai.io>

.agents/api-endpoints-and-auth.md

@@ -284,7 +284,17 @@ Also bump the expected-length count in `api_instructions_test.go` and add the na
 
 ### 3. `capabilities.js` symbol (for new model-config FLAG_* flags)
 
-If your feature needs a new `FLAG_*` usecase flag in `core/config/model_config.go` (so users can filter gallery models by it, and so `/v1/models` surfaces it), also declare the matching symbol in `core/http/react-ui/src/utils/capabilities.js`:
+If your feature needs a new `FLAG_*` usecase flag in `core/config/model_config.go` (so users can filter gallery models by it, and so `/v1/models` surfaces it), you need to update **all** of:
+
+- `Usecase<Name>` string constant in `core/config/backend_capabilities.go`
+- `UsecaseInfoMap` entry mapping the string to its flag + gRPC method
+- `FLAG_<NAME>` bitmask in `core/config/model_config.go`
+- `GetAllModelConfigUsecases()` map entry (otherwise the YAML loader silently ignores the string)
+- `ModalityGroups` membership if the flag should affect `IsMultimodal()` (e.g. realtime_audio is in both speech-input and audio-output groups so a lone flag still reads as multimodal)
+- `GuessUsecases()` branch listing the backends that own this capability
+- `usecaseFilters` in `core/http/routes/ui_api.go` (drives the gallery filter dropdown)
+- `Models.jsx` `FILTERS` array + matching `filters.<camelCase>` i18n key in `core/http/react-ui/public/locales/en/models.json`
+- `core/http/react-ui/src/utils/capabilities.js`:
 
 ```js
 export const CAP_MY_CAPABILITY = 'FLAG_MY_CAPABILITY'

.github/backend-matrix.yml

@@ -278,6 +278,19 @@ include:
     dockerfile: "./backend/Dockerfile.python"
     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-liquid-audio'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "liquid-audio"
+    dockerfile: "./backend/Dockerfile.python"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'cublas'
     cuda-major-version: "12"
     cuda-minor-version: "8"
@@ -808,6 +821,19 @@ include:
     dockerfile: "./backend/Dockerfile.python"
     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-liquid-audio'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "liquid-audio"
+    dockerfile: "./backend/Dockerfile.python"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'cublas'
     cuda-major-version: "13"
     cuda-minor-version: "0"
@@ -1088,6 +1114,19 @@ include:
     backend: "vibevoice"
     dockerfile: "./backend/Dockerfile.python"
     context: "./"
+  - build-type: 'l4t'
+    cuda-major-version: "13"
+    cuda-minor-version: "0"
+    platforms: 'linux/arm64'
+    tag-latest: 'auto'
+    tag-suffix: '-nvidia-l4t-cuda-13-arm64-liquid-audio'
+    runs-on: 'ubuntu-24.04-arm'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    ubuntu-version: '2404'
+    backend: "liquid-audio"
+    dockerfile: "./backend/Dockerfile.python"
+    context: "./"
   - build-type: 'l4t'
     cuda-major-version: "13"
     cuda-minor-version: "0"
@@ -1729,6 +1768,19 @@ include:
     dockerfile: "./backend/Dockerfile.python"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: 'hipblas'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-rocm-hipblas-liquid-audio'
+    runs-on: 'ubuntu-latest'
+    base-image: "rocm/dev-ubuntu-24.04:7.2.1"
+    skip-drivers: 'false'
+    backend: "liquid-audio"
+    dockerfile: "./backend/Dockerfile.python"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'hipblas'
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -2177,6 +2229,19 @@ include:
     dockerfile: "./backend/Dockerfile.python"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: 'intel'
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-gpu-intel-liquid-audio'
+    runs-on: 'ubuntu-latest'
+    base-image: "intel/oneapi-basekit:2025.3.0-0-devel-ubuntu24.04"
+    skip-drivers: 'false'
+    backend: "liquid-audio"
+    dockerfile: "./backend/Dockerfile.python"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: 'intel'
     cuda-major-version: ""
     cuda-minor-version: ""
@@ -3503,6 +3568,20 @@ include:
     dockerfile: "./backend/Dockerfile.python"
     context: "./"
     ubuntu-version: '2404'
+  - build-type: ''
+    cuda-major-version: ""
+    cuda-minor-version: ""
+    platforms: 'linux/amd64'
+    platform-tag: 'amd64'
+    tag-latest: 'auto'
+    tag-suffix: '-cpu-liquid-audio'
+    runs-on: 'ubuntu-latest'
+    base-image: "ubuntu:24.04"
+    skip-drivers: 'false'
+    backend: "liquid-audio"
+    dockerfile: "./backend/Dockerfile.python"
+    context: "./"
+    ubuntu-version: '2404'
   - build-type: ''
     cuda-major-version: ""
     cuda-minor-version: ""

.github/workflows/image.yml

@@ -151,7 +151,11 @@
               ubuntu-codename: 'noble'
 
     core-image-merge:
-      if: github.repository == 'mudler/LocalAI'
+      # !cancelled(): without it, GHA's default `needs:` cascade skips the
+      # merge whenever any matrix cell of the parent build fails or is
+      # cancelled. Same fix as backend.yml's merge jobs — we still want to
+      # publish the manifest list for tag-suffixes whose legs all succeeded.
+      if: ${{ !cancelled() && github.repository == 'mudler/LocalAI' }}
       needs: core-image-build
       uses: ./.github/workflows/image_merge.yml
       with:
@@ -164,7 +168,7 @@
         quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
 
     gpu-vulkan-image-merge:
-      if: github.repository == 'mudler/LocalAI'
+      if: ${{ !cancelled() && github.repository == 'mudler/LocalAI' }}
       needs: core-image-build
       uses: ./.github/workflows/image_merge.yml
       with:
@@ -175,7 +179,91 @@
         dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
         quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
         quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
-  
+
+    # Single-arch server-image merges. Same conceptual fix as the backend
+    # singletons in PR #9781: image_build.yml pushes by canonical digest
+    # only, so without a downstream merge step there's no tag for consumers
+    # (no :latest-gpu-nvidia-cuda-12, no :v<X>-gpu-nvidia-cuda-12, etc.).
+    # Each merge job needs only its parent build matrix and is filtered by
+    # tag-suffix in image_merge.yml's artifact-download pattern.
+    gpu-nvidia-cuda-12-image-merge:
+      if: ${{ !cancelled() && github.repository == 'mudler/LocalAI' }}
+      needs: core-image-build
+      uses: ./.github/workflows/image_merge.yml
+      with:
+        tag-latest: 'auto'
+        tag-suffix: '-gpu-nvidia-cuda-12'
+      secrets:
+        dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+        dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+        quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+        quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+
+    gpu-nvidia-cuda-13-image-merge:
+      if: ${{ !cancelled() && github.repository == 'mudler/LocalAI' }}
+      needs: core-image-build
+      uses: ./.github/workflows/image_merge.yml
+      with:
+        tag-latest: 'auto'
+        tag-suffix: '-gpu-nvidia-cuda-13'
+      secrets:
+        dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+        dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+        quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+        quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+
+    gpu-intel-image-merge:
+      if: ${{ !cancelled() && github.repository == 'mudler/LocalAI' }}
+      needs: core-image-build
+      uses: ./.github/workflows/image_merge.yml
+      with:
+        tag-latest: 'auto'
+        tag-suffix: '-gpu-intel'
+      secrets:
+        dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+        dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+        quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+        quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+
+    gpu-hipblas-image-merge:
+      if: ${{ !cancelled() && github.repository == 'mudler/LocalAI' }}
+      needs: hipblas-jobs
+      uses: ./.github/workflows/image_merge.yml
+      with:
+        tag-latest: 'auto'
+        tag-suffix: '-gpu-hipblas'
+      secrets:
+        dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+        dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+        quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+        quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+
+    nvidia-l4t-arm64-image-merge:
+      if: ${{ !cancelled() && github.repository == 'mudler/LocalAI' }}
+      needs: gh-runner
+      uses: ./.github/workflows/image_merge.yml
+      with:
+        tag-latest: 'auto'
+        tag-suffix: '-nvidia-l4t-arm64'
+      secrets:
+        dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+        dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+        quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+        quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+
+    nvidia-l4t-arm64-cuda-13-image-merge:
+      if: ${{ !cancelled() && github.repository == 'mudler/LocalAI' }}
+      needs: gh-runner
+      uses: ./.github/workflows/image_merge.yml
+      with:
+        tag-latest: 'auto'
+        tag-suffix: '-nvidia-l4t-arm64-cuda-13'
+      secrets:
+        dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+        dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+        quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+        quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+
     gh-runner:
       if: github.repository == 'mudler/LocalAI'
       uses: ./.github/workflows/image_build.yml

.github/workflows/image_build.yml

@@ -185,11 +185,28 @@ jobs:
           digest="${{ steps.build.outputs.digest }}"
           touch "/tmp/digests/${digest#sha256:}"
 
+      # See .github/scripts/anchor-digest-in-cache.sh for why this is needed
+      # and how it interacts with image_merge.yml's cleanup step. Mirrors the
+      # same anchor in backend_build.yml — quay's per-repo manifest GC reaps
+      # untagged manifests in local-ai before the merge runs.
+      - name: Anchor digest in ci-cache so quay GC won't reap before merge
+        if: github.event_name != 'pull_request'
+        env:
+          TAG_SUFFIX: ${{ inputs.tag-suffix == '' && '-core' || inputs.tag-suffix }}
+          PLATFORM_TAG: ${{ inputs.platform-tag || 'single' }}
+          DIGEST: ${{ steps.build.outputs.digest }}
+          SOURCE_IMAGE: quay.io/go-skynet/local-ai
+        run: .github/scripts/anchor-digest-in-cache.sh
+
       - name: Upload digest artifact
         if: github.event_name != 'pull_request'
         uses: actions/upload-artifact@v7
         with:
-          name: digests-localai${{ inputs.tag-suffix == '' && '-core' || inputs.tag-suffix }}-${{ inputs.platform-tag }}
+          # `--` separator + 'single' placeholder for empty platform-tag —
+          # same pattern as backend_build.yml. Prevents prefix collisions
+          # in the merge-side glob (e.g. -nvidia-l4t-arm64 is a prefix of
+          # -nvidia-l4t-arm64-cuda-13).
+          name: digests-localai${{ inputs.tag-suffix == '' && '-core' || inputs.tag-suffix }}--${{ inputs.platform-tag || 'single' }}
           path: /tmp/digests/*
           if-no-files-found: error
           retention-days: 1

.github/workflows/image_merge.yml

@@ -33,10 +33,22 @@ jobs:
     env:
       quay_username: ${{ secrets.quayUsername }}
     steps:
+      # Sparse checkout: needed for .github/scripts/ (the keepalive cleanup
+      # script). Skips the rest of the source tree.
+      - name: Checkout (.github/scripts only)
+        uses: actions/checkout@v6
+        with:
+          sparse-checkout: |
+            .github/scripts
+          sparse-checkout-cone-mode: false
+
       - name: Download digests
         uses: actions/download-artifact@v8
         with:
-          pattern: digests-localai${{ inputs.tag-suffix == '' && '-core' || inputs.tag-suffix }}-*
+          # `--` separator anchors the glob so we don't over-match sibling
+          # tag-suffixes (e.g. -nvidia-l4t-arm64 vs -nvidia-l4t-arm64-cuda-13).
+          # Must stay in sync with image_build.yml's upload-artifact name.
+          pattern: digests-localai${{ inputs.tag-suffix == '' && '-core' || inputs.tag-suffix }}--*
           merge-multiple: true
           path: /tmp/digests
 
@@ -72,6 +84,13 @@ jobs:
             latest=${{ inputs.tag-latest }}
             suffix=${{ inputs.tag-suffix }},onlatest=true
 
+      # Source from ci-cache, not local-ai. See backend_merge.yml for the
+      # detailed rationale — quay's manifest GC is per-repository, so the
+      # untagged digest in local-ai gets reaped while the same content lives
+      # tagged under ci-cache (anchored by image_build.yml). buildx imagetools
+      # create copies the manifest into local-ai (blobs already cross-mounted)
+      # and publishes the manifest list with user-facing tags. End state in
+      # local-ai is self-contained; no embedded reference to ci-cache.
       - name: Create manifest list and push (quay)
         working-directory: /tmp/digests
         run: |
@@ -82,7 +101,7 @@ jobs:
           else
             # shellcheck disable=SC2086
             docker buildx imagetools create $tags \
-              $(printf 'quay.io/go-skynet/local-ai@sha256:%s ' *)
+              $(printf 'quay.io/go-skynet/ci-cache@sha256:%s ' *)
           fi
 
       - name: Create manifest list and push (dockerhub)
@@ -107,6 +126,15 @@ jobs:
             docker buildx imagetools inspect "$first_tag"
           fi
 
+      # See .github/scripts/cleanup-keepalive-tags.sh for the best-effort
+      # semantics — fails soft when the registry credential isn't OAuth-scoped.
+      - name: Cleanup keepalive tags in ci-cache
+        if: github.event_name != 'pull_request' && success()
+        env:
+          TAG_SUFFIX: ${{ inputs.tag-suffix == '' && '-core' || inputs.tag-suffix }}
+          QUAY_TOKEN: ${{ secrets.quayPassword }}
+        run: .github/scripts/cleanup-keepalive-tags.sh
+
       - name: Job summary
         run: |
           set -euo pipefail

.github/workflows/test-extra.yml

@@ -28,6 +28,7 @@ jobs:
       qwen-asr: ${{ steps.detect.outputs.qwen-asr }}
       nemo: ${{ steps.detect.outputs.nemo }}
       voxcpm: ${{ steps.detect.outputs.voxcpm }}
+      liquid-audio: ${{ steps.detect.outputs.liquid-audio }}
       llama-cpp-quantization: ${{ steps.detect.outputs.llama-cpp-quantization }}
       llama-cpp: ${{ steps.detect.outputs.llama-cpp }}
       ik-llama-cpp: ${{ steps.detect.outputs.ik-llama-cpp }}
@@ -447,6 +448,32 @@ jobs:
         run: |
           make --jobs=5 --output-sync=target -C backend/python/voxcpm
           make --jobs=5 --output-sync=target -C backend/python/voxcpm test
+  # liquid-audio: LFM2.5-Audio any-to-any backend. The CI smoke test
+  # exercises Health() and LoadModel(mode:finetune) — fine-tune mode
+  # short-circuits before pulling weights (backend.py:192), so no
+  # HuggingFace download or GPU is needed. The full-inference path is
+  # gated on LIQUID_AUDIO_MODEL_ID, which we don't set here.
+  tests-liquid-audio:
+    needs: detect-changes
+    if: needs.detect-changes.outputs.liquid-audio == 'true' || needs.detect-changes.outputs.run-all == 'true'
+    runs-on: ubuntu-latest
+    steps:
+      - name: Clone
+        uses: actions/checkout@v6
+        with:
+          submodules: true
+      - name: Dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y build-essential ffmpeg
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          pip install --user --no-cache-dir grpcio-tools==1.64.1
+      - name: Test liquid-audio
+        run: |
+          make --jobs=5 --output-sync=target -C backend/python/liquid-audio
+          make --jobs=5 --output-sync=target -C backend/python/liquid-audio test
   tests-llama-cpp-quantization:
     needs: detect-changes
     if: needs.detect-changes.outputs.llama-cpp-quantization == 'true' || needs.detect-changes.outputs.run-all == 'true'

Makefile

@@ -1,5 +1,5 @@
 # Disable parallel execution for backend builds
-.NOTPARALLEL: backends/diffusers backends/llama-cpp backends/turboquant backends/outetts backends/piper backends/stablediffusion-ggml backends/whisper backends/faster-whisper backends/silero-vad backends/local-store backends/huggingface backends/rfdetr 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/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/vibevoice-cpp backends/localvqe backends/tinygrad backends/sherpa-onnx backends/ds4 backends/ds4-darwin
+.NOTPARALLEL: backends/diffusers backends/llama-cpp backends/turboquant backends/outetts backends/piper backends/stablediffusion-ggml backends/whisper backends/faster-whisper backends/silero-vad backends/local-store backends/huggingface backends/rfdetr 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/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/vibevoice-cpp backends/localvqe backends/tinygrad backends/sherpa-onnx backends/ds4 backends/ds4-darwin backends/liquid-audio
 
 GOCMD=go
 GOTEST=$(GOCMD) test
@@ -463,6 +463,7 @@ prepare-test-extra: protogen-python
 	$(MAKE) -C backend/python/vllm-omni
 	$(MAKE) -C backend/python/sglang
 	$(MAKE) -C backend/python/vibevoice
+	$(MAKE) -C backend/python/liquid-audio
 	$(MAKE) -C backend/python/moonshine
 	$(MAKE) -C backend/python/pocket-tts
 	$(MAKE) -C backend/python/qwen-tts
@@ -488,6 +489,7 @@ test-extra: prepare-test-extra
 	$(MAKE) -C backend/python/vllm test
 	$(MAKE) -C backend/python/vllm-omni test
 	$(MAKE) -C backend/python/vibevoice test
+	$(MAKE) -C backend/python/liquid-audio test
 	$(MAKE) -C backend/python/moonshine test
 	$(MAKE) -C backend/python/pocket-tts test
 	$(MAKE) -C backend/python/qwen-tts test
@@ -1092,6 +1094,7 @@ BACKEND_SGLANG = sglang|python|.|false|true
 BACKEND_DIFFUSERS = diffusers|python|.|--progress=plain|true
 BACKEND_CHATTERBOX = chatterbox|python|.|false|true
 BACKEND_VIBEVOICE = vibevoice|python|.|--progress=plain|true
+BACKEND_LIQUID_AUDIO = liquid-audio|python|.|--progress=plain|true
 BACKEND_MOONSHINE = moonshine|python|.|false|true
 BACKEND_POCKET_TTS = pocket-tts|python|.|false|true
 BACKEND_QWEN_TTS = qwen-tts|python|.|false|true
@@ -1169,6 +1172,7 @@ $(eval $(call generate-docker-build-target,$(BACKEND_SGLANG)))
 $(eval $(call generate-docker-build-target,$(BACKEND_DIFFUSERS)))
 $(eval $(call generate-docker-build-target,$(BACKEND_CHATTERBOX)))
 $(eval $(call generate-docker-build-target,$(BACKEND_VIBEVOICE)))
+$(eval $(call generate-docker-build-target,$(BACKEND_LIQUID_AUDIO)))
 $(eval $(call generate-docker-build-target,$(BACKEND_MOONSHINE)))
 $(eval $(call generate-docker-build-target,$(BACKEND_POCKET_TTS)))
 $(eval $(call generate-docker-build-target,$(BACKEND_QWEN_TTS)))
@@ -1197,7 +1201,7 @@ $(eval $(call generate-docker-build-target,$(BACKEND_SHERPA_ONNX)))
 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-ds4 docker-build-rerankers docker-build-vllm docker-build-vllm-omni docker-build-sglang docker-build-transformers docker-build-outetts docker-build-diffusers docker-build-kokoro docker-build-faster-whisper docker-build-coqui docker-build-chatterbox docker-build-vibevoice 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-qwen3-tts-cpp docker-build-vibevoice-cpp docker-build-localvqe docker-build-insightface docker-build-speaker-recognition docker-build-sherpa-onnx
+docker-build-backends: docker-build-llama-cpp docker-build-ik-llama-cpp docker-build-turboquant docker-build-ds4 docker-build-rerankers docker-build-vllm docker-build-vllm-omni docker-build-sglang docker-build-transformers docker-build-outetts docker-build-diffusers docker-build-kokoro docker-build-faster-whisper 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-qwen3-tts-cpp docker-build-vibevoice-cpp docker-build-localvqe docker-build-insightface docker-build-speaker-recognition docker-build-sherpa-onnx
 
 ########################################################
 ### Mock Backend for E2E Tests

backend/backend.proto

@@ -48,6 +48,11 @@ service Backend {
 
   rpc AudioTransform(AudioTransformRequest) returns (AudioTransformResult) {}
   rpc AudioTransformStream(stream AudioTransformFrameRequest) returns (stream AudioTransformFrameResponse) {}
+  // AudioToAudioStream is the bidirectional any-to-any S2S RPC. Backends
+  // that load a speech-to-speech model consume input audio frames and emit
+  // interleaved audio + transcript + tool-call deltas as typed events.
+  // Backends without S2S support return UNIMPLEMENTED.
+  rpc AudioToAudioStream(stream AudioToAudioRequest) returns (stream AudioToAudioResponse) {}
 
   rpc ModelMetadata(ModelOptions) returns (ModelMetadataResponse) {}
 
@@ -768,6 +773,93 @@ message AudioTransformFrameResponse {
   int64 frame_index = 2;
 }
 
+// === AudioToAudioStream messages =========================================
+//
+// Bidirectional stream between the LocalAI core and an any-to-any audio
+// model. The client opens the stream with a Config payload, then alternates
+// Frame (input audio) and Control (turn boundaries, function-call results,
+// session updates) payloads. The server streams back typed events: audio
+// frames carry PCM in `pcm`; transcript / tool-call deltas carry JSON in
+// `meta`; the stream ends with a `response.done` (success) or `error` event.
+
+message AudioToAudioRequest {
+  oneof payload {
+    AudioToAudioConfig  config  = 1;
+    AudioToAudioFrame   frame   = 2;
+    AudioToAudioControl control = 3;
+  }
+}
+
+message AudioToAudioConfig {
+  // PCM format for client→server audio. 0 => backend default
+  // (16 kHz for the LFM2-Audio Conformer encoder).
+  int32 input_sample_rate = 1;
+  // Preferred server→client audio rate. 0 => backend default
+  // (24 kHz for the LFM2-Audio vocoder).
+  int32 output_sample_rate = 2;
+  // Optional system prompt override. Empty => backend chooses based on
+  // mode (e.g. "Respond with interleaved text and audio.").
+  string system_prompt = 3;
+  // Optional baked-voice id. Models that only ship a fixed set of
+  // voices (e.g. LFM2-Audio: us_male/us_female/uk_male/uk_female) match
+  // this against their voice table; an empty string keeps the default.
+  string voice = 4;
+  // JSON-encoded array of tool definitions in OpenAI Chat Completions
+  // format. Empty => no tools.
+  string tools = 5;
+  // Free-form sampling / decoding parameters (temperature, top_k,
+  // max_new_tokens, audio_top_k, etc).
+  map<string, string> params = 6;
+  // True => reset any session-scoped state before processing further
+  // frames on this stream. The first Config implicitly resets.
+  bool reset = 7;
+}
+
+message AudioToAudioFrame {
+  // Raw PCM s16le mono at config.input_sample_rate. Empty pcm + end_of_input
+  // is a valid "user finished speaking" marker without trailing audio.
+  bytes pcm = 1;
+  // Marks the last frame of a user turn. The backend may begin emitting
+  // a response immediately after seeing this.
+  bool end_of_input = 2;
+}
+
+message AudioToAudioControl {
+  // Free-form control event names. Initial set:
+  //   "input_audio_buffer.commit"     — user finished speaking
+  //   "response.cancel"               — abort in-flight generation
+  //   "conversation.item.create"      — inject a non-audio item (e.g.
+  //                                     function_call_output as JSON in
+  //                                     `payload`)
+  //   "session.update"                — re-configure mid-stream
+  string event = 1;
+  // Event-specific JSON payload.
+  bytes payload = 2;
+}
+
+message AudioToAudioResponse {
+  // Event identifies what this frame carries. Mirrors the OpenAI Realtime
+  // API server-event names where applicable. Initial set:
+  //   "response.audio.delta"
+  //   "response.audio_transcript.delta"
+  //   "response.function_call_arguments.delta"
+  //   "response.function_call_arguments.done"
+  //   "response.done"
+  //   "error"
+  string event = 1;
+  // Populated when event = response.audio.delta.
+  bytes pcm = 2;
+  // Populated alongside pcm to identify its rate. 0 => same as the
+  // session's negotiated output_sample_rate.
+  int32 sample_rate = 3;
+  // JSON payload for non-PCM events (transcript chunk, tool args, error
+  // body).
+  bytes meta = 4;
+  // Monotonic per-stream counter, useful for client reordering and
+  // debugging.
+  int64 sequence = 5;
+}
+
 message ModelMetadataResponse {
   bool supports_thinking = 1;
   string rendered_template = 2;  // The rendered chat template with enable_thinking=true (empty if not applicable)

backend/cpp/ds4/Makefile

@@ -1,10 +1,10 @@
 # ds4 backend Makefile.
 #
-# Upstream pin lives below as DS4_VERSION?=f8b4ed635d559b3a5b44bf2df6a77e21b3e9178f
+# Upstream pin lives below as DS4_VERSION?=950e8e6474a1c9fabe04e669d607606a7ef8824f
 # (.github/bump_deps.sh) can find and update it - matches the
 # llama-cpp / ik-llama-cpp / turboquant convention.
 
-DS4_VERSION?=f8b4ed635d559b3a5b44bf2df6a77e21b3e9178f
+DS4_VERSION?=950e8e6474a1c9fabe04e669d607606a7ef8824f
 DS4_REPO?=https://github.com/antirez/ds4
 
 CURRENT_MAKEFILE_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))

backend/cpp/ik-llama-cpp/Makefile

@@ -1,5 +1,5 @@
 
-IK_LLAMA_VERSION?=f9a93c37e2fc021760c3c1aa99cf74c73b7591a7
+IK_LLAMA_VERSION?=5cc0d86c760e9858e4bed4418400bb39dbe025f2
 LLAMA_REPO?=https://github.com/ikawrakow/ik_llama.cpp
 
 CMAKE_ARGS?=

backend/cpp/llama-cpp/Makefile

@@ -1,5 +1,5 @@
 
-LLAMA_VERSION?=1ec7ba0c14f33f17e980daeeda5f35b225d41994
+LLAMA_VERSION?=1348f67c58f561808136e8a152a9eddec168f221
 LLAMA_REPO?=https://github.com/ggerganov/llama.cpp
 
 CMAKE_ARGS?=

backend/cpp/llama-cpp/grpc-server.cpp

@@ -32,6 +32,7 @@
 #include <grpcpp/health_check_service_interface.h>
 #include <grpcpp/security/server_credentials.h>
 #include <regex>
+#include <algorithm>
 #include <atomic>
 #include <cstdlib>
 #include <fstream>
@@ -450,6 +451,8 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
         // vector; the turboquant fork still uses the legacy scalar. The
         // LOCALAI_LEGACY_LLAMA_CPP_SPEC macro is injected by
         // backend/cpp/turboquant/patch-grpc-server.sh for fork builds only.
+        // Upstream renamed COMMON_SPECULATIVE_TYPE_DRAFT -> ..._DRAFT_SIMPLE
+        // in ggml-org/llama.cpp#22964; the fork still uses the old name.
 #ifdef LOCALAI_LEGACY_LLAMA_CPP_SPEC
         if (params.speculative.type == COMMON_SPECULATIVE_TYPE_NONE) {
             params.speculative.type = COMMON_SPECULATIVE_TYPE_DRAFT;
@@ -458,7 +461,7 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
         const bool no_spec_type = params.speculative.types.empty() ||
             (params.speculative.types.size() == 1 && params.speculative.types[0] == COMMON_SPECULATIVE_TYPE_NONE);
         if (no_spec_type) {
-            params.speculative.types = { COMMON_SPECULATIVE_TYPE_DRAFT };
+            params.speculative.types = { COMMON_SPECULATIVE_TYPE_DRAFT_SIMPLE };
         }
 #endif
     }
@@ -685,6 +688,136 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
                     // If conversion fails, keep default value (8)
                 }
             }
+
+        // --- physical batch size (upstream -ub / --ubatch-size) ---
+        // Note: line ~482 already aliases n_ubatch to n_batch as a default; this
+        // option lets users decouple the two (useful for embeddings/rerank).
+        } else if (!strcmp(optname, "n_ubatch") || !strcmp(optname, "ubatch")) {
+            if (optval != NULL) {
+                try { params.n_ubatch = std::stoi(optval_str); } catch (...) {}
+            }
+
+        // --- main-model batch threads (upstream -tb / --threads-batch) ---
+        } else if (!strcmp(optname, "threads_batch") || !strcmp(optname, "n_threads_batch")) {
+            if (optval != NULL) {
+                try {
+                    int n = std::stoi(optval_str);
+                    if (n <= 0) n = (int)std::thread::hardware_concurrency();
+                    params.cpuparams_batch.n_threads = n;
+                } catch (...) {}
+            }
+
+        // --- pooling type for embeddings (upstream --pooling) ---
+        } else if (!strcmp(optname, "pooling_type") || !strcmp(optname, "pooling")) {
+            if (optval != NULL) {
+                if      (optval_str == "none") params.pooling_type = LLAMA_POOLING_TYPE_NONE;
+                else if (optval_str == "mean") params.pooling_type = LLAMA_POOLING_TYPE_MEAN;
+                else if (optval_str == "cls")  params.pooling_type = LLAMA_POOLING_TYPE_CLS;
+                else if (optval_str == "last") params.pooling_type = LLAMA_POOLING_TYPE_LAST;
+                else if (optval_str == "rank") params.pooling_type = LLAMA_POOLING_TYPE_RANK;
+                // unknown values silently leave UNSPECIFIED (auto-detect)
+            }
+
+        // --- llama log verbosity threshold (upstream -lv / --verbosity) ---
+        } else if (!strcmp(optname, "verbosity")) {
+            if (optval != NULL) {
+                try { params.verbosity = std::stoi(optval_str); } catch (...) {}
+            }
+
+        // --- O_DIRECT model loading (upstream --direct-io) ---
+        } else if (!strcmp(optname, "direct_io") || !strcmp(optname, "use_direct_io")) {
+            if (optval_str == "true" || optval_str == "1" || optval_str == "yes" || optval_str == "on" || optval_str == "enabled") {
+                params.use_direct_io = true;
+            } else if (optval_str == "false" || optval_str == "0" || optval_str == "no" || optval_str == "off" || optval_str == "disabled") {
+                params.use_direct_io = false;
+            }
+
+        // --- embedding normalization (upstream --embd-normalize) ---
+        // -1 none, 0 max-abs, 1 taxicab, 2 L2 (default), >2 p-norm
+        } else if (!strcmp(optname, "embd_normalize") || !strcmp(optname, "embedding_normalize")) {
+            if (optval != NULL) {
+                try { params.embd_normalize = std::stoi(optval_str); } catch (...) {}
+            }
+
+        // --- reasoning parser (upstream --reasoning-format) ---
+        // Picks the parser for <think> blocks emitted by reasoning models.
+        // none / auto / deepseek / deepseek-legacy
+        } else if (!strcmp(optname, "reasoning_format")) {
+            if (optval != NULL) {
+                if      (optval_str == "none")             params.reasoning_format = COMMON_REASONING_FORMAT_NONE;
+                else if (optval_str == "auto")             params.reasoning_format = COMMON_REASONING_FORMAT_AUTO;
+                else if (optval_str == "deepseek")         params.reasoning_format = COMMON_REASONING_FORMAT_DEEPSEEK;
+                else if (optval_str == "deepseek-legacy" || optval_str == "deepseek_legacy")
+                                                            params.reasoning_format = COMMON_REASONING_FORMAT_DEEPSEEK_LEGACY;
+                // unknown values silently keep the upstream default (DEEPSEEK)
+            }
+
+        // --- reasoning budget (upstream --reasoning-budget) ---
+        // -1 unlimited, 0 disabled, >0 token budget for thinking blocks.
+        // Distinct from per-request `enable_thinking` (chat_template_kwargs).
+        } else if (!strcmp(optname, "enable_reasoning") || !strcmp(optname, "reasoning_budget")) {
+            if (optval != NULL) {
+                try { params.enable_reasoning = std::stoi(optval_str); } catch (...) {}
+            }
+
+        // --- prefill assistant turn (upstream --no-prefill-assistant) ---
+        } else if (!strcmp(optname, "prefill_assistant")) {
+            if (optval_str == "true" || optval_str == "1" || optval_str == "yes" || optval_str == "on" || optval_str == "enabled") {
+                params.prefill_assistant = true;
+            } else if (optval_str == "false" || optval_str == "0" || optval_str == "no" || optval_str == "off" || optval_str == "disabled") {
+                params.prefill_assistant = false;
+            }
+
+        // --- mmproj GPU offload (upstream --no-mmproj-offload, inverted) ---
+        } else if (!strcmp(optname, "mmproj_use_gpu") || !strcmp(optname, "mmproj_offload")) {
+            if (optval_str == "true" || optval_str == "1" || optval_str == "yes" || optval_str == "on" || optval_str == "enabled") {
+                params.mmproj_use_gpu = true;
+            } else if (optval_str == "false" || optval_str == "0" || optval_str == "no" || optval_str == "off" || optval_str == "disabled") {
+                params.mmproj_use_gpu = false;
+            }
+
+        // --- per-image vision token budget (upstream --image-min/max-tokens) ---
+        } else if (!strcmp(optname, "image_min_tokens")) {
+            if (optval != NULL) {
+                try { params.image_min_tokens = std::stoi(optval_str); } catch (...) {}
+            }
+        } else if (!strcmp(optname, "image_max_tokens")) {
+            if (optval != NULL) {
+                try { params.image_max_tokens = std::stoi(optval_str); } catch (...) {}
+            }
+
+        // --- main-model tensor buffer overrides (upstream --override-tensor) ---
+        // Format: <tensor regex>=<buffer type>,<tensor regex>=<buffer type>,...
+        // Mirrors the existing `draft_override_tensor` parser below.
+        } else if (!strcmp(optname, "override_tensor") || !strcmp(optname, "tensor_buft_overrides")) {
+            ggml_backend_load_all();
+            std::map<std::string, ggml_backend_buffer_type_t> buft_list;
+            for (size_t i = 0; i < ggml_backend_dev_count(); ++i) {
+                auto * dev = ggml_backend_dev_get(i);
+                auto * buft = ggml_backend_dev_buffer_type(dev);
+                if (buft) {
+                    buft_list[ggml_backend_buft_name(buft)] = buft;
+                }
+            }
+            static std::list<std::string> override_names;
+            std::string cur;
+            auto flush = [&](const std::string & spec) {
+                auto pos = spec.find('=');
+                if (pos == std::string::npos) return;
+                const std::string name = spec.substr(0, pos);
+                const std::string type = spec.substr(pos + 1);
+                auto it = buft_list.find(type);
+                if (it == buft_list.end()) return; // unknown buffer type: ignore
+                override_names.push_back(name);
+                params.tensor_buft_overrides.push_back(
+                    {override_names.back().c_str(), it->second});
+            };
+            for (char c : optval_str) {
+                if (c == ',') { if (!cur.empty()) { flush(cur); cur.clear(); } }
+                else { cur.push_back(c); }
+            }
+            if (!cur.empty()) flush(cur);
+
         // Speculative decoding options
         } else if (!strcmp(optname, "spec_type") || !strcmp(optname, "speculative_type")) {
 #ifdef LOCALAI_LEGACY_LLAMA_CPP_SPEC
@@ -701,16 +834,27 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
             // Upstream switched to a vector of types (comma-separated for multi-type
             // chaining via common_speculative_types_from_names). We keep accepting a
             // single value here, but also tolerate comma-separated lists.
+            //
+            // ggml-org/llama.cpp#22964 also renamed the registered names from
+            // underscore- to dash-separated form, and replaced the bare
+            // `draft`/`eagle3` aliases with `draft-simple`/`draft-eagle3`. We
+            // normalize each token here so existing model configs keep working.
+            auto normalize_spec_name = [](std::string s) -> std::string {
+                std::replace(s.begin(), s.end(), '_', '-');
+                if (s == "draft")  return "draft-simple";
+                if (s == "eagle3") return "draft-eagle3";
+                return s;
+            };
             std::vector<std::string> names;
             std::string item;
             for (char c : optval_str) {
                 if (c == ',') {
-                    if (!item.empty()) { names.push_back(item); item.clear(); }
+                    if (!item.empty()) { names.push_back(normalize_spec_name(item)); item.clear(); }
                 } else {
                     item.push_back(c);
                 }
             }
-            if (!item.empty()) names.push_back(item);
+            if (!item.empty()) names.push_back(normalize_spec_name(item));
             auto parsed = common_speculative_types_from_names(names);
             if (!parsed.empty()) {
                 params.speculative.types = parsed;
@@ -2794,7 +2938,9 @@ public:
             }
         }
 
-        int embd_normalize = 2; // default to Euclidean/L2 norm
+        // Honor the load-time embd_normalize set via options:embd_normalize.
+        // -1 none, 0 max-abs, 1 taxicab, 2 L2 (default), >2 p-norm.
+        int embd_normalize = params_base.embd_normalize;
         // create and queue the task
         auto rd = ctx_server.get_response_reader();
         {

backend/cpp/turboquant/Makefile

@@ -1,7 +1,7 @@
 
 # Pinned to the HEAD of feature/turboquant-kv-cache on https://github.com/TheTom/llama-cpp-turboquant.
 # Auto-bumped nightly by .github/workflows/bump_deps.yaml.
-TURBOQUANT_VERSION?=69d8e4be47243e83b3d0d71e932bc7aa61c644dc
+TURBOQUANT_VERSION?=5aeb2fdbe26cd4c534c6fa15de73cb5749bd0403
 LLAMA_REPO?=https://github.com/TheTom/llama-cpp-turboquant
 
 CMAKE_ARGS?=

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?=90e87bc846f17059771efb8aaa31e9ef0cab6f78
+STABLEDIFFUSION_GGML_VERSION?=0b8296915c4094090cff6bd2e09a5e98288c3c7d
 
 CMAKE_ARGS+=-DGGML_MAX_NAME=128
 

backend/go/whisper/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggml-org/whisper.cpp
-WHISPER_CPP_VERSION?=338cce1e58133261753243802a0e7a430118866d
+WHISPER_CPP_VERSION?=968eebe77225d25e57a3f981da7c696310f0e881
 SO_TARGET?=libgowhisper.so
 
 CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF

backend/index.yaml

@@ -847,6 +847,35 @@
     nvidia-l4t-cuda-12: "nvidia-l4t-vibevoice"
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-vibevoice"
   icon: https://avatars.githubusercontent.com/u/6154722?s=200&v=4
+- &liquid-audio
+  urls:
+    - https://github.com/Liquid4All/liquid-audio
+    - https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B
+  description: |
+    LiquidAI LFM2 / LFM2.5 Audio Python backend. End-to-end speech-to-speech, ASR,
+    TTS (4 baked voices), and text chat from a single 1.5B model. Wraps the
+    upstream `liquid-audio` package; supports fine-tuning via LocalAI's
+    /v1/fine-tuning/jobs endpoint.
+  tags:
+    - speech-to-speech
+    - any-to-any
+    - text-to-speech
+    - speech-to-text
+    - TTS
+    - ASR
+    - realtime
+  license: LFM-Open-License-v1.0
+  name: "liquid-audio"
+  alias: "liquid-audio"
+  capabilities:
+    nvidia: "cuda12-liquid-audio"
+    intel: "intel-liquid-audio"
+    amd: "rocm-liquid-audio"
+    default: "cpu-liquid-audio"
+    nvidia-cuda-13: "cuda13-liquid-audio"
+    nvidia-cuda-12: "cuda12-liquid-audio"
+    nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-liquid-audio"
+  icon: https://cdn-avatars.huggingface.co/v1/production/uploads/61b8e2ba285851687028d395/7_6D7rWrLxp2hb6OHSV1p.png
 - &qwen-tts
   urls:
     - https://github.com/QwenLM/Qwen3-TTS
@@ -3437,6 +3466,77 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-vibevoice"
   mirrors:
     - localai/localai-backends:master-metal-darwin-arm64-vibevoice
+## liquid-audio
+- !!merge <<: *liquid-audio
+  name: "liquid-audio-development"
+  capabilities:
+    nvidia: "cuda12-liquid-audio-development"
+    intel: "intel-liquid-audio-development"
+    amd: "rocm-liquid-audio-development"
+    default: "cpu-liquid-audio-development"
+    nvidia-cuda-13: "cuda13-liquid-audio-development"
+    nvidia-cuda-12: "cuda12-liquid-audio-development"
+    nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-liquid-audio-development"
+- !!merge <<: *liquid-audio
+  name: "cpu-liquid-audio"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-liquid-audio"
+  mirrors:
+    - localai/localai-backends:latest-cpu-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "cpu-liquid-audio-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-cpu-liquid-audio"
+  mirrors:
+    - localai/localai-backends:master-cpu-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "cuda12-liquid-audio"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-liquid-audio"
+  mirrors:
+    - localai/localai-backends:latest-gpu-nvidia-cuda-12-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "cuda12-liquid-audio-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-liquid-audio"
+  mirrors:
+    - localai/localai-backends:master-gpu-nvidia-cuda-12-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "cuda13-liquid-audio"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-13-liquid-audio"
+  mirrors:
+    - localai/localai-backends:latest-gpu-nvidia-cuda-13-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "cuda13-liquid-audio-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-13-liquid-audio"
+  mirrors:
+    - localai/localai-backends:master-gpu-nvidia-cuda-13-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "intel-liquid-audio"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-liquid-audio"
+  mirrors:
+    - localai/localai-backends:latest-gpu-intel-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "intel-liquid-audio-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-liquid-audio"
+  mirrors:
+    - localai/localai-backends:master-gpu-intel-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "rocm-liquid-audio"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-liquid-audio"
+  mirrors:
+    - localai/localai-backends:latest-gpu-rocm-hipblas-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "rocm-liquid-audio-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-liquid-audio"
+  mirrors:
+    - localai/localai-backends:master-gpu-rocm-hipblas-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "cuda13-nvidia-l4t-arm64-liquid-audio"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-cuda-13-arm64-liquid-audio"
+  mirrors:
+    - localai/localai-backends:latest-nvidia-l4t-cuda-13-arm64-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "cuda13-nvidia-l4t-arm64-liquid-audio-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-cuda-13-arm64-liquid-audio"
+  mirrors:
+    - localai/localai-backends:master-nvidia-l4t-cuda-13-arm64-liquid-audio
 ## qwen-tts
 - !!merge <<: *qwen-tts
   name: "qwen-tts-development"

backend/python/liquid-audio/Makefile

@@ -0,0 +1,23 @@
+.PHONY: liquid-audio
+liquid-audio:
+	bash install.sh
+
+.PHONY: run
+run: liquid-audio
+	@echo "Running liquid-audio..."
+	bash run.sh
+	@echo "liquid-audio run."
+
+.PHONY: test
+test: liquid-audio
+	@echo "Testing liquid-audio..."
+	bash test.sh
+	@echo "liquid-audio tested."
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: clean
+clean: protogen-clean
+	rm -rf venv __pycache__

backend/python/liquid-audio/backend.py

@@ -0,0 +1,871 @@
+#!/usr/bin/env python3
+"""
+Liquid Audio backend for LocalAI.
+
+Wraps LiquidAI's `liquid-audio` Python package (https://github.com/Liquid4All/liquid-audio).
+The same model serves four roles, selected by the `mode` option at load time:
+chat, asr, tts, s2s. Fine-tuning is exposed via StartFineTune.
+"""
+from concurrent import futures
+import argparse
+import json
+import os
+import queue
+import signal
+import sys
+import threading
+import time
+import traceback
+import uuid
+
+import grpc
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), '..', 'common'))
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), 'common'))
+from grpc_auth import get_auth_interceptors  # noqa: E402
+from python_utils import parse_options  # noqa: E402
+
+import backend_pb2  # noqa: E402
+import backend_pb2_grpc  # noqa: E402
+
+_ONE_DAY_IN_SECONDS = 60 * 60 * 24
+MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
+
+# Voice id → system-prompt suffix. The model only ships these four voices.
+VOICE_PROMPTS = {
+    "us_male":   "Perform TTS. Use the US male voice.",
+    "us_female": "Perform TTS. Use the US female voice.",
+    "uk_male":   "Perform TTS. Use the UK male voice.",
+    "uk_female": "Perform TTS. Use the UK female voice.",
+}
+DEFAULT_VOICE = "us_female"
+
+# Special-token IDs that LFM2-Audio emits to delimit modality boundaries.
+# Sourced from liquid_audio/model/lfm2_audio.py (see generate_sequential/_sample_*).
+TEXT_END_TOKEN = 130        # <|text_end|>
+AUDIO_START_TOKEN = 128     # <|audio_start|>
+IM_END_TOKEN = 7            # <|im_end|>
+AUDIO_EOS_CODE = 2048       # signals end-of-audio in any codebook position
+
+_PATCHED_LOCAL_PATHS = False
+
+
+def _patch_liquid_audio_local_paths():
+    """Make liquid_audio.utils.get_model_dir() tolerate local directories.
+
+    Upstream always passes its argument to huggingface_hub.snapshot_download,
+    which only accepts `owner/repo` ids. LocalAI's gallery hands us absolute
+    paths under <ModelPath>/<owner>/<repo>, so we intercept snapshot_download
+    in the liquid_audio.utils namespace and return the directory as-is when
+    it already exists on disk. Idempotent.
+    """
+    global _PATCHED_LOCAL_PATHS
+    if _PATCHED_LOCAL_PATHS:
+        return
+    import liquid_audio.utils as _la_utils
+    _orig_snapshot_download = _la_utils.snapshot_download
+
+    def _local_first_snapshot_download(repo_id, revision=None, **kwargs):
+        if isinstance(repo_id, (str, os.PathLike)) and os.path.isdir(str(repo_id)):
+            return str(repo_id)
+        return _orig_snapshot_download(repo_id, revision=revision, **kwargs)
+
+    _la_utils.snapshot_download = _local_first_snapshot_download
+    _PATCHED_LOCAL_PATHS = True
+
+
+def _select_device():
+    import torch
+    if torch.cuda.is_available():
+        return "cuda"
+    if hasattr(torch.backends, "mps") and torch.backends.mps.is_available():
+        return "mps"
+    return "cpu"
+
+
+class ActiveJob:
+    """Tracks an in-flight fine-tune so FineTuneProgress can stream from its queue."""
+
+    def __init__(self, job_id):
+        self.job_id = job_id
+        self.progress_queue = queue.Queue()
+        self.thread = None
+        self.stopped = False
+        self.completed = False
+        self.error = None
+
+
+class BackendServicer(backend_pb2_grpc.BackendServicer):
+    def __init__(self):
+        self.processor = None
+        self.model = None
+        self.device = "cpu"
+        self.dtype = None
+        self.options = {}
+        self.model_id = None
+        self.active_job = None
+
+    @property
+    def mode(self):
+        return str(self.options.get("mode", "chat")).lower()
+
+    @property
+    def voice(self):
+        v = str(self.options.get("voice", DEFAULT_VOICE)).lower()
+        return v if v in VOICE_PROMPTS else DEFAULT_VOICE
+
+
+    def Free(self, request, context):
+        # Called by LocalAI when unloading the model. Drop GPU tensors so the
+        # next load starts from a clean state instead of bumping into OOM.
+        try:
+            for attr in ("model", "processor", "tokenizer"):
+                if hasattr(self, attr):
+                    try:
+                        delattr(self, attr)
+                    except Exception:
+                        pass
+            import gc
+            gc.collect()
+            try:
+                import torch
+                if torch.cuda.is_available():
+                    torch.cuda.empty_cache()
+            except Exception:
+                pass
+            return backend_pb2.Result(success=True, message="OK")
+        except Exception as exc:
+            print(f"Free failed: {exc}", file=sys.stderr)
+            return backend_pb2.Result(success=False, message=str(exc))
+
+
+    def Health(self, request, context):
+        return backend_pb2.Reply(message=bytes("OK", 'utf-8'))
+
+
+    def LoadModel(self, request, context):
+        try:
+            import torch
+
+            self.options = parse_options(request.Options)
+            if self.options.get("voice") and self.options["voice"] not in VOICE_PROMPTS:
+                print(f"Warning: unknown voice '{self.options['voice']}'; defaulting to '{DEFAULT_VOICE}'",
+                      file=sys.stderr)
+
+            requested_device = self.options.get("device")
+            self.device = requested_device or _select_device()
+            if self.device == "cuda" and not torch.cuda.is_available():
+                return backend_pb2.Result(success=False, message="CUDA requested but not available")
+            if self.device == "mps" and not (hasattr(torch.backends, "mps") and
+                                             torch.backends.mps.is_available()):
+                print("MPS not available; falling back to CPU", file=sys.stderr)
+                self.device = "cpu"
+
+            dtype_name = str(self.options.get("dtype", "bfloat16")).lower()
+            self.dtype = {
+                "bfloat16": torch.bfloat16,
+                "bf16":     torch.bfloat16,
+                "float16":  torch.float16,
+                "fp16":     torch.float16,
+                "half":     torch.float16,
+                "float32":  torch.float32,
+                "fp32":     torch.float32,
+            }.get(dtype_name, torch.bfloat16)
+
+            # request.Model holds the raw `parameters.model` value (an HF
+            # repo id like "LiquidAI/LFM2.5-Audio-1.5B"); request.ModelFile
+            # is LocalAI's ModelPath-prefixed local copy that exists only
+            # when the gallery supplied a `files:` list. Mirror the
+            # transformers/vibevoice convention: prefer the repo id and
+            # only switch to the local path if it's been staged on disk.
+            model_id = request.Model
+            if not model_id:
+                model_id = request.ModelFile
+            if not model_id:
+                return backend_pb2.Result(success=False, message="No model identifier provided")
+            if request.ModelFile and os.path.isdir(request.ModelFile):
+                model_id = request.ModelFile
+            self.model_id = model_id
+
+            # Pure fine-tune jobs don't need an in-memory inference model — the
+            # Trainer instantiates its own copy at StartFineTune time.
+            if self.mode == "finetune":
+                print(f"Loaded liquid-audio backend in fine-tune mode (model id: {model_id})",
+                      file=sys.stderr)
+                return backend_pb2.Result(success=True, message="OK")
+
+            from liquid_audio import LFM2AudioModel, LFM2AudioProcessor
+
+            # liquid_audio's from_pretrained unconditionally routes through
+            # huggingface_hub.snapshot_download, which rejects local paths
+            # (HFValidationError on `/models/LiquidAI/LFM2.5-Audio-1.5B`).
+            # When LocalAI's gallery has already staged the weights on disk,
+            # short-circuit the download to return the local directory.
+            _patch_liquid_audio_local_paths()
+
+            print(f"Loading liquid-audio model '{model_id}' on {self.device} ({self.dtype})",
+                  file=sys.stderr)
+            self.processor = LFM2AudioProcessor.from_pretrained(model_id, device=self.device).eval()
+            self.model = LFM2AudioModel.from_pretrained(
+                model_id, device=self.device, dtype=self.dtype
+            ).eval()
+
+            print(f"Liquid-audio mode={self.mode}, voice={self.voice}", file=sys.stderr)
+            return backend_pb2.Result(success=True, message="OK")
+
+        except Exception as exc:
+            print(f"LoadModel failed: {exc}", file=sys.stderr)
+            print(traceback.format_exc(), file=sys.stderr)
+            return backend_pb2.Result(success=False, message=str(exc))
+
+
+    def Predict(self, request, context):
+        try:
+            text = "".join(self._generate_text_stream(request))
+            return backend_pb2.Reply(message=text.encode("utf-8"))
+        except Exception as exc:
+            print(f"Predict failed: {exc}", file=sys.stderr)
+            print(traceback.format_exc(), file=sys.stderr)
+            context.set_code(grpc.StatusCode.INTERNAL)
+            context.set_details(str(exc))
+            return backend_pb2.Reply()
+
+    def PredictStream(self, request, context):
+        try:
+            for delta in self._generate_text_stream(request):
+                yield backend_pb2.Reply(message=delta.encode("utf-8"))
+        except Exception as exc:
+            print(f"PredictStream failed: {exc}", file=sys.stderr)
+            print(traceback.format_exc(), file=sys.stderr)
+            context.set_code(grpc.StatusCode.INTERNAL)
+            context.set_details(str(exc))
+
+
+    def VAD(self, request, context):
+        # Stub voice-activity detector: RMS-energy threshold over 30ms frames at
+        # 16 kHz. Good enough for the realtime endpoint's handleVAD loop, which
+        # only inspects segment presence + last segment end. The proper signal
+        # would come from the model's audio encoder, but that ride-along is a
+        # PR-D scope item — until then this keeps the legacy pipeline path
+        # working without forcing the operator to install a separate VAD model.
+        import numpy as np
+        try:
+            audio = np.asarray(request.audio, dtype=np.float32)
+            if audio.size == 0:
+                return backend_pb2.VADResponse(segments=[])
+
+            sample_rate = 16000
+            frame_size = sample_rate * 30 // 1000  # 30ms → 480 samples
+            threshold = float(self.options.get("vad_rms_threshold", 0.01))
+            min_speech_frames = int(self.options.get("vad_min_speech_frames", 2))  # ≥60ms
+            # handleVAD ticks every 300 ms and only inspects segment presence
+            # + last segment end relative to silence_threshold (~500 ms). Cap
+            # the analysed window to the tail of the buffer so we don't redo
+            # the entire growing utterance every tick.
+            window_s = float(self.options.get("vad_window_s", 5.0))
+            window_samples = int(window_s * sample_rate)
+            time_offset_s = 0.0
+            if audio.size > window_samples:
+                time_offset_s = (audio.size - window_samples) / sample_rate
+                audio = audio[-window_samples:]
+
+            n_frames = audio.size // frame_size
+            if n_frames == 0:
+                return backend_pb2.VADResponse(segments=[])
+            frames = audio[: n_frames * frame_size].reshape(n_frames, frame_size)
+            rms = np.sqrt(np.mean(frames ** 2, axis=1))
+            speech = rms > threshold
+
+            def _emit(start_idx, end_idx, out):
+                if end_idx - start_idx >= min_speech_frames:
+                    out.append(backend_pb2.VADSegment(
+                        start=time_offset_s + start_idx * frame_size / sample_rate,
+                        end=time_offset_s + end_idx * frame_size / sample_rate,
+                    ))
+
+            segments = []
+            start_idx = None
+            for i, is_speech in enumerate(speech):
+                if is_speech and start_idx is None:
+                    start_idx = i
+                elif not is_speech and start_idx is not None:
+                    _emit(start_idx, i, segments)
+                    start_idx = None
+            if start_idx is not None:
+                _emit(start_idx, n_frames, segments)
+            return backend_pb2.VADResponse(segments=segments)
+        except Exception as exc:
+            print(f"VAD failed: {exc}", file=sys.stderr)
+            print(traceback.format_exc(), file=sys.stderr)
+            context.set_code(grpc.StatusCode.INTERNAL)
+            context.set_details(str(exc))
+            return backend_pb2.VADResponse(segments=[])
+
+
+    def TTS(self, request, context):
+        try:
+            if self.model is None or self.processor is None:
+                return backend_pb2.Result(success=False, message="Model not loaded")
+
+            import torch
+            import torchaudio
+            from liquid_audio import ChatState
+
+            voice = request.voice.lower() if request.voice else self.voice
+            voice = voice.removeprefix("lfm2:").removeprefix("lfm:")
+            if voice not in VOICE_PROMPTS:
+                voice = self.voice
+            system_prompt = VOICE_PROMPTS[voice]
+
+            chat = ChatState(self.processor)
+            chat.new_turn("system")
+            chat.add_text(system_prompt)
+            chat.end_turn()
+            chat.new_turn("user")
+            chat.add_text(request.text or "")
+            chat.end_turn()
+            chat.new_turn("assistant")
+
+            audio_top_k = int(self.options.get("audio_top_k", 64))
+            audio_temp = float(self.options.get("audio_temperature", 0.8))
+            max_new = int(self.options.get("max_new_tokens", 2048))
+
+            audio_out = []
+            for tok in self.model.generate_sequential(
+                **chat,
+                max_new_tokens=max_new,
+                audio_temperature=audio_temp,
+                audio_top_k=audio_top_k,
+            ):
+                if tok.numel() > 1:
+                    audio_out.append(tok)
+
+            if len(audio_out) <= 1:
+                return backend_pb2.Result(success=False, message="No audio frames generated")
+
+            # Drop the trailing end-of-audio frame, matching the package's examples.
+            audio_codes = torch.stack(audio_out[:-1], 1).unsqueeze(0)
+            waveform = self.processor.decode(audio_codes)
+
+            out_path = request.dst
+            if not out_path:
+                return backend_pb2.Result(success=False, message="dst path is required")
+            os.makedirs(os.path.dirname(out_path) or ".", exist_ok=True)
+            # soundfile in preference to torchaudio.save — the latter routes
+            # through torchcodec, whose native libs need NVIDIA NPP that we
+            # don't bundle in the cuda13 image.
+            import soundfile as _sf
+            _sf.write(out_path, waveform.cpu().numpy().squeeze(0).T, 24_000)
+
+            return backend_pb2.Result(success=True)
+        except Exception as exc:
+            print(f"TTS failed: {exc}", file=sys.stderr)
+            print(traceback.format_exc(), file=sys.stderr)
+            return backend_pb2.Result(success=False, message=str(exc))
+
+
+    def AudioToAudioStream(self, request_iterator, context):
+        """Bidirectional any-to-any speech-to-speech stream.
+
+        See `backend.proto` AudioToAudioStream for the wire protocol. Audio
+        is decoded once per turn here; chunked detokenization for sub-second
+        TTFB is left to a future iteration once the LFM2AudioDetokenizer
+        gains a streaming entry point.
+        """
+        try:
+            yield from self._audio_to_audio_stream(request_iterator, context)
+        except Exception as exc:
+            print(f"AudioToAudioStream failed: {exc}", file=sys.stderr)
+            print(traceback.format_exc(), file=sys.stderr)
+            yield backend_pb2.AudioToAudioResponse(
+                event="error",
+                meta=json.dumps({"message": str(exc)}).encode("utf-8"),
+            )
+
+    def _audio_to_audio_stream(self, request_iterator, context):
+        if self.model is None or self.processor is None:
+            raise RuntimeError("Model not loaded")
+
+        import torch
+        import torchaudio
+        from liquid_audio import ChatState
+
+        cfg = None
+        chat = None
+        input_sample_rate = 16000
+        output_sample_rate = 24000
+        sequence = 0
+
+        def _new_event(event, **kwargs):
+            nonlocal sequence
+            sequence += 1
+            kwargs.setdefault("sequence", sequence)
+            return backend_pb2.AudioToAudioResponse(event=event, **kwargs)
+
+        def _ensure_chat():
+            """Build a fresh ChatState seeded with the system prompt."""
+            nonlocal chat
+            chat = ChatState(self.processor)
+            system_prompt = (cfg.system_prompt if cfg and cfg.system_prompt
+                             else "Respond with interleaved text and audio.")
+            chat.new_turn("system")
+            chat.add_text(system_prompt)
+            chat.end_turn()
+
+        # Buffers for the in-flight user turn
+        pcm_buffer = bytearray()
+
+        def _consume_user_turn():
+            nonlocal pcm_buffer
+            if not pcm_buffer:
+                return
+            # Avoid the bytes(pcm_buffer) copy and let the float widen happen
+            # in-place: numpy view → torch view → in-place divide.
+            import numpy as np
+            arr = np.frombuffer(memoryview(pcm_buffer), dtype=np.int16)
+            wav = torch.from_numpy(arr).to(torch.float32).div_(32768.0).unsqueeze(0)
+            chat.new_turn("user")
+            chat.add_audio(wav, input_sample_rate)
+            chat.end_turn()
+            pcm_buffer = bytearray()
+
+        def _run_generation():
+            """Run generate_interleaved; yield response events as we go."""
+            chat.new_turn("assistant")
+            audio_top_k = int(self.options.get("audio_top_k", 4))
+            audio_temp = float(self.options.get("audio_temperature", 1.0))
+            text_top_k = int(self.options.get("text_top_k", 0)) or None
+            text_temp = float(self.options.get("text_temperature", 0)) or None
+            max_new = int(self.options.get("max_new_tokens", 512))
+
+            audio_tokens = []
+            for tok in self.model.generate_interleaved(
+                **chat,
+                max_new_tokens=max_new,
+                text_temperature=text_temp,
+                text_top_k=text_top_k,
+                audio_temperature=audio_temp,
+                audio_top_k=audio_top_k,
+            ):
+                if tok.numel() == 1:
+                    if tok.item() == IM_END_TOKEN:
+                        break
+                    text = self.processor.text.decode(tok)
+                    if not text:
+                        continue
+                    yield _new_event(
+                        "response.audio_transcript.delta",
+                        meta=json.dumps({"delta": text}).encode("utf-8"),
+                    )
+                else:
+                    audio_tokens.append(tok)
+
+            # Detokenize the accumulated audio at end-of-turn — the
+            # LFM2AudioDetokenizer is non-streaming today.
+            if len(audio_tokens) > 1:
+                audio_codes = torch.stack(audio_tokens[:-1], 1).unsqueeze(0)
+                waveform = self.processor.decode(audio_codes)
+                # Convert to s16le PCM bytes at output_sample_rate
+                if output_sample_rate != 24000:
+                    waveform = torchaudio.functional.resample(
+                        waveform.cpu(), 24000, output_sample_rate
+                    )
+                pcm = (waveform.cpu().squeeze(0).clamp(-1, 1) * 32767.0).to(
+                    torch.int16
+                ).numpy().tobytes()
+                yield _new_event(
+                    "response.audio.delta",
+                    pcm=pcm,
+                    sample_rate=output_sample_rate,
+                )
+
+            yield _new_event("response.done", meta=b"{}")
+
+        for req in request_iterator:
+            if not context.is_active():
+                return
+            payload = req.WhichOneof("payload")
+            if payload == "config":
+                cfg = req.config
+                if cfg.input_sample_rate > 0:
+                    input_sample_rate = cfg.input_sample_rate
+                if cfg.output_sample_rate > 0:
+                    output_sample_rate = cfg.output_sample_rate
+                # The first config implicitly resets state.
+                _ensure_chat()
+                pcm_buffer = bytearray()
+            elif payload == "frame":
+                if chat is None:
+                    _ensure_chat()
+                if req.frame.pcm:
+                    pcm_buffer.extend(req.frame.pcm)
+                if req.frame.end_of_input:
+                    _consume_user_turn()
+                    yield from _run_generation()
+            elif payload == "control":
+                event = req.control.event
+                if event == "input_audio_buffer.commit":
+                    _consume_user_turn()
+                    yield from _run_generation()
+                elif event == "response.cancel":
+                    # Synchronous generation here means cancel can only
+                    # take effect between turns; we ack so the client unblocks.
+                    yield _new_event("response.done", meta=b'{"cancelled":true}')
+                elif event == "session.update":
+                    # Free-form session re-config; treat as a soft reset.
+                    _ensure_chat()
+                    pcm_buffer = bytearray()
+                # Unknown events are ignored — forward-compatible.
+
+
+    def AudioTranscription(self, request, context):
+        try:
+            if self.model is None or self.processor is None:
+                return backend_pb2.TranscriptResult(segments=[], text="")
+
+            import torchaudio
+            from liquid_audio import ChatState
+
+            audio_path = request.dst
+            if not audio_path:
+                return backend_pb2.TranscriptResult(segments=[], text="")
+
+            chat = ChatState(self.processor)
+            chat.new_turn("system")
+            chat.add_text("Perform ASR.")
+            chat.end_turn()
+            chat.new_turn("user")
+            # soundfile in preference to torchaudio.load — the latter routes
+            # through torchcodec which needs NVIDIA NPP libs we don't bundle.
+            import soundfile as _sf
+            import torch
+            audio_np, sr = _sf.read(audio_path, dtype="float32", always_2d=True)
+            wav = torch.from_numpy(audio_np.T)  # (channels, samples)
+            if wav.shape[0] > 1:
+                # Down-mix to mono — the processor expects a single channel
+                wav = wav.mean(dim=0, keepdim=True)
+            chat.add_audio(wav, sr)
+            chat.end_turn()
+            chat.new_turn("assistant")
+
+            max_new = int(self.options.get("max_new_tokens", 1024))
+
+            pieces = []
+            for tok in self.model.generate_sequential(**chat, max_new_tokens=max_new):
+                if tok.numel() == 1:
+                    if tok.item() == IM_END_TOKEN:
+                        break
+                    pieces.append(self.processor.text.decode(tok))
+
+            text = "".join(pieces).strip()
+            duration_ms = int((wav.shape[1] / sr) * 1000)
+            segment = backend_pb2.TranscriptSegment(
+                id=0, start=0, end=duration_ms, text=text, tokens=[],
+            )
+            return backend_pb2.TranscriptResult(segments=[segment], text=text)
+        except Exception as exc:
+            print(f"AudioTranscription failed: {exc}", file=sys.stderr)
+            print(traceback.format_exc(), file=sys.stderr)
+            return backend_pb2.TranscriptResult(segments=[], text="")
+
+
+    def StartFineTune(self, request, context):
+        if self.active_job is not None and not self.active_job.completed:
+            return backend_pb2.FineTuneJobResult(
+                job_id="", success=False,
+                message="A fine-tuning job is already running",
+            )
+
+        job_id = request.job_id or str(uuid.uuid4())
+        job = ActiveJob(job_id)
+        self.active_job = job
+
+        thread = threading.Thread(target=self._run_training, args=(request, job), daemon=True)
+        job.thread = thread
+        thread.start()
+
+        return backend_pb2.FineTuneJobResult(
+            job_id=job_id, success=True, message="Training started",
+        )
+
+    def FineTuneProgress(self, request, context):
+        if self.active_job is None or self.active_job.job_id != request.job_id:
+            context.set_code(grpc.StatusCode.NOT_FOUND)
+            context.set_details(f"Job {request.job_id} not found")
+            return
+
+        job = self.active_job
+        while True:
+            try:
+                update = job.progress_queue.get(timeout=1.0)
+            except queue.Empty:
+                if job.completed or job.stopped:
+                    break
+                if not context.is_active():
+                    break
+                continue
+            if update is None:
+                break
+            yield update
+            if update.status in ("completed", "failed", "stopped"):
+                break
+
+    def StopFineTune(self, request, context):
+        # We can't kill the Accelerate training loop mid-step cleanly from here;
+        # LocalAI's job manager kills the backend process on stop. The flag below
+        # at least lets the progress stream terminate quickly.
+        if self.active_job is not None and self.active_job.job_id == request.job_id:
+            self.active_job.stopped = True
+            self.active_job.progress_queue.put(None)
+        return backend_pb2.Result(success=True, message="OK")
+
+    def _run_training(self, request, job):
+        try:
+            self._do_train(request, job)
+            job.completed = True
+            job.progress_queue.put(backend_pb2.FineTuneProgressUpdate(
+                job_id=job.job_id, status="completed", message="Training completed",
+                progress_percent=100.0,
+            ))
+        except Exception as exc:
+            job.error = str(exc)
+            job.completed = True
+            print(f"Training failed: {exc}", file=sys.stderr)
+            print(traceback.format_exc(), file=sys.stderr)
+            job.progress_queue.put(backend_pb2.FineTuneProgressUpdate(
+                job_id=job.job_id, status="failed", message=str(exc),
+            ))
+        finally:
+            job.progress_queue.put(None)
+
+    def _do_train(self, request, job):
+        from liquid_audio import LFM2AudioModel  # noqa: F401  (sanity import)
+        from liquid_audio.data.dataloader import LFM2DataLoader
+        from liquid_audio.trainer import Trainer
+
+        model_id = request.model or self.model_id or "LiquidAI/LFM2.5-Audio-1.5B"
+
+        dataset_path = request.dataset_source
+        if not dataset_path:
+            raise ValueError("dataset_source is required (path to a preprocessed dataset)")
+
+        extras = dict(request.extra_options) if request.extra_options else {}
+        val_path = extras.get("val_dataset")
+
+        # Map FineTuneRequest hyperparameters to liquid_audio.Trainer constructor args
+        lr = request.learning_rate or 3e-5
+        max_steps = request.max_steps or 1000
+        warmup_steps = request.warmup_steps or min(100, max_steps // 10)
+        batch_size = request.batch_size or 16
+        save_interval = request.save_steps or max(1, max_steps // 4)
+
+        output_dir = request.output_dir or os.path.join(
+            os.environ.get("LIQUID_AUDIO_OUTPUT_DIR", "/tmp"),
+            f"liquid-audio-{job.job_id}",
+        )
+        os.makedirs(output_dir, exist_ok=True)
+
+        job.progress_queue.put(backend_pb2.FineTuneProgressUpdate(
+            job_id=job.job_id, status="loading_dataset",
+            message=f"Loading preprocessed dataset from {dataset_path}",
+        ))
+        train_data = LFM2DataLoader(dataset_path)
+        val_data = LFM2DataLoader(val_path) if val_path else None
+
+        job.progress_queue.put(backend_pb2.FineTuneProgressUpdate(
+            job_id=job.job_id, status="loading_model",
+            message=f"Loading base model {model_id}",
+        ))
+
+        # The Liquid Trainer logs via self.accelerator.print; we subclass it to
+        # also push progress events onto the queue every logging_interval steps.
+        progress_q = job.progress_queue
+
+        class QueuedTrainer(Trainer):
+            def log(self_, model_output):
+                if self_.step > 0 and self_.step % self_.logging_interval == 0:
+                    try:
+                        loss = self_.accelerator.reduce(
+                            model_output.loss.detach(), reduction="mean"
+                        ).item()
+                    except Exception:
+                        loss = float("nan")
+                    lr_now = self_.optimizer.param_groups[0]["lr"]
+                    pct = (self_.step / self_.max_steps * 100.0) if self_.max_steps else 0.0
+                    progress_q.put(backend_pb2.FineTuneProgressUpdate(
+                        job_id=job.job_id,
+                        current_step=int(self_.step),
+                        total_steps=int(self_.max_steps),
+                        current_epoch=float(self_.epoch),
+                        loss=float(loss),
+                        learning_rate=float(lr_now),
+                        progress_percent=float(pct),
+                        status="training",
+                    ))
+                # Honour stop requests: raising here terminates the loop cleanly
+                if job.stopped:
+                    raise KeyboardInterrupt("stop requested")
+                return super().log(model_output)
+
+            def validate(self_):
+                progress_q.put(backend_pb2.FineTuneProgressUpdate(
+                    job_id=job.job_id, current_step=int(self_.step),
+                    total_steps=int(self_.max_steps), status="training",
+                    message=f"Running validation at step {self_.step}",
+                ))
+                return super().validate()
+
+        trainer = QueuedTrainer(
+            model_id=model_id,
+            train_data=train_data,
+            val_data=val_data,
+            lr=lr,
+            max_steps=max_steps,
+            warmup_steps=warmup_steps,
+            batch_size=batch_size,
+            save_interval=save_interval,
+            output_dir=output_dir,
+            weight_decay=request.weight_decay or 0.1,
+        )
+
+        job.progress_queue.put(backend_pb2.FineTuneProgressUpdate(
+            job_id=job.job_id, status="training", message="Training started",
+            total_steps=int(max_steps),
+        ))
+        trainer.train()
+
+        job.progress_queue.put(backend_pb2.FineTuneProgressUpdate(
+            job_id=job.job_id, status="saving",
+            message=f"Saved final model to {output_dir}",
+            checkpoint_path=os.path.join(output_dir, "final"),
+        ))
+
+
+    def _build_chat_state(self, messages, user_prompt, tools_prelude=None):
+        """Build a ChatState from a list of (role, content) tuples plus an optional final user turn.
+
+        tools_prelude, when non-empty, is prepended as an extra system turn carrying
+        the LFM2 tool-list block — mirrors gallery/lfm.yaml's `function:` template
+        so the model sees the same prompt shape whether served via llama-cpp or here.
+        """
+        from liquid_audio import ChatState
+        chat = ChatState(self.processor)
+        if tools_prelude:
+            chat.new_turn("system")
+            chat.add_text(tools_prelude)
+            chat.end_turn()
+        for role, content in messages:
+            chat.new_turn(role)
+            chat.add_text(content)
+            chat.end_turn()
+        if user_prompt:
+            chat.new_turn("user")
+            chat.add_text(user_prompt)
+            chat.end_turn()
+        chat.new_turn("assistant")
+        return chat
+
+    def _collect_messages(self, request):
+        """Translate PredictOptions.Messages into (role, content) tuples."""
+        out = []
+        for m in request.Messages:
+            role = (m.role or "user").lower()
+            if role not in ("system", "user", "assistant"):
+                role = "user"
+            out.append((role, m.content or ""))
+        return out
+
+    def _render_tools_prelude(self, request):
+        """Build the LFM2 `<|tool_list_start|>…<|tool_list_end|>` system prelude
+        from request.Tools (OpenAI Chat-Completions tool JSON). Returns "" when
+        no tools are attached. Output mirrors gallery/lfm.yaml's `function:`
+        template so the model sees the same prompt whether routed via llama-cpp
+        or this backend."""
+        tools_raw = getattr(request, "Tools", "") or ""
+        if not tools_raw:
+            return ""
+        try:
+            tools = json.loads(tools_raw)
+        except json.JSONDecodeError:
+            print(f"liquid-audio: ignoring malformed Tools JSON: {tools_raw[:200]!r}",
+                  file=sys.stderr)
+            return ""
+        if not isinstance(tools, list) or not tools:
+            return ""
+        # The LFM2 chat template uses single-quoted Python-dict-ish syntax in
+        # examples, but the tokenizer treats this whole block as opaque text;
+        # JSON works fine and is what other backends emit.
+        return (
+            "You are a function calling AI model. You are provided with functions to "
+            "execute. You may call one or more functions to assist with the user query. "
+            "Don't make assumptions about what values to plug into functions.\n"
+            "List of tools: <|tool_list_start|>"
+            + json.dumps(tools, separators=(",", ":"))
+            + "<|tool_list_end|>"
+        )
+
+    def _generate_text_stream(self, request):
+        """Yield text-only deltas from generate_sequential. Caller joins for unary Predict."""
+        if self.model is None or self.processor is None:
+            raise RuntimeError("Model not loaded")
+        messages = self._collect_messages(request)
+        user_prompt = request.Prompt or None
+        tools_prelude = self._render_tools_prelude(request)
+        # If the request already carries Messages, Prompt is the templated form
+        # of the same content — don't append a duplicate user turn.
+        chat = self._build_chat_state(
+            messages,
+            user_prompt if not messages else None,
+            tools_prelude=tools_prelude,
+        )
+
+        max_new = request.Tokens if request.Tokens > 0 else int(self.options.get("max_new_tokens", 512))
+        temperature = request.Temperature if request.Temperature > 0 else None
+        top_k = request.TopK if request.TopK > 0 else None
+
+        for tok in self.model.generate_sequential(
+            **chat,
+            max_new_tokens=max_new,
+            text_temperature=temperature,
+            text_top_k=top_k,
+        ):
+            if tok.numel() == 1:
+                if tok.item() == IM_END_TOKEN:
+                    break
+                yield self.processor.text.decode(tok)
+
+
+def serve(address):
+    server = grpc.server(
+        futures.ThreadPoolExecutor(max_workers=MAX_WORKERS),
+        options=[
+            ('grpc.max_message_length', 50 * 1024 * 1024),
+            ('grpc.max_send_message_length', 50 * 1024 * 1024),
+            ('grpc.max_receive_message_length', 50 * 1024 * 1024),
+        ],
+        interceptors=get_auth_interceptors(),
+    )
+    backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
+    server.add_insecure_port(address)
+    server.start()
+    print(f"Liquid-audio backend listening on {address}", file=sys.stderr, flush=True)
+
+    def stop(_signum, _frame):
+        server.stop(0)
+        sys.exit(0)
+
+    signal.signal(signal.SIGTERM, stop)
+    signal.signal(signal.SIGINT, stop)
+
+    try:
+        while True:
+            time.sleep(_ONE_DAY_IN_SECONDS)
+    except KeyboardInterrupt:
+        server.stop(0)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Liquid Audio gRPC backend")
+    parser.add_argument("--addr", default="localhost:50051", help="gRPC server address")
+    args = parser.parse_args()
+    serve(args.addr)

backend/python/liquid-audio/install.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+set -e
+
+# liquid-audio requires Python ≥ 3.12 (per its pyproject.toml); the default
+# portable Python in libbackend.sh is 3.10. Override before sourcing.
+export PYTHON_VERSION="${PYTHON_VERSION:-3.12}"
+export PYTHON_PATCH="${PYTHON_PATCH:-11}"
+
+backend_dir=$(dirname $0)
+if [ -d $backend_dir/common ]; then
+    source $backend_dir/common/libbackend.sh
+else
+    source $backend_dir/../common/libbackend.sh
+fi
+
+# liquid-audio's torch wheels are large; allow upgrades to satisfy transitive pins
+EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
+installRequirements

backend/python/liquid-audio/protogen.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+set -e
+
+backend_dir=$(dirname $0)
+if [ -d $backend_dir/common ]; then
+    source $backend_dir/common/libbackend.sh
+else
+    source $backend_dir/../common/libbackend.sh
+fi
+
+runProtogen

backend/python/liquid-audio/requirements-cpu.txt

@@ -0,0 +1,13 @@
+--extra-index-url https://download.pytorch.org/whl/cpu
+torch>=2.8.0
+torchaudio>=2.8.0
+torchcodec>=0.9.1
+transformers>=4.55.4
+accelerate>=1.10.1
+datasets>=4.8.4
+einops>=0.8.1
+librosa>=0.11.0
+soundfile>=0.12.1
+sentencepiece>=0.2.1
+huggingface-hub>=1.3.0
+liquid-audio>=1.2.0

backend/python/liquid-audio/requirements-cublas12.txt

@@ -0,0 +1,13 @@
+--extra-index-url https://download.pytorch.org/whl/cu121
+torch>=2.8.0
+torchaudio>=2.8.0
+torchcodec>=0.9.1
+transformers>=4.55.4
+accelerate>=1.10.1
+datasets>=4.8.4
+einops>=0.8.1
+librosa>=0.11.0
+soundfile>=0.12.1
+sentencepiece>=0.2.1
+huggingface-hub>=1.3.0
+liquid-audio>=1.2.0

backend/python/liquid-audio/requirements-cublas13.txt

@@ -0,0 +1,13 @@
+--extra-index-url https://download.pytorch.org/whl/cu130
+torch>=2.8.0
+torchaudio>=2.8.0
+torchcodec>=0.9.1
+transformers>=4.55.4
+accelerate>=1.10.1
+datasets>=4.8.4
+einops>=0.8.1
+librosa>=0.11.0
+soundfile>=0.12.1
+sentencepiece>=0.2.1
+huggingface-hub>=1.3.0
+liquid-audio>=1.2.0

backend/python/liquid-audio/requirements-hipblas.txt

@@ -0,0 +1,13 @@
+--extra-index-url https://download.pytorch.org/whl/rocm7.0
+torch>=2.8.0
+torchaudio>=2.8.0
+torchcodec>=0.9.1
+transformers>=4.55.4
+accelerate>=1.10.1
+datasets>=4.8.4
+einops>=0.8.1
+librosa>=0.11.0
+soundfile>=0.12.1
+sentencepiece>=0.2.1
+huggingface-hub>=1.3.0
+liquid-audio>=1.2.0

backend/python/liquid-audio/requirements-l4t13.txt

@@ -0,0 +1,13 @@
+--extra-index-url https://pypi.jetson-ai-lab.io/jp7/cu130
+torch>=2.8.0
+torchaudio>=2.8.0
+torchcodec>=0.9.1
+transformers>=4.55.4
+accelerate>=1.10.1
+datasets>=4.8.4
+einops>=0.8.1
+librosa>=0.11.0
+soundfile>=0.12.1
+sentencepiece>=0.2.1
+huggingface-hub>=1.3.0
+liquid-audio>=1.2.0

backend/python/liquid-audio/requirements-mps.txt

@@ -0,0 +1,12 @@
+torch>=2.8.0
+torchaudio>=2.8.0
+torchcodec>=0.9.1
+transformers>=4.55.4
+accelerate>=1.10.1
+datasets>=4.8.4
+einops>=0.8.1
+librosa>=0.11.0
+soundfile>=0.12.1
+sentencepiece>=0.2.1
+huggingface-hub>=1.3.0
+liquid-audio>=1.2.0

backend/python/liquid-audio/requirements.txt

@@ -0,0 +1,3 @@
+grpcio==1.78.1
+protobuf
+certifi

backend/python/liquid-audio/run.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+
+backend_dir=$(dirname $0)
+if [ -d $backend_dir/common ]; then
+    source $backend_dir/common/libbackend.sh
+else
+    source $backend_dir/../common/libbackend.sh
+fi
+
+startBackend $@

backend/python/liquid-audio/test.py

@@ -0,0 +1,89 @@
+"""Smoke tests for the liquid-audio backend.
+
+These run without contacting HuggingFace or loading model weights:
+they only verify that the gRPC service starts and Health() responds.
+
+To run an end-to-end inference test, set LIQUID_AUDIO_MODEL_ID
+(e.g. "LiquidAI/LFM2.5-Audio-1.5B") in the environment — see test_inference().
+"""
+import os
+import subprocess
+import sys
+import time
+import unittest
+
+import grpc
+
+# Ensure generated protobuf stubs are importable
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+
+import backend_pb2
+import backend_pb2_grpc
+
+
+class TestBackend(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls):
+        addr = os.environ.get("LIQUID_AUDIO_TEST_ADDR", "localhost:50053")
+        cls.addr = addr
+        cls.server = subprocess.Popen(
+            [sys.executable, os.path.join(os.path.dirname(__file__), "backend.py"), "--addr", addr],
+        )
+        time.sleep(2)  # Give the server a moment to bind
+
+    @classmethod
+    def tearDownClass(cls):
+        cls.server.terminate()
+        try:
+            cls.server.wait(timeout=5)
+        except subprocess.TimeoutExpired:
+            cls.server.kill()
+
+    def _stub(self):
+        channel = grpc.insecure_channel(self.addr)
+        return backend_pb2_grpc.BackendStub(channel)
+
+    def test_health(self):
+        stub = self._stub()
+        reply = stub.Health(backend_pb2.HealthMessage(), timeout=5)
+        self.assertEqual(reply.message, b"OK")
+
+    def test_load_finetune_mode_without_weights(self):
+        """Loading in fine-tune mode should succeed without pulling model weights."""
+        stub = self._stub()
+        result = stub.LoadModel(
+            backend_pb2.ModelOptions(
+                Model="LiquidAI/LFM2.5-Audio-1.5B",
+                Options=["mode:finetune"],
+            ),
+            timeout=10,
+        )
+        self.assertTrue(result.success, msg=result.message)
+
+    @unittest.skipUnless(os.environ.get("LIQUID_AUDIO_MODEL_ID"),
+                         "Set LIQUID_AUDIO_MODEL_ID to run an end-to-end inference smoke test")
+    def test_inference(self):
+        """End-to-end: load a real LFM2-Audio model and run one short prediction."""
+        stub = self._stub()
+        model_id = os.environ["LIQUID_AUDIO_MODEL_ID"]
+        result = stub.LoadModel(
+            backend_pb2.ModelOptions(
+                Model=model_id,
+                Options=["mode:chat"],
+            ),
+            timeout=600,
+        )
+        self.assertTrue(result.success, msg=result.message)
+        reply = stub.Predict(
+            backend_pb2.PredictOptions(
+                Prompt="Hello!",
+                Tokens=8,
+                Temperature=0.0,
+            ),
+            timeout=120,
+        )
+        self.assertGreater(len(reply.message), 0)
+
+
+if __name__ == "__main__":
+    unittest.main()

backend/python/liquid-audio/test.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+set -e
+
+backend_dir=$(dirname $0)
+if [ -d $backend_dir/common ]; then
+    source $backend_dir/common/libbackend.sh
+else
+    source $backend_dir/../common/libbackend.sh
+fi
+
+runUnittests

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

@@ -3,5 +3,5 @@
 # on a cu130 host. Pull the cu130-flavoured wheel from vLLM's per-tag index
 # instead — the cublas13 case in install.sh adds --index-strategy=unsafe-best-match
 # so uv consults this index alongside PyPI.
---extra-index-url https://wheels.vllm.ai/0.20.2/cu130
-vllm==0.20.2
+--extra-index-url https://wheels.vllm.ai/0.21.0/cu130
+vllm==0.21.0

core/application/distributed.go

@@ -169,7 +169,7 @@ func initDistributed(cfg *config.ApplicationConfig, authDB *gorm.DB, configLoade
 		cfg.Distributed.HealthCheckIntervalOrDefault(),
 		cfg.Distributed.StaleNodeThresholdOrDefault(),
 		routerAuthToken,
-		cfg.Distributed.PerModelHealthCheck,
+		!cfg.Distributed.DisablePerModelHealthCheck,
 	)
 
 	// Initialize job store

core/config/backend_capabilities.go

@@ -24,6 +24,7 @@ const (
 	UsecaseVAD             = "vad"
 	UsecaseAudioTransform  = "audio_transform"
 	UsecaseDiarization     = "diarization"
+	UsecaseRealtimeAudio   = "realtime_audio"
 )
 
 // GRPCMethod identifies a Backend service RPC from backend.proto.
@@ -45,6 +46,7 @@ const (
 	MethodVAD                GRPCMethod = "VAD"
 	MethodAudioTransform     GRPCMethod = "AudioTransform"
 	MethodDiarize            GRPCMethod = "Diarize"
+	MethodAudioToAudioStream GRPCMethod = "AudioToAudioStream"
 )
 
 // UsecaseInfo describes a single known_usecase value and how it maps
@@ -147,6 +149,11 @@ var UsecaseInfoMap = map[string]UsecaseInfo{
 		GRPCMethod:  MethodDiarize,
 		Description: "Speaker diarization (who-spoke-when, per-speaker segments) via the Diarize RPC.",
 	},
+	UsecaseRealtimeAudio: {
+		Flag:        FLAG_REALTIME_AUDIO,
+		GRPCMethod:  MethodAudioToAudioStream,
+		Description: "Self-contained any-to-any audio model for the Realtime API — accepts microphone audio and emits speech + transcript (+ optional function calls) from a single backend via the AudioToAudioStream RPC.",
+	},
 }
 
 // BackendCapability describes which gRPC methods and usecases a backend supports.
@@ -397,6 +404,15 @@ var BackendCapabilities = map[string]BackendCapability{
 		Description:      "Meta MusicGen via transformers — music generation from text",
 	},
 
+	// --- Any-to-any audio backends ---
+	"liquid-audio": {
+		GRPCMethods:      []GRPCMethod{MethodPredict, MethodPredictStream, MethodAudioTranscription, MethodTTS, MethodAudioToAudioStream, MethodVAD},
+		PossibleUsecases: []string{UsecaseChat, UsecaseCompletion, UsecaseTranscript, UsecaseTTS, UsecaseRealtimeAudio, UsecaseVAD},
+		DefaultUsecases:  []string{UsecaseRealtimeAudio, UsecaseChat, UsecaseTranscript, UsecaseTTS, UsecaseVAD},
+		AcceptsAudios:    true,
+		Description:      "LFM2 / LFM2.5-Audio — self-contained any-to-any audio model for the Realtime API; also exposes chat, transcription, TTS and a stub energy-based VAD endpoint",
+	},
+
 	// --- Audio transform backends ---
 	"localvqe": {
 		GRPCMethods:      []GRPCMethod{MethodAudioTransform},

core/config/distributed_config.go

@@ -31,7 +31,15 @@ type DistributedConfig struct {
 	DrainTimeout        time.Duration // Time to wait for in-flight requests during drain (default 30s)
 	HealthCheckInterval time.Duration // Health monitor check interval (default 15s)
 	StaleNodeThreshold  time.Duration // Time before a node is considered stale (default 60s)
-	PerModelHealthCheck bool          // Enable per-model backend health checking (default false)
+	// DisablePerModelHealthCheck turns off the health monitor's per-model
+	// gRPC probe. When enabled (the default), the monitor pings each model's
+	// gRPC address and removes stale node_models rows whose backend has
+	// crashed even though the worker's node-level heartbeat is still arriving.
+	// Without per-model probing, /embeddings and /completions can be dispatched
+	// to a backend that silently returns garbage (see also the cascading
+	// model-row cleanup on MarkUnhealthy / MarkDraining).
+	DisablePerModelHealthCheck bool
+
 	MCPCIJobTimeout     time.Duration // MCP CI job execution timeout (default 10m)
 
 	MaxUploadSize int64 // Maximum upload body size in bytes (default 50 GB)

core/config/model_config.go

@@ -636,6 +636,7 @@ const (
 	FLAG_SPEAKER_RECOGNITION ModelConfigUsecase = 0b1000000000000000
 	FLAG_AUDIO_TRANSFORM     ModelConfigUsecase = 0b10000000000000000
 	FLAG_DIARIZATION         ModelConfigUsecase = 0b100000000000000000
+	FLAG_REALTIME_AUDIO      ModelConfigUsecase = 0b1000000000000000000
 
 	// Common Subsets
 	FLAG_LLM ModelConfigUsecase = FLAG_CHAT | FLAG_COMPLETION | FLAG_EDIT
@@ -645,12 +646,12 @@ const (
 // Flags within the same group are NOT orthogonal (e.g., chat and completion are
 // both text/language). A model is multimodal when its usecases span 2+ groups.
 var ModalityGroups = []ModelConfigUsecase{
-	FLAG_CHAT | FLAG_COMPLETION | FLAG_EDIT, // text/language
-	FLAG_VISION | FLAG_DETECTION,            // visual understanding
-	FLAG_TRANSCRIPT,                         // speech input
-	FLAG_TTS | FLAG_SOUND_GENERATION,        // audio output
-	FLAG_AUDIO_TRANSFORM,                    // audio in/out transforms
-	FLAG_IMAGE | FLAG_VIDEO,                 // visual generation
+	FLAG_CHAT | FLAG_COMPLETION | FLAG_EDIT,    // text/language
+	FLAG_VISION | FLAG_DETECTION,               // visual understanding
+	FLAG_TRANSCRIPT | FLAG_REALTIME_AUDIO,      // speech input — realtime_audio is any-to-any, so it counts here too
+	FLAG_TTS | FLAG_SOUND_GENERATION | FLAG_REALTIME_AUDIO, // audio output — and here, so a lone realtime_audio flag still reads as multimodal
+	FLAG_AUDIO_TRANSFORM,                       // audio in/out transforms
+	FLAG_IMAGE | FLAG_VIDEO,                    // visual generation
 }
 
 // IsMultimodal returns true if the given usecases span two or more orthogonal
@@ -692,6 +693,7 @@ func GetAllModelConfigUsecases() map[string]ModelConfigUsecase {
 		"FLAG_SPEAKER_RECOGNITION": FLAG_SPEAKER_RECOGNITION,
 		"FLAG_AUDIO_TRANSFORM":     FLAG_AUDIO_TRANSFORM,
 		"FLAG_DIARIZATION":         FLAG_DIARIZATION,
+		"FLAG_REALTIME_AUDIO":      FLAG_REALTIME_AUDIO,
 	}
 }
 
@@ -866,6 +868,16 @@ func (c *ModelConfig) GuessUsecases(u ModelConfigUsecase) bool {
 		}
 	}
 
+	if (u & FLAG_REALTIME_AUDIO) == FLAG_REALTIME_AUDIO {
+		// Backends that own a single any-to-any loop and implement
+		// AudioToAudioStream — listed here so models without an explicit
+		// known_usecases still surface on the Talk page.
+		realtimeAudioBackends := []string{"liquid-audio"}
+		if !slices.Contains(realtimeAudioBackends, c.Backend) {
+			return false
+		}
+	}
+
 	return true
 }
 

core/gallery/importers/importers.go

@@ -130,6 +130,8 @@ var defaultImporters = []Importer{
 	// and would otherwise swallow the C++ port's GGUF bundles.
 	&VibeVoiceCppImporter{},
 	&VibeVoiceImporter{},
+	// LiquidAudio (Python) — keep before LlamaCPP so non-GGUF LFM2-Audio repos route here.
+	&LiquidAudioImporter{},
 	&CoquiImporter{},
 	// Image/Video (Batch 3)
 	&StableDiffusionGGMLImporter{},

core/gallery/importers/liquid-audio.go

@@ -0,0 +1,145 @@
+package importers
+
+import (
+	"encoding/json"
+	"path/filepath"
+	"strings"
+
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/gallery"
+	"github.com/mudler/LocalAI/core/schema"
+	"go.yaml.in/yaml/v2"
+)
+
+var _ Importer = &LiquidAudioImporter{}
+
+// LiquidAudioImporter recognises LiquidAI's LFM2-Audio family (LFM2-Audio-1.5B,
+// LFM2.5-Audio-1.5B, community finetunes) and routes them to the Python
+// `liquid-audio` backend. Detection is by repo-name substring so third-party
+// mirrors still match. preferences.backend="liquid-audio" overrides detection.
+//
+// Once upstream llama.cpp PR #18641 lands and the GGUF gallery entries are
+// added, GGUF mirrors of these models should route to llama-cpp; that's
+// handled by ordering LlamaCPPImporter after this one and by the explicit
+// "-gguf" exclusion below.
+type LiquidAudioImporter struct{}
+
+func (i *LiquidAudioImporter) Name() string      { return "liquid-audio" }
+func (i *LiquidAudioImporter) Modality() string  { return "tts" }
+func (i *LiquidAudioImporter) AutoDetects() bool { return true }
+
+func (i *LiquidAudioImporter) Match(details Details) bool {
+	preferences, err := details.Preferences.MarshalJSON()
+	if err != nil {
+		return false
+	}
+	preferencesMap := make(map[string]any)
+	if len(preferences) > 0 {
+		if err := json.Unmarshal(preferences, &preferencesMap); err != nil {
+			return false
+		}
+	}
+
+	if b, ok := preferencesMap["backend"].(string); ok && b == "liquid-audio" {
+		return true
+	}
+
+	matchRepo := func(repo string) bool {
+		r := strings.ToLower(repo)
+		// Cede GGUF mirrors to the (later-ordered) llama-cpp importer.
+		if strings.HasSuffix(r, "-gguf") {
+			return false
+		}
+		return strings.Contains(r, "lfm2-audio") || strings.Contains(r, "lfm2.5-audio")
+	}
+
+	if details.HuggingFace != nil {
+		repoName := details.HuggingFace.ModelID
+		if idx := strings.Index(repoName, "/"); idx >= 0 {
+			repoName = repoName[idx+1:]
+		}
+		if matchRepo(repoName) {
+			return true
+		}
+	}
+
+	if _, repo, ok := HFOwnerRepoFromURI(details.URI); ok {
+		return matchRepo(repo)
+	}
+	return false
+}
+
+func (i *LiquidAudioImporter) Import(details Details) (gallery.ModelConfig, error) {
+	preferences, err := details.Preferences.MarshalJSON()
+	if err != nil {
+		return gallery.ModelConfig{}, err
+	}
+	preferencesMap := make(map[string]any)
+	if len(preferences) > 0 {
+		if err := json.Unmarshal(preferences, &preferencesMap); err != nil {
+			return gallery.ModelConfig{}, err
+		}
+	}
+
+	name, ok := preferencesMap["name"].(string)
+	if !ok {
+		name = filepath.Base(details.URI)
+	}
+
+	description, ok := preferencesMap["description"].(string)
+	if !ok {
+		description = "Imported from " + details.URI
+	}
+
+	model := details.URI
+	if details.HuggingFace != nil && details.HuggingFace.ModelID != "" {
+		model = details.HuggingFace.ModelID
+	}
+
+	// Preferences may pin the mode (chat / asr / tts / s2s / finetune).
+	// Default to s2s — the headline any-to-any use case.
+	mode, _ := preferencesMap["mode"].(string)
+	if mode == "" {
+		mode = "s2s"
+	}
+
+	options := []string{"mode:" + mode}
+	if voice, ok := preferencesMap["voice"].(string); ok && voice != "" {
+		options = append(options, "voice:"+voice)
+	}
+
+	usecases := []string{"chat"}
+	switch mode {
+	case "asr":
+		usecases = []string{"transcript"}
+	case "tts":
+		usecases = []string{"tts"}
+	case "s2s":
+		// realtime_audio surfaces the model on the Talk page; chat/tts/
+		// transcript/vad keep the standalone OpenAI-compatible endpoints
+		// working since liquid-audio implements all of them.
+		usecases = []string{"realtime_audio", "chat", "tts", "transcript", "vad"}
+	}
+
+	modelConfig := config.ModelConfig{
+		Name:                name,
+		Description:         description,
+		Backend:             "liquid-audio",
+		KnownUsecaseStrings: usecases,
+		Options:             options,
+		PredictionOptions: schema.PredictionOptions{
+			BasicModelRequest: schema.BasicModelRequest{Model: model},
+		},
+	}
+
+	data, err := yaml.Marshal(modelConfig)
+	if err != nil {
+		return gallery.ModelConfig{}, err
+	}
+
+	return gallery.ModelConfig{
+		Name:        name,
+		Description: description,
+		ConfigFile:  string(data),
+	}, nil
+}

core/gallery/importers/liquid-audio_test.go

@@ -0,0 +1,91 @@
+package importers_test
+
+import (
+	"encoding/json"
+	"fmt"
+
+	"github.com/mudler/LocalAI/core/gallery/importers"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("LiquidAudioImporter", func() {
+	Context("detection from HuggingFace", func() {
+		It("matches LiquidAI/LFM2.5-Audio-1.5B", func() {
+			uri := "https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B"
+			preferences := json.RawMessage(`{}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("backend: liquid-audio"))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("LiquidAI/LFM2.5-Audio-1.5B"))
+		})
+
+		It("matches LiquidAI/LFM2-Audio-1.5B (older variant)", func() {
+			uri := "https://huggingface.co/LiquidAI/LFM2-Audio-1.5B"
+			preferences := json.RawMessage(`{}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("backend: liquid-audio"))
+		})
+
+		It("cedes -GGUF mirrors to the llama-cpp importer", func() {
+			// LiquidAI/LFM2.5-Audio-1.5B-GGUF should NOT route to liquid-audio.
+			// Once upstream PR #18641 lands and the GGUF gallery entry exists,
+			// this is the path that lets users opt into the C++ runtime.
+			uri := "https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B-GGUF"
+			preferences := json.RawMessage(`{}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).ToNot(ContainSubstring("backend: liquid-audio"),
+				fmt.Sprintf("GGUF repo should not match Python importer; got: %s", modelConfig.ConfigFile))
+		})
+	})
+
+	Context("preference override", func() {
+		It("honours preferences.backend=liquid-audio for arbitrary URIs", func() {
+			uri := "https://example.com/some-unrelated-model"
+			preferences := json.RawMessage(`{"backend": "liquid-audio"}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("backend: liquid-audio"))
+		})
+
+		It("picks up the mode preference", func() {
+			uri := "https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B"
+			preferences := json.RawMessage(`{"mode": "asr"}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("mode:asr"))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("transcript"))
+		})
+
+		It("picks up the voice preference", func() {
+			uri := "https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B"
+			preferences := json.RawMessage(`{"mode": "tts", "voice": "uk_male"}`)
+
+			modelConfig, err := importers.DiscoverModelConfig(uri, preferences)
+
+			Expect(err).ToNot(HaveOccurred(), fmt.Sprintf("Error: %v", err))
+			Expect(modelConfig.ConfigFile).To(ContainSubstring("voice:uk_male"))
+		})
+	})
+
+	Context("Importer interface metadata", func() {
+		It("exposes name/modality/autodetect", func() {
+			imp := &importers.LiquidAudioImporter{}
+			Expect(imp.Name()).To(Equal("liquid-audio"))
+			Expect(imp.Modality()).To(Equal("tts"))
+			Expect(imp.AutoDetects()).To(BeTrue())
+		})
+	})
+})

core/http/app.go

@@ -443,6 +443,25 @@ func API(application *application.Application) (*echo.Echo, error) {
 					baseTag := `<base href="` + httpMiddleware.SecureBaseHref(baseURL) + `" />`
 					indexHTML = []byte(strings.Replace(string(indexHTML), "<head>", "<head>\n  "+baseTag, 1))
 				}
+				// <base href> only changes how relative URLs resolve; path-absolute
+				// URLs (those starting with `/`) still resolve against the origin
+				// and would bypass the reverse-proxy prefix. Rewrite the internal
+				// path-absolute references emitted by the build so the browser
+				// requests them through the proxy under the prefix.
+				//
+				// HTML-escape the prefix before interpolating it into attributes:
+				// BasePathPrefix already gates X-Forwarded-Prefix via
+				// SafeForwardedPrefix, but the validator only blocks open-redirect
+				// shapes (// prefix, backslashes, control chars), not attribute
+				// breakout characters like `"`. Escaping makes this resilient
+				// even if the validator ever loosens.
+				if prefix := httpMiddleware.BasePathPrefix(c); prefix != "/" {
+					safePrefix := httpMiddleware.SecureBaseHref(prefix)
+					html := string(indexHTML)
+					html = strings.ReplaceAll(html, `="/assets/`, `="`+safePrefix+`assets/`)
+					html = strings.ReplaceAll(html, `="/favicon.svg"`, `="`+safePrefix+`favicon.svg"`)
+					indexHTML = []byte(html)
+				}
 				return c.HTMLBlob(http.StatusOK, indexHTML)
 			}
 

core/http/app_test.go

@@ -446,6 +446,42 @@ var _ = Describe("API test", func() {
 				Expect(sc).To(Equal(200), "status code")
 				Expect(string(body)).To(ContainSubstring(`<base href="https://example.org/myprefix/" />`), "body")
 			})
+
+			// Caddy's `handle_path` (and similar directives) strip the matched
+			// prefix before forwarding upstream, so LocalAI receives the
+			// already-stripped path together with X-Forwarded-Prefix. The base
+			// href and asset URLs must still include the prefix so the browser
+			// requests them through the proxy.
+			It("Should support reverse-proxy when prefix is stripped by the proxy", func() {
+
+				err, sc, body := getRequest("http://127.0.0.1:9090/app", http.Header{
+					"X-Forwarded-Proto":  {"https"},
+					"X-Forwarded-Host":   {"example.org"},
+					"X-Forwarded-Prefix": {"/myprefix"},
+				})
+				Expect(err).To(BeNil(), "error")
+				Expect(sc).To(Equal(200), "status code")
+				Expect(string(body)).To(ContainSubstring(`<base href="https://example.org/myprefix/" />`), "body")
+				Expect(string(body)).ToNot(ContainSubstring(`="/assets/`), "asset URLs must include the prefix")
+				Expect(string(body)).ToNot(ContainSubstring(`="/favicon.svg"`), "favicon URL must include the prefix")
+			})
+
+			// X-Forwarded-Prefix is attacker controllable on misconfigured
+			// proxy chains. A value like "//evil.com" would otherwise turn the
+			// asset URL rewrite into a protocol-relative URL that loads JS
+			// from a foreign origin. BasePathPrefix must reject these via
+			// SafeForwardedPrefix and fall back to "/".
+			It("Should ignore an unsafe X-Forwarded-Prefix and not poison asset URLs", func() {
+				err, sc, body := getRequest("http://127.0.0.1:9090/app", http.Header{
+					"X-Forwarded-Proto":  {"https"},
+					"X-Forwarded-Host":   {"example.org"},
+					"X-Forwarded-Prefix": {"//evil.com"},
+				})
+				Expect(err).To(BeNil(), "error")
+				Expect(sc).To(Equal(200), "status code")
+				Expect(string(body)).ToNot(ContainSubstring("evil.com"), "unsafe prefix must not leak into the response")
+				Expect(string(body)).ToNot(ContainSubstring(`="//`), "asset URLs must not become protocol-relative")
+			})
 		})
 
 		Context("Applying models", func() {

core/http/endpoints/localai/video.go

@@ -22,12 +22,19 @@ import (
 	"github.com/mudler/LocalAI/core/backend"
 
 	model "github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/utils"
 	"github.com/mudler/xlog"
 )
 
+var videoDownloadClient = http.Client{Timeout: 30 * time.Second}
+
 func downloadFile(url string) (string, error) {
+	if err := utils.ValidateExternalURL(url); err != nil {
+		return "", fmt.Errorf("URL validation failed: %w", err)
+	}
+
 	// Get the data
-	resp, err := http.Get(url)
+	resp, err := videoDownloadClient.Get(url)
 	if err != nil {
 		return "", err
 	}

core/http/endpoints/openai/chat.go

@@ -131,13 +131,19 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 				delta.Reasoning = &reasoningDelta
 			}
 
+			// Usage rides as a struct field for the consumer to track the
+			// running cumulative — it is stripped before JSON marshal so the
+			// wire chunk stays spec-compliant (no `usage` on intermediate
+			// chunks). The dedicated trailer chunk (when include_usage=true)
+			// carries the final totals.
+			usageForChunk := usage
 			resp := schema.OpenAIResponse{
 				ID:      id,
 				Created: created,
 				Model:   req.Model, // we have to return what the user sent here, due to OpenAI spec.
 				Choices: []schema.Choice{{Delta: delta, Index: 0, FinishReason: nil}},
 				Object:  "chat.completion.chunk",
-				Usage:   usage,
+				Usage:   &usageForChunk,
 			}
 
 			responses <- resp
@@ -164,7 +170,7 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 		hasChatDeltaToolCalls := false
 		hasChatDeltaContent := false
 
-		_, tokenUsage, chatDeltas, err := ComputeChoices(req, prompt, config, cl, startupOptions, loader, func(s string, c *[]schema.Choice) {}, func(s string, usage backend.TokenUsage) bool {
+		_, _, chatDeltas, err := ComputeChoices(req, prompt, config, cl, startupOptions, loader, func(s string, c *[]schema.Choice) {}, func(s string, usage backend.TokenUsage) bool {
 			result += s
 
 			// Track whether ChatDeltas from the C++ autoparser contain
@@ -387,16 +393,11 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 
 		switch {
 		case noActionToRun:
-			usage := schema.OpenAIUsage{
-				PromptTokens:     tokenUsage.Prompt,
-				CompletionTokens: tokenUsage.Completion,
-				TotalTokens:      tokenUsage.Prompt + tokenUsage.Completion,
-			}
-			if extraUsage {
-				usage.TimingTokenGeneration = tokenUsage.TimingTokenGeneration
-				usage.TimingPromptProcessing = tokenUsage.TimingPromptProcessing
-			}
-
+			// Token-cumulative usage is communicated to the streaming
+			// consumer via the per-token callback's chunk struct (stripped
+			// before wire marshal). The final usage trailer — when the
+			// caller opted in with stream_options.include_usage — is built
+			// by the outer streaming loop, not here.
 			var result string
 			if !sentInitialRole {
 				var hqErr error
@@ -409,7 +410,7 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 			for _, chunk := range buildNoActionFinalChunks(
 				id, req.Model, created,
 				sentInitialRole, sentReasoning,
-				result, reasoning, usage,
+				result, reasoning,
 			) {
 				responses <- chunk
 			}
@@ -724,7 +725,13 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 							xlog.Debug("No choices in the response, skipping")
 							continue
 						}
-						usage = &ev.Usage // Copy a pointer to the latest usage chunk so that the stop message can reference it
+						// Capture the running cumulative usage from this chunk
+						// (when present) so the include_usage trailer can carry
+						// the final totals. Usage is stripped before marshal
+						// below so the wire chunk stays spec-compliant.
+						if ev.Usage != nil {
+							usage = ev.Usage
+						}
 						if len(ev.Choices[0].Delta.ToolCalls) > 0 {
 							toolsCalled = true
 							// Collect and merge tool call deltas for MCP execution
@@ -740,6 +747,11 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 								collectedContent += *sp
 							}
 						}
+						// OpenAI streaming spec: intermediate chunks must NOT
+						// carry a `usage` field. Strip the tracking copy
+						// before marshalling — usage is delivered via the
+						// dedicated trailer chunk when include_usage=true.
+						ev.Usage = nil
 						respData, err := json.Marshal(ev)
 						if err != nil {
 							xlog.Debug("Failed to marshal response", "error", err)
@@ -888,6 +900,9 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 					finishReason = FinishReasonFunctionCall
 				}
 
+				// Final delta chunk: empty delta with finish_reason set. Per
+				// OpenAI streaming spec this chunk does NOT carry usage —
+				// the optional trailer (below) does, gated on include_usage.
 				resp := &schema.OpenAIResponse{
 					ID:      id,
 					Created: created,
@@ -899,11 +914,18 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 							Delta:        &schema.Message{},
 						}},
 					Object: "chat.completion.chunk",
-					Usage:  *usage,
 				}
 				respData, _ := json.Marshal(resp)
-
 				fmt.Fprintf(c.Response().Writer, "data: %s\n\n", respData)
+
+				// Trailing usage chunk per OpenAI spec: emit only when the
+				// caller opted in via stream_options.include_usage. Shape:
+				// {"choices":[],"usage":{...},"object":"chat.completion.chunk",...}
+				if input.StreamOptions != nil && input.StreamOptions.IncludeUsage && usage != nil {
+					trailer := streamUsageTrailerJSON(id, input.Model, created, *usage)
+					_, _ = fmt.Fprintf(c.Response().Writer, "data: %s\n\n", trailer)
+				}
+
 				fmt.Fprintf(c.Response().Writer, "data: [DONE]\n\n")
 				c.Response().Flush()
 				xlog.Debug("Stream ended")
@@ -1263,7 +1285,7 @@ func ChatEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 					Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
 					Choices: result,
 					Object:  "chat.completion",
-					Usage:   usage,
+					Usage:   &usage,
 				}
 				respData, _ := json.Marshal(resp)
 				xlog.Debug("Response", "response", string(respData))

core/http/endpoints/openai/chat_emit.go

@@ -1,12 +1,45 @@
 package openai
 
 import (
+	"encoding/json"
 	"fmt"
 
 	"github.com/mudler/LocalAI/core/schema"
 	"github.com/mudler/LocalAI/pkg/functions"
 )
 
+// streamUsageTrailerJSON returns the bytes of the OpenAI-spec trailing usage
+// chunk emitted in streaming completions when the request opts in via
+// `stream_options.include_usage: true`. The shape is:
+//
+//	{"id":"...","object":"chat.completion.chunk","created":N,
+//	 "model":"...","choices":[],"usage":{...}}
+//
+// `choices` is intentionally an empty array (not absent, not null) — that is
+// what the OpenAI spec mandates, and what consumers like the official OpenAI
+// SDK and Continue's openai-adapter look for to recognise this as the usage
+// chunk rather than a content chunk. schema.OpenAIResponse has `omitempty`
+// on Choices, so we cannot reuse it for the trailer.
+func streamUsageTrailerJSON(id, model string, created int, usage schema.OpenAIUsage) []byte {
+	trailer := struct {
+		ID      string             `json:"id"`
+		Created int                `json:"created"`
+		Model   string             `json:"model"`
+		Object  string             `json:"object"`
+		Choices []schema.Choice    `json:"choices"`
+		Usage   schema.OpenAIUsage `json:"usage"`
+	}{
+		ID:      id,
+		Created: created,
+		Model:   model,
+		Object:  "chat.completion.chunk",
+		Choices: []schema.Choice{},
+		Usage:   usage,
+	}
+	b, _ := json.Marshal(trailer)
+	return b
+}
+
 // hasRealCall reports whether functionResults contains at least one
 // entry whose Name is something other than the noAction sentinel.
 // Used by processTools to decide between the "answer the question"
@@ -25,10 +58,10 @@ func hasRealCall(functionResults []functions.FuncCallResults, noAction string) b
 // pseudo-function or emitted no tool calls at all).
 //
 // When content was already streamed (contentAlreadyStreamed=true) the
-// helper emits a single trailing usage chunk, optionally carrying
-// reasoning that was produced but not streamed incrementally. When
-// content was not streamed it emits a role chunk followed by a
-// content+reasoning+usage chunk — the "send everything at once" fallback.
+// helper emits a trailing reasoning chunk if any non-streamed reasoning
+// remains, else nothing. When content was not streamed it emits a role
+// chunk followed by a content (+reasoning) chunk — the "send everything
+// at once" fallback.
 //
 // Reasoning re-emission is guarded by reasoningAlreadyStreamed, not by
 // probing the extractor's Go-side state: the C++ autoparser delivers
@@ -36,6 +69,10 @@ func hasRealCall(functionResults []functions.FuncCallResults, noAction string) b
 // separate accumulator that extractor.Reasoning() does not expose.
 // Without this guard the callback would stream reasoning incrementally
 // and the final chunk would duplicate it.
+//
+// The returned chunks intentionally do NOT carry a `usage` field. The
+// usage trailer is emitted separately by the streaming handler when
+// `stream_options.include_usage` is true, per OpenAI spec.
 func buildNoActionFinalChunks(
 	id, model string,
 	created int,
@@ -43,26 +80,26 @@ func buildNoActionFinalChunks(
 	reasoningAlreadyStreamed bool,
 	content string,
 	reasoning string,
-	usage schema.OpenAIUsage,
 ) []schema.OpenAIResponse {
 	var out []schema.OpenAIResponse
 
 	if contentAlreadyStreamed {
-		delta := &schema.Message{}
-		if reasoning != "" && !reasoningAlreadyStreamed {
-			r := reasoning
-			delta.Reasoning = &r
+		if reasoning == "" || reasoningAlreadyStreamed {
+			return nil
 		}
+		r := reasoning
 		out = append(out, schema.OpenAIResponse{
 			ID: id, Created: created, Model: model,
-			Choices: []schema.Choice{{Delta: delta, Index: 0}},
-			Object:  "chat.completion.chunk",
-			Usage:   usage,
+			Choices: []schema.Choice{{
+				Delta: &schema.Message{Reasoning: &r},
+				Index: 0,
+			}},
+			Object: "chat.completion.chunk",
 		})
 		return out
 	}
 
-	// Content was not streamed — send role, then content (+reasoning) + usage.
+	// Content was not streamed — send role, then content (+reasoning).
 	out = append(out, schema.OpenAIResponse{
 		ID: id, Created: created, Model: model,
 		Choices: []schema.Choice{{
@@ -82,7 +119,6 @@ func buildNoActionFinalChunks(
 		ID: id, Created: created, Model: model,
 		Choices: []schema.Choice{{Delta: delta, Index: 0}},
 		Object:  "chat.completion.chunk",
-		Usage:   usage,
 	})
 	return out
 }

core/http/endpoints/openai/chat_emit_test.go

@@ -609,54 +609,52 @@ var _ = Describe("buildNoActionFinalChunks", func() {
 		testModel   = "test-model"
 		testCreated = 1700000000
 	)
-	usage := schema.OpenAIUsage{PromptTokens: 5, CompletionTokens: 7, TotalTokens: 12}
 
-	Describe("Content streamed — trailing usage chunk", func() {
-		It("emits just one chunk with usage, no content, no reasoning when reasoning was streamed", func() {
+	Describe("Content streamed — trailing reasoning only", func() {
+		It("emits nothing when content and reasoning were already streamed", func() {
+			// Before the streaming-usage-spec fix this branch emitted a
+			// content-less chunk solely to carry `usage`. Per the OpenAI
+			// spec usage no longer rides on delta chunks; the dedicated
+			// trailer (when include_usage=true) carries it instead — so
+			// with nothing to deliver the helper returns no chunks.
 			chunks := buildNoActionFinalChunks(
 				testID, testModel, testCreated,
 				true, true,
-				"", "already-streamed-reasoning", usage,
+				"", "already-streamed-reasoning",
 			)
-
-			Expect(chunks).To(HaveLen(1))
-			Expect(chunks[0].Usage.TotalTokens).To(Equal(12))
-			Expect(contentOf(chunks[0])).To(BeEmpty())
-			Expect(reasoningOf(chunks[0])).To(BeEmpty(),
-				"reasoning must not be re-emitted once it was streamed via the callback")
+			Expect(chunks).To(BeEmpty())
 		})
 
 		It("emits a trailing reasoning delivery when reasoning came only at end", func() {
 			chunks := buildNoActionFinalChunks(
 				testID, testModel, testCreated,
 				true, false,
-				"", "autoparser final reasoning", usage,
+				"", "autoparser final reasoning",
 			)
 
 			Expect(chunks).To(HaveLen(1))
 			Expect(reasoningOf(chunks[0])).To(Equal("autoparser final reasoning"))
 			Expect(contentOf(chunks[0])).To(BeEmpty())
-			Expect(chunks[0].Usage.TotalTokens).To(Equal(12))
+			Expect(chunks[0].Usage).To(BeNil(),
+				"intermediate chunks must not carry usage per OpenAI spec")
 		})
 
-		It("omits reasoning when it's empty regardless of streamed flag", func() {
+		It("returns no chunks when reasoning is empty and content was streamed", func() {
 			chunks := buildNoActionFinalChunks(
 				testID, testModel, testCreated,
 				true, false,
-				"", "", usage,
+				"", "",
 			)
-
-			Expect(chunks).To(HaveLen(1))
-			Expect(reasoningOf(chunks[0])).To(BeEmpty())
+			Expect(chunks).To(BeEmpty())
 		})
 	})
 
-	Describe("Content not streamed — role, then content+usage", func() {
+	Describe("Content not streamed — role, then content", func() {
 		It("emits role chunk then content chunk without reasoning when reasoning was streamed", func() {
 			chunks := buildNoActionFinalChunks(
 				testID, testModel, testCreated,
 				false, true,
-				"the answer", "already-streamed-reasoning", usage,
+				"the answer", "already-streamed-reasoning",
 			)
 
 			Expect(chunks).To(HaveLen(2))
@@ -666,14 +664,14 @@ var _ = Describe("buildNoActionFinalChunks", func() {
 			Expect(contentOf(chunks[1])).To(Equal("the answer"))
 			Expect(reasoningOf(chunks[1])).To(BeEmpty(),
 				"reasoning must not be re-emitted if it was streamed earlier")
-			Expect(chunks[1].Usage.TotalTokens).To(Equal(12))
+			Expect(chunks[1].Usage).To(BeNil())
 		})
 
 		It("emits role, then content+reasoning when reasoning was not streamed", func() {
 			chunks := buildNoActionFinalChunks(
 				testID, testModel, testCreated,
 				false, false,
-				"the answer", "autoparser final reasoning", usage,
+				"the answer", "autoparser final reasoning",
 			)
 
 			Expect(chunks).To(HaveLen(2))
@@ -681,14 +679,14 @@ var _ = Describe("buildNoActionFinalChunks", func() {
 
 			Expect(contentOf(chunks[1])).To(Equal("the answer"))
 			Expect(reasoningOf(chunks[1])).To(Equal("autoparser final reasoning"))
-			Expect(chunks[1].Usage.TotalTokens).To(Equal(12))
+			Expect(chunks[1].Usage).To(BeNil())
 		})
 
 		It("still emits content even when reasoning is empty", func() {
 			chunks := buildNoActionFinalChunks(
 				testID, testModel, testCreated,
 				false, false,
-				"just an answer", "", usage,
+				"just an answer", "",
 			)
 
 			Expect(chunks).To(HaveLen(2))
@@ -702,7 +700,7 @@ var _ = Describe("buildNoActionFinalChunks", func() {
 			chunks := buildNoActionFinalChunks(
 				testID, testModel, testCreated,
 				false, false,
-				"hi", "reasoning", usage,
+				"hi", "reasoning",
 			)
 			for i, ch := range chunks {
 				Expect(ch.ID).To(Equal(testID), "chunk[%d] ID", i)

core/http/endpoints/openai/chat_stream_usage_test.go

@@ -0,0 +1,179 @@
+package openai
+
+import (
+	"encoding/json"
+
+	"github.com/mudler/LocalAI/core/schema"
+	"github.com/mudler/LocalAI/pkg/functions"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+// These tests pin LocalAI's streaming chunks to the OpenAI spec for the
+// `usage` field. The regression that motivated them (issue #8546) was that
+// LocalAI emitted `"usage":{...zeros...}` on every chunk, which made the
+// official OpenAI Node SDK consumers (Continue, Kilo Code, Roo Code, Zed,
+// IntelliJ Continue) drop every content chunk via the filter at
+// continuedev/continue packages/openai-adapters/src/apis/OpenAI.ts:275-288.
+//
+// Per OpenAI's chat-completion streaming contract:
+//   - intermediate chunks MUST NOT carry a `usage` field
+//   - usage is only delivered when the request opts in via
+//     `stream_options.include_usage: true`, on a final extra chunk whose
+//     `choices` is an empty array.
+
+var _ = Describe("streaming usage spec compliance", func() {
+	Describe("OpenAIResponse JSON shape", func() {
+		It("does not emit a 'usage' key when Usage is unset", func() {
+			// A typical intermediate token chunk: no Usage populated.
+			content := "hello"
+			resp := schema.OpenAIResponse{
+				ID:      "req-1",
+				Created: 1,
+				Model:   "m",
+				Object:  "chat.completion.chunk",
+				Choices: []schema.Choice{{
+					Index: 0,
+					Delta: &schema.Message{Content: &content},
+				}},
+			}
+			data, err := json.Marshal(resp)
+			Expect(err).ToNot(HaveOccurred())
+
+			var raw map[string]any
+			Expect(json.Unmarshal(data, &raw)).To(Succeed())
+			_, present := raw["usage"]
+			Expect(present).To(BeFalse(),
+				"intermediate chunk must not include a 'usage' key; got: %s", string(data))
+		})
+
+		It("emits the usage object when Usage is explicitly set", func() {
+			usage := &schema.OpenAIUsage{PromptTokens: 11, CompletionTokens: 22, TotalTokens: 33}
+			resp := schema.OpenAIResponse{
+				ID:      "req-1",
+				Created: 1,
+				Model:   "m",
+				Object:  "chat.completion.chunk",
+				Usage:   usage,
+			}
+			data, err := json.Marshal(resp)
+			Expect(err).ToNot(HaveOccurred())
+
+			var raw map[string]any
+			Expect(json.Unmarshal(data, &raw)).To(Succeed())
+			u, ok := raw["usage"].(map[string]any)
+			Expect(ok).To(BeTrue(), "expected 'usage' object, got: %s", string(data))
+			Expect(u["prompt_tokens"]).To(BeNumerically("==", 11))
+			Expect(u["completion_tokens"]).To(BeNumerically("==", 22))
+			Expect(u["total_tokens"]).To(BeNumerically("==", 33))
+		})
+	})
+
+	Describe("buildNoActionFinalChunks", func() {
+		It("returns chunks with no Usage embedded", func() {
+			// Whatever the caller is doing, helpers must not bake usage
+			// into intermediate or final delta chunks. The usage trailer
+			// (when requested via include_usage) is emitted separately.
+			chunks := buildNoActionFinalChunks(
+				"req-1", "m", 1,
+				false, false,
+				"hi", "",
+			)
+			Expect(chunks).ToNot(BeEmpty())
+			for i, ch := range chunks {
+				Expect(ch.Usage).To(BeNil(),
+					"chunk[%d] must not carry Usage; got %+v", i, ch.Usage)
+			}
+		})
+
+		It("returns chunks with no Usage when only trailing reasoning needs delivery", func() {
+			chunks := buildNoActionFinalChunks(
+				"req-1", "m", 1,
+				true, false,
+				"", "autoparser late reasoning",
+			)
+			Expect(chunks).ToNot(BeEmpty())
+			for i, ch := range chunks {
+				Expect(ch.Usage).To(BeNil(),
+					"chunk[%d] must not carry Usage; got %+v", i, ch.Usage)
+			}
+		})
+	})
+
+	Describe("buildDeferredToolCallChunks", func() {
+		It("returns chunks with no Usage embedded", func() {
+			calls := []functions.FuncCallResults{{
+				Name: "do_thing", Arguments: `{"x":1}`,
+			}}
+			chunks := buildDeferredToolCallChunks(
+				"req-1", "m", 1, calls, 0,
+				false, "", false, "",
+			)
+			Expect(chunks).ToNot(BeEmpty())
+			for i, ch := range chunks {
+				Expect(ch.Usage).To(BeNil(),
+					"chunk[%d] must not carry Usage; got %+v", i, ch.Usage)
+			}
+		})
+	})
+
+	Describe("streamUsageTrailerJSON", func() {
+		It("produces JSON matching the OpenAI spec for the trailer chunk", func() {
+			// Trailing usage chunk shape (OpenAI streaming spec):
+			//   {"id":"...","object":"chat.completion.chunk","created":...,
+			//    "model":"...","choices":[],"usage":{...}}
+			usage := schema.OpenAIUsage{
+				PromptTokens: 18, CompletionTokens: 14, TotalTokens: 32,
+			}
+			data := streamUsageTrailerJSON("req-1", "m", 1, usage)
+
+			var raw map[string]any
+			Expect(json.Unmarshal(data, &raw)).To(Succeed(),
+				"trailer must be valid JSON, got: %s", string(data))
+
+			Expect(raw["id"]).To(Equal("req-1"))
+			Expect(raw["model"]).To(Equal("m"))
+			Expect(raw["object"]).To(Equal("chat.completion.chunk"))
+			Expect(raw["created"]).To(BeNumerically("==", 1))
+
+			// `choices` MUST be present as an empty array (not absent, not null).
+			rawChoices, present := raw["choices"]
+			Expect(present).To(BeTrue(), "choices key must be present, got: %s", string(data))
+			choicesArr, ok := rawChoices.([]any)
+			Expect(ok).To(BeTrue(), "choices must serialize as an array, got: %s", string(data))
+			Expect(choicesArr).To(BeEmpty(), "choices must be empty in usage trailer, got: %s", string(data))
+
+			// `usage` MUST be present and non-null with the populated counts.
+			u, ok := raw["usage"].(map[string]any)
+			Expect(ok).To(BeTrue(), "usage object must be present, got: %s", string(data))
+			Expect(u["prompt_tokens"]).To(BeNumerically("==", 18))
+			Expect(u["completion_tokens"]).To(BeNumerically("==", 14))
+			Expect(u["total_tokens"]).To(BeNumerically("==", 32))
+		})
+	})
+
+	Describe("OpenAIRequest.StreamOptions", func() {
+		It("parses stream_options.include_usage=true", func() {
+			body := []byte(`{
+                "model": "m",
+                "stream": true,
+                "stream_options": {"include_usage": true},
+                "messages": []
+            }`)
+			var req schema.OpenAIRequest
+			Expect(json.Unmarshal(body, &req)).To(Succeed())
+			Expect(req.StreamOptions).ToNot(BeNil())
+			Expect(req.StreamOptions.IncludeUsage).To(BeTrue())
+		})
+
+		It("defaults IncludeUsage to false when stream_options is absent", func() {
+			body := []byte(`{"model":"m","stream":true,"messages":[]}`)
+			var req schema.OpenAIRequest
+			Expect(json.Unmarshal(body, &req)).To(Succeed())
+			// Either a nil StreamOptions or one with IncludeUsage=false is acceptable.
+			if req.StreamOptions != nil {
+				Expect(req.StreamOptions.IncludeUsage).To(BeFalse())
+			}
+		})
+	})
+})

core/http/endpoints/openai/completion.go

@@ -39,6 +39,10 @@ func CompletionEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, eva
 				usage.TimingTokenGeneration = tokenUsage.TimingTokenGeneration
 				usage.TimingPromptProcessing = tokenUsage.TimingPromptProcessing
 			}
+			// Usage rides on the struct for the consumer to track the
+			// running cumulative; the consumer strips it before marshalling
+			// so intermediate chunks stay OpenAI-spec compliant.
+			usageForChunk := usage
 			resp := schema.OpenAIResponse{
 				ID:      id,
 				Created: created,
@@ -51,7 +55,7 @@ func CompletionEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, eva
 					},
 				},
 				Object: "text_completion",
-				Usage:  usage,
+				Usage:  &usageForChunk,
 			}
 			xlog.Debug("Sending goroutine", "text", s)
 
@@ -127,6 +131,8 @@ func CompletionEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, eva
 				ended <- process(id, predInput, input, config, ml, responses, extraUsage)
 			}()
 
+			var latestUsage *schema.OpenAIUsage
+
 		LOOP:
 			for {
 				select {
@@ -135,6 +141,14 @@ func CompletionEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, eva
 						xlog.Debug("No choices in the response, skipping")
 						continue
 					}
+					// Capture running cumulative usage for the optional trailer
+					// emitted after the final stop chunk when include_usage=true.
+					if ev.Usage != nil {
+						latestUsage = ev.Usage
+					}
+					// OpenAI streaming spec: intermediate chunks must NOT
+					// carry a `usage` field. Strip the tracking copy now.
+					ev.Usage = nil
 					respData, err := json.Marshal(ev)
 					if err != nil {
 						xlog.Debug("Failed to marshal response", "error", err)
@@ -194,8 +208,15 @@ func CompletionEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, eva
 				Object: "text_completion",
 			}
 			respData, _ := json.Marshal(resp)
-
 			fmt.Fprintf(c.Response().Writer, "data: %s\n\n", respData)
+
+			// Trailing usage chunk per OpenAI spec: emit only when the caller
+			// opted in via stream_options.include_usage.
+			if input.StreamOptions != nil && input.StreamOptions.IncludeUsage && latestUsage != nil {
+				trailer := streamUsageTrailerJSON(id, input.Model, created, *latestUsage)
+				_, _ = fmt.Fprintf(c.Response().Writer, "data: %s\n\n", trailer)
+			}
+
 			fmt.Fprintf(c.Response().Writer, "data: [DONE]\n\n")
 			c.Response().Flush()
 			return nil
@@ -247,7 +268,7 @@ func CompletionEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, eva
 			Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
 			Choices: result,
 			Object:  "text_completion",
-			Usage:   usage,
+			Usage:   &usage,
 		}
 
 		jsonResult, _ := json.Marshal(resp)

core/http/endpoints/openai/edit.go

@@ -92,7 +92,7 @@ func EditEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, evaluator
 			Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
 			Choices: result,
 			Object:  "edit",
-			Usage:   usage,
+			Usage:   &usage,
 		}
 
 		jsonResult, _ := json.Marshal(resp)

core/http/endpoints/openai/image.go

@@ -233,7 +233,7 @@ func ImageEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, appConfi
 			ID:      id,
 			Created: created,
 			Data:    result,
-			Usage: schema.OpenAIUsage{
+			Usage: &schema.OpenAIUsage{
 				PromptTokens:     0,
 				CompletionTokens: 0,
 				TotalTokens:      0,

core/http/endpoints/openai/inpainting.go

@@ -258,7 +258,7 @@ func InpaintingEndpoint(cl *config.ModelConfigLoader, ml *model.ModelLoader, app
 			Data: []schema.Item{{
 				URL: imgPath,
 			}},
-			Usage: schema.OpenAIUsage{
+			Usage: &schema.OpenAIUsage{
 				PromptTokens:     0,
 				CompletionTokens: 0,
 				TotalTokens:      0,

core/http/endpoints/openai/realtime.go

@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"math"
 	"os"
+	"strconv"
 	"sync"
 	"time"
 
@@ -20,6 +21,8 @@ import (
 	"github.com/mudler/LocalAI/core/backend"
 
 	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/http/auth"
+	mcpTools "github.com/mudler/LocalAI/core/http/endpoints/mcp"
 	"github.com/mudler/LocalAI/core/http/endpoints/openai/types"
 	"github.com/mudler/LocalAI/core/schema"
 	"github.com/mudler/LocalAI/core/templates"
@@ -51,6 +54,30 @@ const (
 		"Avoid parenthetical asides, URLs, and anything that cannot be clearly vocalized."
 )
 
+// resolveOutputModalities returns the effective output modalities for a
+// response: response-level overrides session-level, and the OpenAI Realtime
+// spec default is ["audio"] when neither is set.
+func resolveOutputModalities(session, response []types.Modality) []types.Modality {
+	if len(response) > 0 {
+		return response
+	}
+	if len(session) > 0 {
+		return session
+	}
+	return []types.Modality{types.ModalityAudio}
+}
+
+// modalitiesContainAudio reports whether the resolved modalities include audio
+// output.
+func modalitiesContainAudio(m []types.Modality) bool {
+	for _, x := range m {
+		if x == types.ModalityAudio {
+			return true
+		}
+	}
+	return false
+}
+
 // A model can be "emulated" that is: transcribe audio to text -> feed text to the LLM -> generate audio as result
 // If the model support instead audio-to-audio, we will use the specific gRPC calls instead
 
@@ -79,6 +106,30 @@ type Session struct {
 	InputSampleRate  int
 	OutputSampleRate int
 	MaxOutputTokens  types.IntOrInf
+	// OutputModalities mirrors the OpenAI Realtime spec field of the same
+	// name. Empty means "use the spec default" (audio). ["text"] suppresses
+	// TTS so the client receives only response.output_text.* events.
+	OutputModalities []types.Modality
+	// MaxHistoryItems caps the number of MessageItems passed to the LLM each
+	// turn (0 = unlimited). Small models — especially the LFM2.5-Audio 1.5B
+	// served via the liquid-audio backend — degrade quickly past a handful
+	// of turns. Counted from the tail; FunctionCall + FunctionCallOutput
+	// pairs are kept together so we never feed an orphaned tool result.
+	MaxHistoryItems int
+
+	// AssistantExecutor is non-nil when the session opted into the in-process
+	// LocalAI Assistant tool surface. Tool calls whose name matches this
+	// executor's catalog are run inproc and their output is fed back to the
+	// model server-side; the client never sees a function_call_arguments
+	// event for those. Mirrors the chat handler's metadata.localai_assistant
+	// path.
+	AssistantExecutor mcpTools.ToolExecutor
+
+	// AssistantTools is the cached ToolUnion slice we injected at session
+	// creation. Re-applied after every client session.update so a
+	// client-driven tool refresh (e.g. toggling a client MCP server) doesn't
+	// silently strip Manage Mode's tools.
+	AssistantTools []types.ToolUnion
 
 	// Response cancellation: protects activeResponseCancel/activeResponseDone
 	responseMu           sync.Mutex
@@ -139,13 +190,14 @@ func (s *Session) ToServer() types.SessionUnion {
 	} else {
 		return types.SessionUnion{
 			Realtime: &types.RealtimeSession{
-				ID:              s.ID,
-				Object:          "realtime.session",
-				Model:           s.Model,
-				Instructions:    s.Instructions,
-				Tools:           s.Tools,
-				ToolChoice:      s.ToolChoice,
-				MaxOutputTokens: s.MaxOutputTokens,
+				ID:               s.ID,
+				Object:           "realtime.session",
+				Model:            s.Model,
+				Instructions:     s.Instructions,
+				Tools:            s.Tools,
+				ToolChoice:       s.ToolChoice,
+				MaxOutputTokens:  s.MaxOutputTokens,
+				OutputModalities: s.OutputModalities,
 				Audio: &types.RealtimeSessionAudio{
 					Input: &types.SessionAudioInput{
 						TurnDetection: s.TurnDetection,
@@ -205,6 +257,19 @@ func RealtimeTranscriptionSession(application *application.Application) echo.Han
 	}
 }
 
+// RealtimeSessionOptions bundles per-session knobs decoded from the WS query
+// string (or the WebRTC handshake body). Mirrors what chat.go pulls off
+// `metadata.localai_assistant` — admin-only opt-in to the in-process
+// management tool surface.
+type RealtimeSessionOptions struct {
+	LocalAIAssistant bool
+	// AuthEnabled mirrors chat.go's requireAssistantAccess gate. We resolve
+	// admin role at handshake time (where the echo.Context has the auth
+	// cookie/Bearer) and drop the result here so runRealtimeSession can
+	// decide without holding onto the request.
+	IsAdmin bool
+}
+
 func Realtime(application *application.Application) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		ws, err := upgrader.Upgrade(c.Response(), c.Request(), nil)
@@ -218,25 +283,105 @@ func Realtime(application *application.Application) echo.HandlerFunc {
 
 		// Extract query parameters from Echo context before passing to websocket handler
 		model := c.QueryParam("model")
+		assistantFlag, _ := strconv.ParseBool(c.QueryParam("localai_assistant"))
+		opts := RealtimeSessionOptions{
+			LocalAIAssistant: assistantFlag,
+			IsAdmin:          isCurrentUserAdmin(c, application),
+		}
 
-		registerRealtime(application, model)(ws)
+		registerRealtime(application, model, opts)(ws)
 		return nil
 	}
 }
 
-func registerRealtime(application *application.Application, model string) func(c *websocket.Conn) {
+// isCurrentUserAdmin replicates the chat-side admin check at the realtime
+// handshake. When auth is disabled, every caller is treated as admin (same
+// as chat's requireAssistantAccess).
+func isCurrentUserAdmin(c echo.Context, application *application.Application) bool {
+	if application == nil || application.ApplicationConfig() == nil || !application.ApplicationConfig().Auth.Enabled {
+		return true
+	}
+	user := auth.GetUser(c)
+	return user != nil && user.Role == auth.RoleAdmin
+}
+
+func registerRealtime(application *application.Application, model string, opts RealtimeSessionOptions) func(c *websocket.Conn) {
 	return func(conn *websocket.Conn) {
 		t := NewWebSocketTransport(conn)
 		evaluator := application.TemplatesEvaluator()
 		xlog.Debug("Realtime WebSocket connection established", "address", conn.RemoteAddr().String(), "model", model)
-		runRealtimeSession(application, t, model, evaluator)
+		runRealtimeSession(application, t, model, evaluator, opts)
 	}
 }
 
+// defaultMaxHistoryItems picks a sensible default cap for the session.
+// Small any-to-any audio models degrade quickly past a handful of turns;
+// legacy pipelines composing larger LLMs keep the historical "unlimited"
+// default and rely on the LLM's own context window.
+func defaultMaxHistoryItems(cfg *config.ModelConfig) int {
+	if cfg != nil && cfg.HasUsecases(config.FLAG_REALTIME_AUDIO) {
+		return 6
+	}
+	return 0
+}
+
+// trimRealtimeItems returns the tail of items capped at maxItems (0 = no cap).
+// Walks backwards keeping function_call + function_call_output pairs together
+// so we never feed the LLM an orphaned tool result that references a call it
+// can't see.
+func trimRealtimeItems(items []*types.MessageItemUnion, maxItems int) []*types.MessageItemUnion {
+	if maxItems <= 0 || len(items) <= maxItems {
+		return items
+	}
+	// Find the cut point starting from len-maxItems and pull it left until
+	// we're not in the middle of a tool-call pair.
+	cut := len(items) - maxItems
+	for cut > 0 && items[cut] != nil && items[cut].FunctionCallOutput != nil {
+		cut--
+	}
+	return items[cut:]
+}
+
+// prepareRealtimeConfig validates a model config for use in a realtime session
+// and fills in pipeline slots for self-contained any-to-any models. It returns
+// an error code + message pair suitable for sendError; the bool indicates
+// whether the caller should proceed. Extracted from runRealtimeSession so the
+// gate logic can be exercised in unit tests without a full Application.
+func prepareRealtimeConfig(cfg *config.ModelConfig) (errCode, errMsg string, ok bool) {
+	if cfg == nil {
+		return "invalid_model", "Model is not a pipeline model", false
+	}
+
+	// Self-contained any-to-any models (e.g. liquid-audio) own the whole
+	// loop in one engine — surface them by populating empty pipeline slots
+	// with the model's own name so newModel can resolve a config for each
+	// role. The user can still pin individual slots (e.g. Pipeline.VAD =
+	// silero-vad) and those wins.
+	if cfg.HasUsecases(config.FLAG_REALTIME_AUDIO) {
+		if cfg.Pipeline.VAD == "" {
+			cfg.Pipeline.VAD = cfg.Name
+		}
+		if cfg.Pipeline.Transcription == "" {
+			cfg.Pipeline.Transcription = cfg.Name
+		}
+		if cfg.Pipeline.LLM == "" {
+			cfg.Pipeline.LLM = cfg.Name
+		}
+		if cfg.Pipeline.TTS == "" {
+			cfg.Pipeline.TTS = cfg.Name
+		}
+		return "", "", true
+	}
+
+	if cfg.Pipeline.VAD == "" && cfg.Pipeline.Transcription == "" && cfg.Pipeline.TTS == "" && cfg.Pipeline.LLM == "" {
+		return "invalid_model", "Model is not a pipeline model", false
+	}
+	return "", "", true
+}
+
 // runRealtimeSession runs the main event loop for a realtime session.
 // It is transport-agnostic and works with both WebSocket and WebRTC.
-func runRealtimeSession(application *application.Application, t Transport, model string, evaluator *templates.Evaluator) {
-	// TODO: Allow any-to-any model to be specified
+func runRealtimeSession(application *application.Application, t Transport, model string, evaluator *templates.Evaluator, opts RealtimeSessionOptions) {
 	cl := application.ModelConfigLoader()
 	cfg, err := cl.LoadModelConfigFileByNameDefaultOptions(model, application.ApplicationConfig())
 	if err != nil {
@@ -245,22 +390,79 @@ func runRealtimeSession(application *application.Application, t Transport, model
 		return
 	}
 
-	if cfg == nil || (cfg.Pipeline.VAD == "" && cfg.Pipeline.Transcription == "" && cfg.Pipeline.TTS == "" && cfg.Pipeline.LLM == "") {
+	if code, msg, ok := prepareRealtimeConfig(cfg); !ok {
 		xlog.Error("model is not a pipeline", "model", model)
-		sendError(t, "invalid_model", "Model is not a pipeline model", "", "")
+		sendError(t, code, msg, "", "")
 		return
 	}
 
+	// LocalAI Assistant opt-in: gate on admin (same rule as chat.go's
+	// requireAssistantAccess) and grab the process-wide holder's executor.
+	// We collect tools + system prompt here and merge them into the session
+	// below so they're live from the first response.create.
+	var assistantTools []types.ToolUnion
+	var assistantSystemPrompt string
+	var assistantExecutor mcpTools.ToolExecutor
+	if opts.LocalAIAssistant {
+		if !opts.IsAdmin {
+			sendError(t, "forbidden", "localai_assistant requires admin", "", "")
+			return
+		}
+		appCfg := application.ApplicationConfig()
+		if appCfg != nil && appCfg.DisableLocalAIAssistant {
+			sendError(t, "unavailable", "LocalAI Assistant is disabled on this server", "", "")
+			return
+		}
+		holder := application.LocalAIAssistant()
+		if holder == nil || !holder.HasTools() {
+			sendError(t, "unavailable", "LocalAI Assistant is not available on this server", "", "")
+			return
+		}
+		exec := holder.Executor()
+		fns, discErr := exec.DiscoverTools(context.Background())
+		if discErr != nil {
+			xlog.Error("realtime: failed to discover LocalAI Assistant tools", "error", discErr)
+			sendError(t, "tool_discovery_failed", "failed to discover assistant tools: "+discErr.Error(), "", "")
+			return
+		}
+		assistantExecutor = exec
+		assistantSystemPrompt = holder.SystemPrompt()
+		assistantTools = make([]types.ToolUnion, 0, len(fns))
+		for _, fn := range fns {
+			fnCopy := fn
+			assistantTools = append(assistantTools, types.ToolUnion{
+				Function: &types.ToolFunction{
+					Name:        fnCopy.Name,
+					Description: fnCopy.Description,
+					Parameters:  fnCopy.Parameters,
+				},
+			})
+		}
+		xlog.Debug("realtime: LocalAI Assistant tools injected", "count", len(fns))
+	}
+
 	sttModel := cfg.Pipeline.Transcription
 
+	// Compose the system prompt: prepend the assistant prompt when we have
+	// one (it teaches the model the safety rules and tool recipes), then the
+	// session's default voice instructions. Order matches chat.go's
+	// hasSystemMessage check — assistant prompt comes first.
+	instructions := defaultInstructions
+	if assistantSystemPrompt != "" {
+		instructions = assistantSystemPrompt + "\n\n" + defaultInstructions
+	}
+
 	sessionID := generateSessionID()
 	session := &Session{
 		ID:                sessionID,
 		TranscriptionOnly: false,
 		Model:             model,
 		Voice:             cfg.TTSConfig.Voice,
-		Instructions:      defaultInstructions,
+		Instructions:      instructions,
 		ModelConfig:       cfg,
+		Tools:             assistantTools,
+		AssistantTools:    assistantTools,
+		AssistantExecutor: assistantExecutor,
 		TurnDetection: &types.TurnDetectionUnion{
 			ServerVad: &types.ServerVad{
 				Threshold:         0.5,
@@ -275,6 +477,7 @@ func runRealtimeSession(application *application.Application, t Transport, model
 		Conversations:    make(map[string]*Conversation),
 		InputSampleRate:  defaultRemoteSampleRate,
 		OutputSampleRate: defaultRemoteSampleRate,
+		MaxHistoryItems:  defaultMaxHistoryItems(cfg),
 	}
 
 	// Create a default conversation
@@ -810,7 +1013,28 @@ func updateSession(session *Session, update *types.SessionUnion, cl *config.Mode
 	}
 
 	if rt.Tools != nil {
-		session.Tools = rt.Tools
+		// Manage Mode tools survive a client-driven session.update — the
+		// alternative is silently dropping them whenever the user toggles
+		// a client MCP server, which would break the modality mid-session.
+		// Names from rt.Tools win on collision (the client is explicit;
+		// we preserve, we don't override).
+		merged := append([]types.ToolUnion(nil), rt.Tools...)
+		seen := make(map[string]struct{}, len(merged))
+		for _, t := range merged {
+			if t.Function != nil {
+				seen[t.Function.Name] = struct{}{}
+			}
+		}
+		for _, t := range session.AssistantTools {
+			if t.Function == nil {
+				continue
+			}
+			if _, ok := seen[t.Function.Name]; ok {
+				continue
+			}
+			merged = append(merged, t)
+		}
+		session.Tools = merged
 	}
 	if rt.ToolChoice != nil {
 		session.ToolChoice = rt.ToolChoice
@@ -820,6 +1044,10 @@ func updateSession(session *Session, update *types.SessionUnion, cl *config.Mode
 		session.MaxOutputTokens = rt.MaxOutputTokens
 	}
 
+	if len(rt.OutputModalities) > 0 {
+		session.OutputModalities = rt.OutputModalities
+	}
+
 	return nil
 }
 
@@ -1104,7 +1332,17 @@ func generateResponse(ctx context.Context, session *Session, utt []byte, transcr
 	triggerResponse(ctx, session, conv, t, nil)
 }
 
+// maxAssistantToolTurns caps the server-side agentic loop. Mirrors the
+// chat-page maxToolTurns:10 from useChat.js — the model gets up to this
+// many consecutive tool round-trips before we return control to the user
+// without another response cycle.
+const maxAssistantToolTurns = 10
+
 func triggerResponse(ctx context.Context, session *Session, conv *Conversation, t Transport, overrides *types.ResponseCreateParams) {
+	triggerResponseAtTurn(ctx, session, conv, t, overrides, 0)
+}
+
+func triggerResponseAtTurn(ctx context.Context, session *Session, conv *Conversation, t Transport, overrides *types.ResponseCreateParams, toolTurn int) {
 	config := session.ModelInterface.PredictConfig()
 
 	// Default values
@@ -1155,7 +1393,8 @@ func triggerResponse(ctx context.Context, session *Session, conv *Conversation,
 
 	imgIndex := 0
 	conv.Lock.Lock()
-	for _, item := range conv.Items {
+	items := trimRealtimeItems(conv.Items, session.MaxHistoryItems)
+	for _, item := range items {
 		if item.User != nil {
 			msg := schema.Message{
 				Role: string(types.MessageRoleUser),
@@ -1448,106 +1687,130 @@ func triggerResponse(ctx context.Context, session *Session, conv *Conversation,
 			})
 		}
 
-		// Check for cancellation before TTS
-		if ctx.Err() != nil {
-			xlog.Debug("Response cancelled before TTS (barge-in)")
-			sendCancelledResponse()
-			return
-		}
-
-		audioFilePath, res, err := session.ModelInterface.TTS(ctx, finalSpeech, session.Voice, session.InputAudioTranscription.Language)
-		if err != nil {
-			if ctx.Err() != nil {
-				xlog.Debug("TTS cancelled (barge-in)")
-				sendCancelledResponse()
-				return
-			}
-			xlog.Error("TTS failed", "error", err)
-			sendError(t, "tts_error", fmt.Sprintf("TTS generation failed: %v", err), "", item.Assistant.ID)
-			return
-		}
-		if !res.Success {
-			xlog.Error("TTS failed", "message", res.Message)
-			sendError(t, "tts_error", fmt.Sprintf("TTS generation failed: %s", res.Message), "", item.Assistant.ID)
-			return
-		}
-		defer os.Remove(audioFilePath)
-
-		audioBytes, err := os.ReadFile(audioFilePath)
-		if err != nil {
-			xlog.Error("failed to read TTS file", "error", err)
-			sendError(t, "tts_error", fmt.Sprintf("Failed to read TTS audio: %v", err), "", item.Assistant.ID)
-			return
-		}
-
-		// Parse WAV header to get raw PCM and the actual sample rate from the TTS backend.
-		pcmData, ttsSampleRate := laudio.ParseWAV(audioBytes)
-		if ttsSampleRate == 0 {
-			ttsSampleRate = localSampleRate
-		}
-		xlog.Debug("TTS audio parsed", "raw_bytes", len(audioBytes), "pcm_bytes", len(pcmData), "sample_rate", ttsSampleRate)
-
-		// SendAudio (WebRTC) passes PCM at the TTS sample rate directly to the
-		// Opus encoder, which resamples to 48kHz internally. This avoids a
-		// lossy intermediate resample through 16kHz.
-		// XXX: This is a noop in websocket mode; it's included in the JSON instead
-		if err := t.SendAudio(ctx, pcmData, ttsSampleRate); err != nil {
-			if ctx.Err() != nil {
-				xlog.Debug("Audio playback cancelled (barge-in)")
-				sendCancelledResponse()
-				return
-			}
-			xlog.Error("failed to send audio via transport", "error", err)
-		}
-
-		_, isWebRTC := t.(*WebRTCTransport)
-
-		// For WebSocket clients, resample to the session's output rate and
-		// deliver audio as base64 in JSON events. WebRTC clients already
-		// received audio over the RTP track, so skip the base64 payload.
 		var audioString string
-		if !isWebRTC {
-			wsPCM := pcmData
-			if ttsSampleRate != session.OutputSampleRate {
-				samples := sound.BytesToInt16sLE(pcmData)
-				resampled := sound.ResampleInt16(samples, ttsSampleRate, session.OutputSampleRate)
-				wsPCM = sound.Int16toBytesLE(resampled)
-			}
-			audioString = base64.StdEncoding.EncodeToString(wsPCM)
+		_, isWebRTC := t.(*WebRTCTransport)
+		var respMods []types.Modality
+		if overrides != nil {
+			respMods = overrides.OutputModalities
 		}
+		modalities := resolveOutputModalities(session.OutputModalities, respMods)
+		if modalitiesContainAudio(modalities) {
+			// Check for cancellation before TTS
+			if ctx.Err() != nil {
+				xlog.Debug("Response cancelled before TTS (barge-in)")
+				sendCancelledResponse()
+				return
+			}
 
-		sendEvent(t, types.ResponseOutputAudioTranscriptDeltaEvent{
-			ServerEventBase: types.ServerEventBase{},
-			ResponseID:      responseID,
-			ItemID:          item.Assistant.ID,
-			OutputIndex:     0,
-			ContentIndex:    0,
-			Delta:           finalSpeech,
-		})
-		sendEvent(t, types.ResponseOutputAudioTranscriptDoneEvent{
-			ServerEventBase: types.ServerEventBase{},
-			ResponseID:      responseID,
-			ItemID:          item.Assistant.ID,
-			OutputIndex:     0,
-			ContentIndex:    0,
-			Transcript:      finalSpeech,
-		})
+			audioFilePath, res, err := session.ModelInterface.TTS(ctx, finalSpeech, session.Voice, session.InputAudioTranscription.Language)
+			if err != nil {
+				if ctx.Err() != nil {
+					xlog.Debug("TTS cancelled (barge-in)")
+					sendCancelledResponse()
+					return
+				}
+				xlog.Error("TTS failed", "error", err)
+				sendError(t, "tts_error", fmt.Sprintf("TTS generation failed: %v", err), "", item.Assistant.ID)
+				return
+			}
+			if !res.Success {
+				xlog.Error("TTS failed", "message", res.Message)
+				sendError(t, "tts_error", fmt.Sprintf("TTS generation failed: %s", res.Message), "", item.Assistant.ID)
+				return
+			}
+			defer func() { _ = os.Remove(audioFilePath) }()
 
-		if !isWebRTC {
-			sendEvent(t, types.ResponseOutputAudioDeltaEvent{
+			audioBytes, err := os.ReadFile(audioFilePath)
+			if err != nil {
+				xlog.Error("failed to read TTS file", "error", err)
+				sendError(t, "tts_error", fmt.Sprintf("Failed to read TTS audio: %v", err), "", item.Assistant.ID)
+				return
+			}
+
+			// Parse WAV header to get raw PCM and the actual sample rate from the TTS backend.
+			pcmData, ttsSampleRate := laudio.ParseWAV(audioBytes)
+			if ttsSampleRate == 0 {
+				ttsSampleRate = localSampleRate
+			}
+			xlog.Debug("TTS audio parsed", "raw_bytes", len(audioBytes), "pcm_bytes", len(pcmData), "sample_rate", ttsSampleRate)
+
+			// SendAudio (WebRTC) passes PCM at the TTS sample rate directly to the
+			// Opus encoder, which resamples to 48kHz internally. This avoids a
+			// lossy intermediate resample through 16kHz.
+			// XXX: This is a noop in websocket mode; it's included in the JSON instead
+			if err := t.SendAudio(ctx, pcmData, ttsSampleRate); err != nil {
+				if ctx.Err() != nil {
+					xlog.Debug("Audio playback cancelled (barge-in)")
+					sendCancelledResponse()
+					return
+				}
+				xlog.Error("failed to send audio via transport", "error", err)
+			}
+
+			// For WebSocket clients, resample to the session's output rate and
+			// deliver audio as base64 in JSON events. WebRTC clients already
+			// received audio over the RTP track, so skip the base64 payload.
+			if !isWebRTC {
+				wsPCM := pcmData
+				if ttsSampleRate != session.OutputSampleRate {
+					samples := sound.BytesToInt16sLE(pcmData)
+					resampled := sound.ResampleInt16(samples, ttsSampleRate, session.OutputSampleRate)
+					wsPCM = sound.Int16toBytesLE(resampled)
+				}
+				audioString = base64.StdEncoding.EncodeToString(wsPCM)
+			}
+
+			sendEvent(t, types.ResponseOutputAudioTranscriptDeltaEvent{
 				ServerEventBase: types.ServerEventBase{},
 				ResponseID:      responseID,
 				ItemID:          item.Assistant.ID,
 				OutputIndex:     0,
 				ContentIndex:    0,
-				Delta:           audioString,
+				Delta:           finalSpeech,
 			})
-			sendEvent(t, types.ResponseOutputAudioDoneEvent{
+			sendEvent(t, types.ResponseOutputAudioTranscriptDoneEvent{
 				ServerEventBase: types.ServerEventBase{},
 				ResponseID:      responseID,
 				ItemID:          item.Assistant.ID,
 				OutputIndex:     0,
 				ContentIndex:    0,
+				Transcript:      finalSpeech,
+			})
+
+			if !isWebRTC {
+				sendEvent(t, types.ResponseOutputAudioDeltaEvent{
+					ServerEventBase: types.ServerEventBase{},
+					ResponseID:      responseID,
+					ItemID:          item.Assistant.ID,
+					OutputIndex:     0,
+					ContentIndex:    0,
+					Delta:           audioString,
+				})
+				sendEvent(t, types.ResponseOutputAudioDoneEvent{
+					ServerEventBase: types.ServerEventBase{},
+					ResponseID:      responseID,
+					ItemID:          item.Assistant.ID,
+					OutputIndex:     0,
+					ContentIndex:    0,
+				})
+			}
+		} else {
+			// Text-only mode: skip TTS, emit only the text events.
+			sendEvent(t, types.ResponseOutputTextDeltaEvent{
+				ServerEventBase: types.ServerEventBase{},
+				ResponseID:      responseID,
+				ItemID:          item.Assistant.ID,
+				OutputIndex:     0,
+				ContentIndex:    0,
+				Delta:           finalSpeech,
+			})
+			sendEvent(t, types.ResponseOutputTextDoneEvent{
+				ServerEventBase: types.ServerEventBase{},
+				ResponseID:      responseID,
+				ItemID:          item.Assistant.ID,
+				OutputIndex:     0,
+				ContentIndex:    0,
+				Text:            finalSpeech,
 			})
 		}
 
@@ -1575,8 +1838,16 @@ func triggerResponse(ctx context.Context, session *Session, conv *Conversation,
 		})
 	}
 
-	// Handle Tool Calls
+	// Handle Tool Calls. Two paths:
+	//   - LocalAI Assistant tools (session.AssistantExecutor.IsTool) run
+	//     server-side; we append both the call and its output to conv.Items
+	//     and re-trigger a follow-up response so the model can speak the
+	//     result. The client only sees observability events.
+	//   - All other tools follow the standard OpenAI flow: emit
+	//     function_call_arguments.done and wait for the client to send
+	//     conversation.item.create back.
 	xlog.Debug("About to handle tool calls", "finalToolCallsCount", len(finalToolCalls))
+	executedAssistantTool := false
 	for i, tc := range finalToolCalls {
 		toolCallID := generateItemID()
 		callID := "call_" + generateUniqueID() // OpenAI uses call_xyz
@@ -1608,6 +1879,51 @@ func triggerResponse(ctx context.Context, session *Session, conv *Conversation,
 			Item:            fcItem,
 		})
 
+		serverSide := session.AssistantExecutor != nil && session.AssistantExecutor.IsTool(tc.Name)
+		if serverSide {
+			output, execErr := session.AssistantExecutor.ExecuteTool(ctx, tc.Name, tc.Arguments)
+			if execErr != nil {
+				output = "Error: " + execErr.Error()
+				xlog.Error("realtime: assistant tool execution failed", "tool", tc.Name, "error", execErr)
+			}
+			foItem := types.MessageItemUnion{
+				FunctionCallOutput: &types.MessageItemFunctionCallOutput{
+					ID:     generateItemID(),
+					CallID: callID,
+					Output: output,
+					Status: types.ItemStatusCompleted,
+				},
+			}
+			conv.Lock.Lock()
+			conv.Items = append(conv.Items, &foItem)
+			conv.Lock.Unlock()
+			// Close the call out and emit the output as its own paired
+			// added/done — the OpenAI spec pairs every item-done with a
+			// preceding item-added, so we re-pair here for the output.
+			// The UI renders the transcript entry on item.done for both
+			// shapes (FunctionCall + FunctionCallOutput).
+			sendEvent(t, types.ResponseOutputItemDoneEvent{
+				ServerEventBase: types.ServerEventBase{},
+				ResponseID:      responseID,
+				OutputIndex:     outputIndex,
+				Item:            fcItem,
+			})
+			sendEvent(t, types.ResponseOutputItemAddedEvent{
+				ServerEventBase: types.ServerEventBase{},
+				ResponseID:      responseID,
+				OutputIndex:     outputIndex,
+				Item:            foItem,
+			})
+			sendEvent(t, types.ResponseOutputItemDoneEvent{
+				ServerEventBase: types.ServerEventBase{},
+				ResponseID:      responseID,
+				OutputIndex:     outputIndex,
+				Item:            foItem,
+			})
+			executedAssistantTool = true
+			continue
+		}
+
 		sendEvent(t, types.ResponseFunctionCallArgumentsDeltaEvent{
 			ServerEventBase: types.ServerEventBase{},
 			ResponseID:      responseID,
@@ -1643,6 +1959,19 @@ func triggerResponse(ctx context.Context, session *Session, conv *Conversation,
 			Status: types.ResponseStatusCompleted,
 		},
 	})
+
+	// If we executed any assistant tools inproc, run another response cycle
+	// so the model can speak the result. Mirrors the chat-side agentic loop
+	// but driven server-side rather than by client round-trip. Bounded so a
+	// degenerate "model keeps calling tools" doesn't blow the stack.
+	if executedAssistantTool {
+		if toolTurn+1 >= maxAssistantToolTurns {
+			xlog.Warn("realtime: assistant tool-turn limit reached, stopping the agentic loop",
+				"limit", maxAssistantToolTurns, "model", session.Model)
+			return
+		}
+		triggerResponseAtTurn(ctx, session, conv, t, nil, toolTurn+1)
+	}
 }
 
 // Helper functions to generate unique IDs

core/http/endpoints/openai/realtime_gate_test.go

@@ -0,0 +1,153 @@
+package openai
+
+import (
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/http/endpoints/openai/types"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+// withUsecases returns a *ModelConfigUsecase pointing at the OR of the given flags.
+// Helper so each spec keeps its intent obvious.
+func withUsecases(flags ...config.ModelConfigUsecase) *config.ModelConfigUsecase {
+	var u config.ModelConfigUsecase
+	for _, f := range flags {
+		u |= f
+	}
+	return &u
+}
+
+var _ = Describe("prepareRealtimeConfig", func() {
+	It("rejects a nil config", func() {
+		code, msg, ok := prepareRealtimeConfig(nil)
+		Expect(ok).To(BeFalse())
+		Expect(code).To(Equal("invalid_model"))
+		Expect(msg).To(ContainSubstring("not a pipeline model"))
+	})
+
+	It("rejects a model with no pipeline slots and no realtime_audio usecase", func() {
+		cfg := &config.ModelConfig{Name: "plain-chat"}
+		code, msg, ok := prepareRealtimeConfig(cfg)
+		Expect(ok).To(BeFalse())
+		Expect(code).To(Equal("invalid_model"))
+		Expect(msg).To(ContainSubstring("not a pipeline model"))
+	})
+
+	It("accepts a model with a fully populated legacy pipeline", func() {
+		cfg := &config.ModelConfig{
+			Name: "legacy",
+			Pipeline: config.Pipeline{
+				VAD:           "silero",
+				Transcription: "whisper",
+				LLM:           "llama",
+				TTS:           "piper",
+			},
+		}
+		_, _, ok := prepareRealtimeConfig(cfg)
+		Expect(ok).To(BeTrue())
+		Expect(cfg.Pipeline.LLM).To(Equal("llama"), "user-supplied pipeline slot must not be overwritten")
+	})
+
+	It("accepts a self-contained realtime_audio model and self-pipelines empty slots", func() {
+		cfg := &config.ModelConfig{
+			Name:          "lfm2.5-audio-realtime",
+			KnownUsecases: withUsecases(config.FLAG_REALTIME_AUDIO),
+		}
+		_, _, ok := prepareRealtimeConfig(cfg)
+		Expect(ok).To(BeTrue())
+		Expect(cfg.Pipeline.VAD).To(Equal("lfm2.5-audio-realtime"))
+		Expect(cfg.Pipeline.Transcription).To(Equal("lfm2.5-audio-realtime"))
+		Expect(cfg.Pipeline.LLM).To(Equal("lfm2.5-audio-realtime"))
+		Expect(cfg.Pipeline.TTS).To(Equal("lfm2.5-audio-realtime"))
+	})
+
+	It("preserves user-pinned pipeline slots on a realtime_audio model", func() {
+		// A user might want a dedicated silero-vad and let the realtime_audio
+		// model own only STT/LLM/TTS.
+		cfg := &config.ModelConfig{
+			Name:          "lfm-with-external-vad",
+			KnownUsecases: withUsecases(config.FLAG_REALTIME_AUDIO),
+			Pipeline: config.Pipeline{
+				VAD: "silero-vad",
+			},
+		}
+		_, _, ok := prepareRealtimeConfig(cfg)
+		Expect(ok).To(BeTrue())
+		Expect(cfg.Pipeline.VAD).To(Equal("silero-vad"))
+		Expect(cfg.Pipeline.Transcription).To(Equal("lfm-with-external-vad"))
+		Expect(cfg.Pipeline.LLM).To(Equal("lfm-with-external-vad"))
+		Expect(cfg.Pipeline.TTS).To(Equal("lfm-with-external-vad"))
+	})
+
+	It("accepts a model with at least one legacy pipeline slot set", func() {
+		// Pre-existing behaviour: the gate only rejected when ALL four slots
+		// were empty. Lock that in so the change doesn't tighten the gate.
+		cfg := &config.ModelConfig{
+			Name: "partial",
+			Pipeline: config.Pipeline{
+				LLM: "llama",
+			},
+		}
+		_, _, ok := prepareRealtimeConfig(cfg)
+		Expect(ok).To(BeTrue())
+	})
+})
+
+var _ = Describe("defaultMaxHistoryItems", func() {
+	It("caps realtime_audio sessions at 6", func() {
+		cfg := &config.ModelConfig{KnownUsecases: withUsecases(config.FLAG_REALTIME_AUDIO)}
+		Expect(defaultMaxHistoryItems(cfg)).To(Equal(6))
+	})
+	It("leaves legacy pipelines unlimited", func() {
+		cfg := &config.ModelConfig{Pipeline: config.Pipeline{LLM: "llama"}}
+		Expect(defaultMaxHistoryItems(cfg)).To(Equal(0))
+	})
+	It("tolerates nil", func() {
+		Expect(defaultMaxHistoryItems(nil)).To(Equal(0))
+	})
+})
+
+var _ = Describe("trimRealtimeItems", func() {
+	user := func(id string) *types.MessageItemUnion {
+		return &types.MessageItemUnion{User: &types.MessageItemUser{ID: id}}
+	}
+	assistant := func(id string) *types.MessageItemUnion {
+		return &types.MessageItemUnion{Assistant: &types.MessageItemAssistant{ID: id}}
+	}
+	fnCall := func(id, callID string) *types.MessageItemUnion {
+		return &types.MessageItemUnion{FunctionCall: &types.MessageItemFunctionCall{ID: id, CallID: callID}}
+	}
+	fnOut := func(id, callID string) *types.MessageItemUnion {
+		return &types.MessageItemUnion{FunctionCallOutput: &types.MessageItemFunctionCallOutput{ID: id, CallID: callID}}
+	}
+
+	It("returns the input unchanged when cap is zero", func() {
+		in := []*types.MessageItemUnion{user("u1"), assistant("a1")}
+		Expect(trimRealtimeItems(in, 0)).To(Equal(in))
+	})
+
+	It("returns the input unchanged when under the cap", func() {
+		in := []*types.MessageItemUnion{user("u1"), assistant("a1")}
+		Expect(trimRealtimeItems(in, 4)).To(Equal(in))
+	})
+
+	It("keeps the tail when over the cap", func() {
+		in := []*types.MessageItemUnion{user("u1"), assistant("a1"), user("u2"), assistant("a2"), user("u3")}
+		out := trimRealtimeItems(in, 3)
+		Expect(out).To(HaveLen(3))
+		Expect(out[0].User.ID).To(Equal("u2"))
+		Expect(out[2].User.ID).To(Equal("u3"))
+	})
+
+	It("pulls the cut left to keep a function_call paired with its output", func() {
+		// 0:user 1:fc 2:fc_out 3:assistant — cap=2 would otherwise start at
+		// index 2 (orphan fc_out). Helper must roll back to include 1.
+		in := []*types.MessageItemUnion{user("u1"), fnCall("fc1", "c1"), fnOut("fo1", "c1"), assistant("a1")}
+		out := trimRealtimeItems(in, 2)
+		// Expect at least the fc + fc_out + assistant (3 items, cap was 2)
+		// — the rollback prefers correctness over the cap.
+		Expect(len(out)).To(BeNumerically(">=", 3))
+		Expect(out[0].FunctionCall).NotTo(BeNil())
+		Expect(out[1].FunctionCallOutput).NotTo(BeNil())
+	})
+})

core/http/endpoints/openai/realtime_modality_test.go

@@ -0,0 +1,39 @@
+package openai
+
+import (
+	"github.com/mudler/LocalAI/core/http/endpoints/openai/types"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("resolveOutputModalities", func() {
+	It("defaults to audio when neither session nor response specify", func() {
+		got := resolveOutputModalities(nil, nil)
+		Expect(got).To(ConsistOf(types.ModalityAudio))
+	})
+
+	It("uses session modalities when response omits them", func() {
+		sess := []types.Modality{types.ModalityText}
+		got := resolveOutputModalities(sess, nil)
+		Expect(got).To(ConsistOf(types.ModalityText))
+	})
+
+	It("response modalities override session", func() {
+		sess := []types.Modality{types.ModalityAudio}
+		resp := []types.Modality{types.ModalityText}
+		got := resolveOutputModalities(sess, resp)
+		Expect(got).To(ConsistOf(types.ModalityText))
+	})
+
+	It("returns false from modalitiesContainAudio for text-only", func() {
+		Expect(modalitiesContainAudio([]types.Modality{types.ModalityText})).To(BeFalse())
+	})
+
+	It("returns true from modalitiesContainAudio for audio (default)", func() {
+		Expect(modalitiesContainAudio([]types.Modality{types.ModalityAudio})).To(BeTrue())
+	})
+
+	It("returns true when both audio and text are present", func() {
+		Expect(modalitiesContainAudio([]types.Modality{types.ModalityText, types.ModalityAudio})).To(BeTrue())
+	})
+})

core/http/endpoints/openai/realtime_webrtc.go

@@ -15,6 +15,10 @@ import (
 type RealtimeCallRequest struct {
 	SDP   string `json:"sdp"`
 	Model string `json:"model"`
+	// LocalAIAssistant opts the session into the in-process admin tool
+	// surface (same modality as the chat page's "Manage Mode"). Admin-only;
+	// the realtime entry point gates it the same way the chat handler does.
+	LocalAIAssistant bool `json:"localai_assistant,omitempty"`
 }
 
 // RealtimeCallResponse is the JSON response for POST /v1/realtime/calls.
@@ -165,9 +169,13 @@ func RealtimeCalls(application *application.Application) echo.HandlerFunc {
 
 		// Start the realtime session in a goroutine
 		evaluator := application.TemplatesEvaluator()
+		opts := RealtimeSessionOptions{
+			LocalAIAssistant: req.LocalAIAssistant,
+			IsAdmin:          isCurrentUserAdmin(c, application),
+		}
 		go func() {
 			defer transport.Close()
-			runRealtimeSession(application, transport, req.Model, evaluator)
+			runRealtimeSession(application, transport, req.Model, evaluator, opts)
 		}()
 
 		return c.JSON(http.StatusCreated, RealtimeCallResponse{

core/http/middleware/baseurl.go

@@ -6,20 +6,55 @@ import (
 	"github.com/labstack/echo/v4"
 )
 
+// BasePathPrefix returns the URL path prefix that the request was reached
+// under (e.g. "/myprefix/"). It always returns a value that starts and ends
+// with `/`, defaulting to "/" when the app is not behind a path prefix.
+//
+// It first looks at the path StripPathPrefix removed (when the proxy forwards
+// the prefix in the URL), then falls back to the X-Forwarded-Prefix header
+// (when the proxy strips the prefix before forwarding, e.g. Caddy's
+// handle_path).
+//
+// The header fallback is gated through SafeForwardedPrefix because the value
+// flows into the SPA HTML response (both <base href> and the path-absolute
+// asset URL rewrite in serveIndex). X-Forwarded-Prefix is attacker
+// controllable on misconfigured proxy chains; without that gate a value like
+// "//evil.com" turns the asset rewrite into a protocol-relative URL that
+// loads JS from a foreign origin.
+func BasePathPrefix(c echo.Context) string {
+	path := c.Path()
+	origPath := c.Request().URL.Path
+
+	if storedPath, ok := c.Get("_original_path").(string); ok && storedPath != "" {
+		origPath = storedPath
+	}
+
+	if path != origPath && strings.HasSuffix(origPath, path) && len(path) > 0 {
+		prefixLen := len(origPath) - len(path)
+		if prefixLen > 0 {
+			pathPrefix := origPath[:prefixLen]
+			if !strings.HasSuffix(pathPrefix, "/") {
+				pathPrefix += "/"
+			}
+			return pathPrefix
+		}
+	}
+
+	if validated, ok := SafeForwardedPrefix(c.Request().Header.Get("X-Forwarded-Prefix")); ok {
+		if !strings.HasSuffix(validated, "/") {
+			validated += "/"
+		}
+		return validated
+	}
+
+	return "/"
+}
+
 // BaseURL returns the base URL for the given HTTP request context.
 // It takes into account that the app may be exposed by a reverse-proxy under a different protocol, host and path.
 // The returned URL is guaranteed to end with `/`.
 // The method should be used in conjunction with the StripPathPrefix middleware.
 func BaseURL(c echo.Context) string {
-	path := c.Path()
-	origPath := c.Request().URL.Path
-
-	// Check if StripPathPrefix middleware stored the original path
-	if storedPath, ok := c.Get("_original_path").(string); ok && storedPath != "" {
-		origPath = storedPath
-	}
-
-	// Check X-Forwarded-Proto for scheme
 	scheme := "http"
 	if c.Request().Header.Get("X-Forwarded-Proto") == "https" {
 		scheme = "https"
@@ -27,22 +62,10 @@ func BaseURL(c echo.Context) string {
 		scheme = "https"
 	}
 
-	// Check X-Forwarded-Host for host
 	host := c.Request().Host
 	if forwardedHost := c.Request().Header.Get("X-Forwarded-Host"); forwardedHost != "" {
 		host = forwardedHost
 	}
 
-	if path != origPath && strings.HasSuffix(origPath, path) && len(path) > 0 {
-		prefixLen := len(origPath) - len(path)
-		if prefixLen > 0 && prefixLen <= len(origPath) {
-			pathPrefix := origPath[:prefixLen]
-			if !strings.HasSuffix(pathPrefix, "/") {
-				pathPrefix += "/"
-			}
-			return scheme + "://" + host + pathPrefix
-		}
-	}
-
-	return scheme + "://" + host + "/"
+	return scheme + "://" + host + BasePathPrefix(c)
 }

core/http/middleware/baseurl_test.go

@@ -55,4 +55,84 @@ var _ = Describe("BaseURL", func() {
 			Expect(actualURL).To(Equal("http://example.com/myprefix/"), "base URL")
 		})
 	})
+
+	// Caddy's handle_path (and similar reverse-proxy directives) strips the
+	// matched prefix before forwarding upstream, so LocalAI receives the
+	// already-stripped path together with X-Forwarded-Prefix. In that case
+	// StripPathPrefix never stores _original_path, but BaseURL must still
+	// honor the header so that <base href> and asset URLs include the prefix.
+	Context("with X-Forwarded-Prefix header but pre-stripped path", func() {
+		It("should return base URL with prefix from header", func() {
+			app := echo.New()
+			actualURL := ""
+
+			routePath := "/app"
+			app.GET(routePath, func(c echo.Context) error {
+				actualURL = BaseURL(c)
+				return nil
+			})
+
+			req := httptest.NewRequest("GET", "/app", nil)
+			req.Header.Set("X-Forwarded-Prefix", "/localai")
+			rec := httptest.NewRecorder()
+			app.ServeHTTP(rec, req)
+
+			Expect(rec.Code).To(Equal(200), "response status code")
+			Expect(actualURL).To(Equal("http://example.com/localai/"), "base URL")
+		})
+
+		It("should normalize a prefix that already ends with a slash", func() {
+			app := echo.New()
+			actualURL := ""
+
+			routePath := "/app"
+			app.GET(routePath, func(c echo.Context) error {
+				actualURL = BaseURL(c)
+				return nil
+			})
+
+			req := httptest.NewRequest("GET", "/app", nil)
+			req.Header.Set("X-Forwarded-Prefix", "/localai/")
+			rec := httptest.NewRecorder()
+			app.ServeHTTP(rec, req)
+
+			Expect(rec.Code).To(Equal(200), "response status code")
+			Expect(actualURL).To(Equal("http://example.com/localai/"), "base URL")
+		})
+	})
+
+	// X-Forwarded-Prefix is attacker controllable on misconfigured proxy
+	// chains, and the value flows into the SPA HTML response (<base href>
+	// and asset URLs). BasePathPrefix must gate the header through
+	// SafeForwardedPrefix so values that turn the prefix into an open
+	// redirect or a protocol-relative URL are ignored and the base falls
+	// back to "/".
+	Context("with unsafe X-Forwarded-Prefix header", func() {
+		DescribeTable("falls back to / when the header is unsafe",
+			func(header string) {
+				app := echo.New()
+				actualURL := ""
+
+				app.GET("/app", func(c echo.Context) error {
+					actualURL = BaseURL(c)
+					return nil
+				})
+
+				req := httptest.NewRequest("GET", "/app", nil)
+				req.Header.Set("X-Forwarded-Prefix", header)
+				rec := httptest.NewRecorder()
+				app.ServeHTTP(rec, req)
+
+				Expect(rec.Code).To(Equal(200), "response status code")
+				Expect(actualURL).To(Equal("http://example.com/"), "base URL")
+			},
+			Entry("protocol-relative URL", "//evil.com"),
+			Entry("protocol-relative URL with path", "//evil.com/assets"),
+			Entry("backslash path", `/foo\bar`),
+			Entry("embedded NUL", "/foo\x00bar"),
+			Entry("CR injection", "/foo\rbar"),
+			Entry("LF injection", "/foo\nbar"),
+			Entry("missing leading slash", "evil"),
+		)
+	})
 })

core/http/middleware/request.go

@@ -14,7 +14,6 @@ import (
 	"github.com/mudler/LocalAI/core/schema"
 	"github.com/mudler/LocalAI/core/services/galleryop"
 	"github.com/mudler/LocalAI/core/templates"
-	"github.com/mudler/LocalAI/pkg/functions"
 	"github.com/mudler/LocalAI/pkg/model"
 	"github.com/mudler/LocalAI/pkg/utils"
 	"github.com/mudler/xlog"
@@ -241,6 +240,28 @@ func (re *RequestExtractor) SetOpenAIRequest(c echo.Context) error {
 	return nil
 }
 
+// extractToolChoiceFunctionName parses a tool_choice map and returns the
+// specific function name. Accepts both the OpenAI-spec nested shape
+// ({type:function, function:{name:...}}) and the legacy/Anthropic-compat
+// flat shape ({type:function, name:...}); the nested form wins when both
+// are present. Returns "" for malformed input or when the shape names a
+// mode rather than a specific tool.
+func extractToolChoiceFunctionName(m map[string]any) string {
+	tcType, ok := m["type"].(string)
+	if !ok || tcType != "function" {
+		return ""
+	}
+	if fn, ok := m["function"].(map[string]any); ok {
+		if n, ok := fn["name"].(string); ok && n != "" {
+			return n
+		}
+	}
+	if n, ok := m["name"].(string); ok {
+		return n
+	}
+	return ""
+}
+
 func mergeOpenAIRequestAndModelConfig(config *config.ModelConfig, input *schema.OpenAIRequest) error {
 	if input.Echo {
 		config.Echo = input.Echo
@@ -320,17 +341,55 @@ func mergeOpenAIRequestAndModelConfig(config *config.ModelConfig, input *schema.
 	}
 
 	if input.ToolsChoice != nil {
-		var toolChoice functions.Tool
-
+		// OpenAI tool_choice has three valid shapes plus one tolerated
+		// non-spec form seen in the wild:
+		//
+		//   1. string mode:    "auto" | "none" | "required"
+		//   2. specific tool:  {"type":"function", "function":{"name":"..."}}  (current spec)
+		//   3. legacy:         {"type":"function", "name":"..."}                (older / Anthropic-compat)
+		//   4. double-encoded: "{\"type\":\"function\", ...}"                   (some clients serialize the object)
+		//
+		// The pre-#9559 code unmarshalled the string case through
+		// json.Unmarshal([]byte(content), &functions.Tool{}), which:
+		//   - failed for plain string modes (so "required" / "none" were
+		//     silently ignored and tools stayed enabled regardless), but
+		//   - happened to handle shape 4 by accident.
+		// It also could not parse shape 3 because functions.Tool has no
+		// flat top-level Name field.
+		//
+		// Mirror the parsing pattern from MergeOpenResponsesConfig (#9509),
+		// route results through the existing input.FunctionCall string/map
+		// dispatch downstream (see the switch on input.FunctionCall in this
+		// same function), and preserve the shape-4 fallback so non-spec
+		// clients don't silently break. Tracked in #9508; sibling fix in #9526.
 		switch content := input.ToolsChoice.(type) {
 		case string:
-			_ = json.Unmarshal([]byte(content), &toolChoice)
+			// "auto" is the default and needs no override. "none" and "required"
+			// both reach SetFunctionCallString via the input.FunctionCall string
+			// branch below; ShouldUseFunctions() then returns false for "none"
+			// (tools disabled) and true for "required" (mode engaged).
+			//
+			// If the string looks like a JSON object, try shape 4 first: parse
+			// it as a tool_choice map and use the resulting name. Falling back
+			// to mode-string handling when the parse yields no usable name keeps
+			// genuinely-malformed input from accidentally engaging a mode.
+			if content == "" || content == "auto" {
+				break
+			}
+			if strings.HasPrefix(strings.TrimSpace(content), "{") {
+				var nested map[string]any
+				if err := json.Unmarshal([]byte(content), &nested); err == nil {
+					if name := extractToolChoiceFunctionName(nested); name != "" {
+						input.FunctionCall = map[string]any{"name": name}
+						break
+					}
+				}
+			}
+			input.FunctionCall = content
 		case map[string]any:
-			dat, _ := json.Marshal(content)
-			_ = json.Unmarshal(dat, &toolChoice)
-		}
-		input.FunctionCall = map[string]any{
-			"name": toolChoice.Function.Name,
+			if name := extractToolChoiceFunctionName(content); name != "" {
+				input.FunctionCall = map[string]any{"name": name}
+			}
 		}
 	}
 

core/http/middleware/request_test.go

@@ -306,3 +306,248 @@ var _ = Describe("MergeOpenResponsesConfig tool_choice parsing", func() {
 		})
 	})
 })
+
+// ---------------------------------------------------------------------------
+// SetModelAndConfig + SetOpenAIRequest - /v1/chat/completions tool_choice parsing
+// ---------------------------------------------------------------------------
+//
+// Parallel to the MergeOpenResponsesConfig specs above, but for the chat
+// completions path. The parsing block lives in mergeOpenAIRequestAndModelConfig
+// (called from SetOpenAIRequest), so these tests drive the full middleware
+// chain the way the production /v1/chat/completions route does.
+//
+// What we assert per shape:
+//   - "required"                                  -> ShouldUseFunctions=true,  no specific name
+//   - "none"                                      -> ShouldUseFunctions=false (tools disabled)
+//   - "auto"                                      -> ShouldUseFunctions=true,  no specific name
+//   - {type:function, function:{name:"X"}} (spec) -> ShouldCallSpecificFunction=true, FunctionToCall="X"
+//   - {type:function, name:"X"}        (legacy)   -> ShouldCallSpecificFunction=true, FunctionToCall="X"
+//   - nested+flat both present                    -> nested wins
+//   - malformed (no type / no name)               -> no-op
+var _ = Describe("SetModelAndConfig tool_choice parsing (chat completions)", func() {
+	var (
+		app            *echo.Echo
+		modelDir       string
+		capturedConfig *config.ModelConfig
+	)
+
+	BeforeEach(func() {
+		var err error
+		modelDir, err = os.MkdirTemp("", "localai-test-models-*")
+		Expect(err).ToNot(HaveOccurred())
+
+		cfgContent := []byte("name: test-model\nbackend: llama-cpp\n")
+		Expect(os.WriteFile(filepath.Join(modelDir, "test-model.yaml"), cfgContent, 0644)).To(Succeed())
+
+		ss := &system.SystemState{
+			Model: system.Model{ModelsPath: modelDir},
+		}
+		appConfig := config.NewApplicationConfig()
+		appConfig.SystemState = ss
+
+		mcl := config.NewModelConfigLoader(modelDir)
+		ml := model.NewModelLoader(ss)
+		re := NewRequestExtractor(mcl, ml, appConfig)
+
+		capturedConfig = nil
+		app = echo.New()
+		app.POST("/v1/chat/completions",
+			func(c echo.Context) error {
+				if cfg, ok := c.Get(CONTEXT_LOCALS_KEY_MODEL_CONFIG).(*config.ModelConfig); ok {
+					capturedConfig = cfg
+				}
+				return c.String(http.StatusOK, "ok")
+			},
+			re.SetModelAndConfig(func() schema.LocalAIRequest { return new(schema.OpenAIRequest) }),
+			func(next echo.HandlerFunc) echo.HandlerFunc {
+				return func(c echo.Context) error {
+					if err := re.SetOpenAIRequest(c); err != nil {
+						return err
+					}
+					return next(c)
+				}
+			},
+		)
+	})
+
+	AfterEach(func() {
+		_ = os.RemoveAll(modelDir)
+	})
+
+	// chatReq wraps a tool_choice JSON fragment in a minimal valid chat-completions
+	// payload. The tools array is non-empty so downstream code paths that gate on
+	// len(input.Functions) see something to work with.
+	chatReq := func(toolChoiceJSON string) string {
+		return `{"model":"test-model",` +
+			`"messages":[{"role":"user","content":"hi"}],` +
+			`"tools":[{"type":"function","function":{"name":"get_weather"}}],` +
+			`"tool_choice":` + toolChoiceJSON + `}`
+	}
+
+	Context("string tool_choice", func() {
+		It("engages mode for tool_choice=\"required\"", func() {
+			rec := postJSON(app, "/v1/chat/completions", chatReq(`"required"`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+			Expect(capturedConfig.ShouldUseFunctions()).To(BeTrue())
+		})
+
+		It("disables tools for tool_choice=\"none\"", func() {
+			// Before #9559 this was a silent no-op (json.Unmarshal of "none"
+			// into functions.Tool failed); now "none" is honored per OpenAI spec.
+			rec := postJSON(app, "/v1/chat/completions", chatReq(`"none"`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldUseFunctions()).To(BeFalse())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+		})
+
+		It("leaves config untouched for tool_choice=\"auto\"", func() {
+			rec := postJSON(app, "/v1/chat/completions", chatReq(`"auto"`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+			// "auto" is the default: tools available, model decides.
+			Expect(capturedConfig.ShouldUseFunctions()).To(BeTrue())
+			Expect(capturedConfig.FunctionToCall()).To(Equal(""))
+		})
+	})
+
+	Context("specific-function tool_choice (OpenAI spec shape)", func() {
+		It("parses {type:function, function:{name:...}} and forces the named function", func() {
+			rec := postJSON(app, "/v1/chat/completions",
+				chatReq(`{"type":"function","function":{"name":"get_weather"}}`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			// Key invariant: a correctly-formed OpenAI tool_choice must engage
+			// grammar-based forcing via SetFunctionCallNameString.
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeTrue())
+			Expect(capturedConfig.FunctionToCall()).To(Equal("get_weather"))
+		})
+
+		It("prefers the nested function.name over a stray top-level name", func() {
+			rec := postJSON(app, "/v1/chat/completions",
+				chatReq(`{"type":"function","function":{"name":"correct_name"},"name":"legacy_name"}`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.FunctionToCall()).To(Equal("correct_name"))
+		})
+	})
+
+	Context("specific-function tool_choice (legacy Anthropic-compat shape)", func() {
+		It("parses {type:function, name:...} and forces the named function", func() {
+			rec := postJSON(app, "/v1/chat/completions",
+				chatReq(`{"type":"function","name":"get_weather"}`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeTrue())
+			Expect(capturedConfig.FunctionToCall()).To(Equal("get_weather"))
+		})
+	})
+
+	// Some non-spec clients send the object form serialized as a JSON string.
+	// The pre-#9559 code accepted that by accident; this Context locks in
+	// continued tolerance so those clients do not silently regress.
+	Context("double-encoded tool_choice (JSON string of an object, non-spec)", func() {
+		It("parses a serialized OpenAI-spec nested object", func() {
+			// tool_choice value is itself a JSON-encoded string containing the
+			// object form. Use json.Marshal of the inner blob so the escapes
+			// are correct regardless of the test reader.
+			inner := `{"type":"function","function":{"name":"get_weather"}}`
+			encoded, err := json.Marshal(inner)
+			Expect(err).ToNot(HaveOccurred())
+			rec := postJSON(app, "/v1/chat/completions", chatReq(string(encoded)))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeTrue())
+			Expect(capturedConfig.FunctionToCall()).To(Equal("get_weather"))
+		})
+
+		It("parses a serialized legacy/Anthropic flat object", func() {
+			inner := `{"type":"function","name":"get_weather"}`
+			encoded, err := json.Marshal(inner)
+			Expect(err).ToNot(HaveOccurred())
+			rec := postJSON(app, "/v1/chat/completions", chatReq(string(encoded)))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeTrue())
+			Expect(capturedConfig.FunctionToCall()).To(Equal("get_weather"))
+		})
+
+		It("falls back to mode-string handling when the JSON string parses but has no usable name", func() {
+			// A JSON-string that decodes to a map without a function name
+			// should not engage specific-function forcing. We expect it to
+			// fall through to the mode-string path; the resulting mode is
+			// the raw blob (nonsense), but ShouldCallSpecificFunction stays
+			// false - the invariant that matters.
+			inner := `{"type":"function"}`
+			encoded, err := json.Marshal(inner)
+			Expect(err).ToNot(HaveOccurred())
+			rec := postJSON(app, "/v1/chat/completions", chatReq(string(encoded)))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+		})
+	})
+
+	Context("malformed tool_choice", func() {
+		It("is a no-op when type is missing", func() {
+			rec := postJSON(app, "/v1/chat/completions",
+				chatReq(`{"function":{"name":"get_weather"}}`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+		})
+
+		It("is a no-op when type is not \"function\"", func() {
+			rec := postJSON(app, "/v1/chat/completions",
+				chatReq(`{"type":"object","function":{"name":"get_weather"}}`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+		})
+
+		It("is a no-op when name is missing from both shapes", func() {
+			rec := postJSON(app, "/v1/chat/completions",
+				chatReq(`{"type":"function","function":{}}`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+			Expect(capturedConfig.FunctionToCall()).To(Equal(""))
+		})
+
+		It("is a no-op when name is empty string", func() {
+			rec := postJSON(app, "/v1/chat/completions",
+				chatReq(`{"type":"function","function":{"name":""}}`))
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+		})
+	})
+
+	Context("nil tool_choice", func() {
+		It("is a no-op", func() {
+			rec := postJSON(app, "/v1/chat/completions",
+				`{"model":"test-model","messages":[{"role":"user","content":"hi"}]}`)
+
+			Expect(rec.Code).To(Equal(http.StatusOK))
+			Expect(capturedConfig).ToNot(BeNil())
+			Expect(capturedConfig.ShouldCallSpecificFunction()).To(BeFalse())
+			Expect(capturedConfig.FunctionToCall()).To(Equal(""))
+		})
+	})
+})

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

@@ -24,6 +24,7 @@
     "diarization": "Diarization",
     "soundGen": "Sound",
     "audioTransform": "Audio FX",
+    "realtimeAudio": "Realtime Audio",
     "embedding": "Embeddings",
     "rerank": "Rerank",
     "detection": "Detection",

core/http/react-ui/src/hooks/useChat.js

@@ -255,7 +255,10 @@ export function useChat(initialModel = '') {
     )
     messages.push(...historyForApi, { role: 'user', content: messageContent })
 
-    const requestBody = { model, messages, stream: true }
+    // include_usage tells LocalAI to emit a trailing chunk with token totals;
+    // without it the spec-compliant server drops `usage` from the stream and
+    // the token-count badge would never populate.
+    const requestBody = { model, messages, stream: true, stream_options: { include_usage: true } }
     if (temperature !== null && temperature !== undefined) requestBody.temperature = temperature
     if (topP !== null && topP !== undefined) requestBody.top_p = topP
     if (topK !== null && topK !== undefined) requestBody.top_k = topK

core/http/react-ui/src/pages/FineTune.jsx

@@ -732,6 +732,9 @@ export default function FineTune() {
   const [seed, setSeed] = useState(0)
   const [mixedPrecision, setMixedPrecision] = useState('')
   const [extraOptions, setExtraOptions] = useState([])
+  // liquid-audio specific knobs (folded into extra_options on submit)
+  const [liquidAudioVoice, setLiquidAudioVoice] = useState('')
+  const [liquidAudioValDataset, setLiquidAudioValDataset] = useState('')
   const [hfToken, setHfToken] = useState('')
   const [showAdvanced, setShowAdvanced] = useState(false)
   const [resumeFromCheckpoint, setResumeFromCheckpoint] = useState('')
@@ -801,6 +804,12 @@ export default function FineTune() {
       for (const { key, value } of extraOptions) {
         if (key.trim()) extra[key.trim()] = value
       }
+      // Fold liquid-audio specific fields into extra_options. The Python
+      // backend reads `voice` and `val_dataset` directly from there.
+      if (backend === 'liquid-audio') {
+        if (liquidAudioVoice) extra.voice = liquidAudioVoice
+        if (liquidAudioValDataset.trim()) extra.val_dataset = liquidAudioValDataset.trim()
+      }
 
       const isAdapter = ['lora', 'loha', 'lokr'].includes(trainingType)
 
@@ -872,6 +881,10 @@ export default function FineTune() {
     for (const { key, value } of extraOptions) {
       if (key.trim()) extra[key.trim()] = value
     }
+    if (backend === 'liquid-audio') {
+      if (liquidAudioVoice) extra.voice = liquidAudioVoice
+      if (liquidAudioValDataset.trim()) extra.val_dataset = liquidAudioValDataset.trim()
+    }
     return {
       model,
       backend,
@@ -965,10 +978,15 @@ export default function FineTune() {
       setSaveTotalLimit(Number(config.extra_options.save_total_limit))
     }
 
+    // Restore liquid-audio specific extras (also filtered out of the
+    // freeform list below).
+    if (config.extra_options?.voice != null) setLiquidAudioVoice(String(config.extra_options.voice))
+    if (config.extra_options?.val_dataset != null) setLiquidAudioValDataset(String(config.extra_options.val_dataset))
+
     // Convert extra_options object to [{key, value}] entries, filtering out handled keys
     if (config.extra_options && typeof config.extra_options === 'object') {
       const entries = Object.entries(config.extra_options)
-        .filter(([k]) => !['max_seq_length', 'save_total_limit', 'hf_token', 'eval_strategy', 'eval_steps', 'eval_split', 'eval_dataset_source', 'eval_split_ratio'].includes(k))
+        .filter(([k]) => !['max_seq_length', 'save_total_limit', 'hf_token', 'eval_strategy', 'eval_steps', 'eval_split', 'eval_dataset_source', 'eval_split_ratio', 'voice', 'val_dataset'].includes(k))
         .map(([key, value]) => ({ key, value: String(value) }))
       setExtraOptions(entries)
     }
@@ -1458,6 +1476,31 @@ export default function FineTune() {
                   </div>
                 )}
 
+                {backend === 'liquid-audio' && (
+                  <div style={{ marginBottom: 'var(--spacing-md)' }}>
+                    <label className="form-label">Liquid Audio</label>
+                    <div style={{ fontSize: '0.8125rem', color: 'var(--color-text-muted)', marginBottom: 'var(--spacing-sm)' }}>
+                      Dataset must be preprocessed by <code>LFM2AudioChatMapper</code> (a directory of LFM2DataLoader-ready arrow files). See <code>liquid_audio/examples/preprocess_jenny_tts.py</code> for the conversion recipe.
+                    </div>
+                    <div style={{ display: 'grid', gridTemplateColumns: 'repeat(auto-fit, minmax(220px, 1fr))', gap: 'var(--spacing-sm)' }}>
+                      <div>
+                        <label className="form-label">TTS Voice (optional)</label>
+                        <select value={liquidAudioVoice} onChange={e => setLiquidAudioVoice(e.target.value)} className="input">
+                          <option value="">— inherit from system prompt —</option>
+                          <option value="us_male">us_male</option>
+                          <option value="us_female">us_female</option>
+                          <option value="uk_male">uk_male</option>
+                          <option value="uk_female">uk_female</option>
+                        </select>
+                      </div>
+                      <div>
+                        <label className="form-label">Validation Dataset (path)</label>
+                        <input type="text" value={liquidAudioValDataset} onChange={e => setLiquidAudioValDataset(e.target.value)} placeholder="e.g. /data/jenny_tts/val" className="input" />
+                      </div>
+                    </div>
+                  </div>
+                )}
+
                 <div>
                   <label className="form-label">Extra Options (backend-specific key-value pairs)</label>
                   <KeyValueEditor entries={extraOptions} onChange={setExtraOptions} />

core/http/react-ui/src/pages/Models.jsx

@@ -28,6 +28,7 @@ const FILTERS = [
   { key: 'diarization', labelKey: 'filters.diarization', icon: 'fa-users' },
   { key: 'sound_generation', labelKey: 'filters.soundGen', icon: 'fa-music' },
   { key: 'audio_transform', labelKey: 'filters.audioTransform', icon: 'fa-sliders' },
+  { key: 'realtime_audio', labelKey: 'filters.realtimeAudio', icon: 'fa-tower-broadcast' },
   { key: 'embeddings', labelKey: 'filters.embedding', icon: 'fa-vector-square' },
   { key: 'rerank', labelKey: 'filters.rerank', icon: 'fa-sort' },
   { key: 'detection', labelKey: 'filters.detection', icon: 'fa-bullseye' },

core/http/react-ui/src/pages/Talk.jsx

@@ -2,6 +2,10 @@ import { useState, useRef, useEffect, useCallback, useMemo } from 'react'
 import { useOutletContext, useNavigate } from 'react-router-dom'
 import { realtimeApi } from '../utils/api'
 import ModelSelector from '../components/ModelSelector'
+import ClientMCPDropdown from '../components/ClientMCPDropdown'
+import { useMCPClient } from '../hooks/useMCPClient'
+import { loadClientMCPServers } from '../utils/mcpClientStorage'
+import { useAuth } from '../context/AuthContext'
 
 const STATUS_STYLES = {
   disconnected: { icon: 'fa-solid fa-circle', color: 'var(--color-text-secondary)', bg: 'transparent' },
@@ -40,6 +44,27 @@ export default function Talk() {
   const [voiceEdited, setVoiceEdited] = useState(false)
   const [language, setLanguage] = useState('')
 
+  // Client MCP — mirrors the chat page's wiring (useMCPClient + ClientMCPDropdown).
+  // Talk has a single ephemeral session, so the active server set lives in component
+  // state rather than per-chat config.
+  const [clientMCPServers, setClientMCPServers] = useState(() => loadClientMCPServers())
+  const [activeMCPIds, setActiveMCPIds] = useState([])
+  const {
+    connect: mcpConnect,
+    disconnect: mcpDisconnect,
+    getToolsForLLM,
+    isClientTool,
+    executeTool,
+    connectionStatuses,
+    getConnectedTools,
+  } = useMCPClient()
+
+  // LocalAI Assistant ("Manage Mode") — mirrors the chat-page toggle.
+  // Admin-only; the realtime endpoint enforces the gate too. When on, the
+  // backend mounts the in-process MCP admin tool surface for this session.
+  const { isAdmin } = useAuth()
+  const [manageMode, setManageMode] = useState(false)
+
   // Diagnostics
   const [diagVisible, setDiagVisible] = useState(false)
 
@@ -75,7 +100,7 @@ export default function Talk() {
           if (!voiceEdited) setVoice(models[0].voice || '')
         }
       })
-      .catch(err => addToast(`Failed to load pipeline models: ${err.message}`, 'error', 5000, { link: { href: '/app/traces?tab=backend', text: 'View traces' } }))
+      .catch(err => addToast(`Failed to load realtime models: ${err.message}`, 'error', 5000, { link: { href: '/app/traces?tab=backend', text: 'View traces' } }))
       .finally(() => setModelsLoading(false))
   }, [])
 
@@ -84,6 +109,32 @@ export default function Talk() {
     transcriptEndRef.current?.scrollIntoView({ behavior: 'smooth' })
   }, [transcript])
 
+  // Mirror Chat.jsx: connect / disconnect client MCP servers as the user toggles them.
+  useEffect(() => {
+    const activeSet = new Set(activeMCPIds)
+    for (const server of clientMCPServers) {
+      const status = connectionStatuses[server.id]?.status
+      if (activeSet.has(server.id) && status !== 'connected' && status !== 'connecting') {
+        mcpConnect(server)
+      } else if (!activeSet.has(server.id) && (status === 'connected' || status === 'connecting')) {
+        mcpDisconnect(server.id)
+      }
+    }
+  }, [activeMCPIds.join(','), clientMCPServers, connectionStatuses, mcpConnect, mcpDisconnect])
+
+  const handleClientMCPToggle = useCallback((serverId) => {
+    setActiveMCPIds(prev => prev.includes(serverId) ? prev.filter(s => s !== serverId) : [...prev, serverId])
+  }, [])
+  const handleClientMCPServerAdded = useCallback((server) => {
+    setClientMCPServers(loadClientMCPServers())
+    setActiveMCPIds(prev => prev.includes(server.id) ? prev : [...prev, server.id])
+  }, [])
+  const handleClientMCPServerRemoved = useCallback(async (id) => {
+    await mcpDisconnect(id)
+    setClientMCPServers(loadClientMCPServers())
+    setActiveMCPIds(prev => prev.filter(s => s !== id))
+  }, [mcpDisconnect])
+
   const selectedModelInfo = pipelineModels.find(m => m.name === selectedModel)
 
   // ── Status helper ──
@@ -96,7 +147,9 @@ export default function Talk() {
   const sendSessionUpdate = useCallback(() => {
     const dc = dcRef.current
     if (!dc || dc.readyState !== 'open') return
-    if (!instructions.trim() && !voice.trim() && !language.trim()) return
+
+    const tools = getToolsForLLM()
+    if (!instructions.trim() && !voice.trim() && !language.trim() && tools.length === 0) return
 
     const session = {}
     if (instructions.trim()) session.instructions = instructions.trim()
@@ -105,9 +158,57 @@ export default function Talk() {
       if (voice.trim()) session.audio.output = { voice: voice.trim() }
       if (language.trim()) session.audio.input = { transcription: { language: language.trim() } }
     }
+    // Pass MCP-server-advertised tools straight through. Server-side they
+    // get rendered into the model's prompt via the function:/argument_regex
+    // pair on the model config (gallery/lfm.yaml for LFM2.5-Audio).
+    if (tools.length > 0) session.tools = tools
 
     dc.send(JSON.stringify({ type: 'session.update', session }))
-  }, [instructions, voice, language])
+  }, [instructions, voice, language, getToolsForLLM])
+
+  // Re-send session.update whenever the tool set changes mid-session so the
+  // model sees newly-toggled MCP servers without a reconnect.
+  useEffect(() => {
+    if (isConnected) sendSessionUpdate()
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [activeMCPIds.join(',')])
+
+  // ── Function-call dispatcher ──
+  // Mirrors the chat-page agentic loop: collect args from the model's
+  // function_call_arguments.done event, hand them to the MCP client's
+  // executeTool, then echo the result back via conversation.item.create +
+  // response.create so the model can complete its turn with the tool output.
+  const handleFunctionCall = useCallback(async (event) => {
+    const dc = dcRef.current
+    if (!dc || dc.readyState !== 'open') return
+    const { call_id: callId, name, arguments: argsJson } = event
+    if (!callId || !name) return
+    if (!isClientTool(name)) {
+      // No MCP server advertises this tool — let the model know so it can
+      // recover instead of hanging.
+      dc.send(JSON.stringify({
+        type: 'conversation.item.create',
+        item: { type: 'function_call_output', call_id: callId, output: `Error: unknown tool "${name}"` },
+      }))
+      dc.send(JSON.stringify({ type: 'response.create' }))
+      return
+    }
+    updateStatus('thinking', `Running tool ${name}...`)
+    try {
+      const result = await executeTool(name, argsJson)
+      dc.send(JSON.stringify({
+        type: 'conversation.item.create',
+        item: { type: 'function_call_output', call_id: callId, output: typeof result === 'string' ? result : JSON.stringify(result) },
+      }))
+      dc.send(JSON.stringify({ type: 'response.create' }))
+    } catch (err) {
+      dc.send(JSON.stringify({
+        type: 'conversation.item.create',
+        item: { type: 'function_call_output', call_id: callId, output: `Error: ${err?.message || err}` },
+      }))
+      dc.send(JSON.stringify({ type: 'response.create' }))
+    }
+  }, [executeTool, isClientTool, updateStatus])
 
   // ── Server event handler ──
   const handleServerEvent = useCallback((event) => {
@@ -163,6 +264,32 @@ export default function Talk() {
       case 'response.output_audio.delta':
         updateStatus('speaking', 'Speaking...')
         break
+      case 'response.output_item.done': {
+        // Server-executed tools (Manage Mode) surface as output items —
+        // FunctionCall when the model invokes a tool, FunctionCallOutput
+        // once the server has run it. Render both on `done` so we get
+        // each transcript entry exactly once.
+        const item = event.item
+        if (!item) break
+        if (item.FunctionCall) {
+          setTranscript(prev => [...prev, {
+            role: 'tool_call',
+            text: `${item.FunctionCall.name}(${item.FunctionCall.arguments || ''})`,
+          }])
+        } else if (item.FunctionCallOutput) {
+          let preview = item.FunctionCallOutput.output || ''
+          // Pretty-print JSON for readability; fall back to raw string.
+          try { preview = JSON.stringify(JSON.parse(preview), null, 2) } catch (_) { /* keep raw */ }
+          setTranscript(prev => [...prev, { role: 'tool_result', text: preview }])
+          streamingRef.current = null  // tool result ends the current assistant text run
+        }
+        break
+      }
+      case 'response.function_call_arguments.done':
+        // Don't await — keep the event loop free; handleFunctionCall sends
+        // conversation.item.create + response.create when it's done.
+        handleFunctionCall(event)
+        break
       case 'response.done':
         updateStatus('listening', 'Listening...')
         break
@@ -171,12 +298,12 @@ export default function Talk() {
         updateStatus('error', 'Error: ' + (event.error?.message || 'Unknown error'))
         break
     }
-  }, [sendSessionUpdate, updateStatus])
+  }, [sendSessionUpdate, updateStatus, handleFunctionCall])
 
   // ── Connect ──
   const connect = useCallback(async () => {
     if (!selectedModel) {
-      addToast('Please select a pipeline model first.', 'warning')
+      addToast('Please select a realtime model first.', 'warning')
       return
     }
     if (!navigator.mediaDevices?.getUserMedia) {
@@ -237,6 +364,7 @@ export default function Talk() {
       const data = await realtimeApi.call({
         sdp: pc.localDescription.sdp,
         model: selectedModel,
+        localai_assistant: manageMode,
       })
 
       await pc.setRemoteDescription({ type: 'answer', sdp: data.sdp })
@@ -245,7 +373,7 @@ export default function Talk() {
       updateStatus('error', 'Connection failed: ' + err.message)
       disconnect()
     }
-  }, [selectedModel, diagVisible, handleServerEvent, updateStatus, addToast])
+  }, [selectedModel, manageMode, diagVisible, handleServerEvent, updateStatus, addToast])
 
   // ── Disconnect ──
   const disconnect = useCallback(() => {
@@ -508,8 +636,58 @@ export default function Talk() {
             </button>
           </div>
 
+          {/* Tools (client-side MCP servers, mirroring the chat page) */}
+          <div style={{ marginBottom: 'var(--spacing-md)' }}>
+            <label className="form-label" style={{ fontSize: '0.8125rem' }}>
+              <i className="fas fa-screwdriver-wrench" style={{ color: 'var(--color-primary)', marginRight: 4 }} /> Tools
+            </label>
+            <ClientMCPDropdown
+              activeServerIds={activeMCPIds}
+              onToggleServer={handleClientMCPToggle}
+              onServerAdded={handleClientMCPServerAdded}
+              onServerRemoved={handleClientMCPServerRemoved}
+              connectionStatuses={connectionStatuses}
+              getConnectedTools={getConnectedTools}
+            />
+            {isAdmin && (
+              <label style={{
+                display: 'flex', alignItems: 'center', gap: 'var(--spacing-xs)',
+                marginTop: 'var(--spacing-xs)', fontSize: '0.8125rem',
+                cursor: isConnected ? 'not-allowed' : 'pointer',
+                color: isConnected ? 'var(--color-text-secondary)' : 'var(--color-text)',
+              }}>
+                <input
+                  type="checkbox"
+                  checked={manageMode}
+                  disabled={isConnected}
+                  onChange={(e) => setManageMode(e.target.checked)}
+                />
+                <i className="fas fa-user-shield" style={{ color: 'var(--color-primary)' }} />
+                Manage Mode
+                <span style={{ color: 'var(--color-text-secondary)', fontSize: '0.75rem' }}>
+                  — let the model query LocalAI (models, backends, system info)
+                </span>
+              </label>
+            )}
+          </div>
+
           {/* Pipeline details */}
-          {selectedModelInfo && (
+          {selectedModelInfo && selectedModelInfo.self_contained && (
+            <div style={{
+              background: 'var(--color-bg-secondary)', borderRadius: 'var(--radius-sm)',
+              padding: 'var(--spacing-xs) var(--spacing-sm)', border: '1px solid var(--color-border)',
+              marginBottom: 'var(--spacing-xs)', fontSize: '0.75rem',
+              display: 'flex', alignItems: 'center', gap: 'var(--spacing-xs)',
+            }}>
+              <i className="fas fa-tower-broadcast" style={{ color: 'var(--color-primary)' }} />
+              <span style={{ color: 'var(--color-text-secondary)' }}>Self-contained any-to-any —</span>
+              <span style={{ fontFamily: 'var(--font-mono)', overflow: 'hidden', textOverflow: 'ellipsis', whiteSpace: 'nowrap' }}>
+                {selectedModelInfo.name}
+              </span>
+              <span style={{ color: 'var(--color-text-secondary)', marginLeft: 'auto' }}>handles VAD · STT · LLM · TTS</span>
+            </div>
+          )}
+          {selectedModelInfo && !selectedModelInfo.self_contained && (
             <div style={{
               display: 'grid', gridTemplateColumns: 'repeat(4, 1fr)', gap: 'var(--spacing-xs)',
               marginBottom: 'var(--spacing-xs)', fontSize: '0.75rem',
@@ -533,7 +711,8 @@ export default function Talk() {
           {selectedModelInfo && !isConnected && (
             <div style={{ marginBottom: 'var(--spacing-md)' }}>
               <button className="btn btn-secondary btn-sm" onClick={() => navigate(`/app/model-editor/${encodeURIComponent(selectedModel)}`)}>
-                <i className="fas fa-pen-to-square" style={{ marginRight: 'var(--spacing-xs)' }} /> Edit Pipeline
+                <i className="fas fa-pen-to-square" style={{ marginRight: 'var(--spacing-xs)' }} />
+                {selectedModelInfo.self_contained ? ' Edit Model Config' : ' Edit Pipeline'}
               </button>
             </div>
           )}
@@ -600,16 +779,28 @@ export default function Talk() {
                 Conversation will appear here...
               </p>
             )}
-            {transcript.map((entry, i) => (
-              <div key={i} style={{ display: 'flex', alignItems: 'flex-start', gap: 'var(--spacing-xs)' }}>
-                <i className={entry.role === 'user' ? 'fa-solid fa-user' : 'fa-solid fa-robot'}
-                  style={{
-                    color: entry.role === 'user' ? 'var(--color-primary)' : 'var(--color-accent)',
-                    marginTop: 3, flexShrink: 0, fontSize: '0.75rem',
-                  }} />
-                <p style={{ margin: 0 }}>{entry.text}</p>
-              </div>
-            ))}
+            {transcript.map((entry, i) => {
+              const isToolCall = entry.role === 'tool_call'
+              const isToolResult = entry.role === 'tool_result'
+              const isUser = entry.role === 'user'
+              const iconClass = isToolCall ? 'fa-solid fa-screwdriver-wrench'
+                              : isToolResult ? 'fa-solid fa-clipboard-list'
+                              : isUser ? 'fa-solid fa-user' : 'fa-solid fa-robot'
+              const iconColor = isToolCall || isToolResult ? 'var(--color-text-secondary)'
+                              : isUser ? 'var(--color-primary)' : 'var(--color-accent)'
+              return (
+                <div key={i} style={{ display: 'flex', alignItems: 'flex-start', gap: 'var(--spacing-xs)' }}>
+                  <i className={iconClass} style={{ color: iconColor, marginTop: 3, flexShrink: 0, fontSize: '0.75rem' }} />
+                  <p style={{
+                    margin: 0,
+                    fontFamily: (isToolCall || isToolResult) ? 'var(--font-mono)' : undefined,
+                    fontSize: (isToolCall || isToolResult) ? '0.8125rem' : undefined,
+                    color: (isToolCall || isToolResult) ? 'var(--color-text-secondary)' : undefined,
+                    whiteSpace: isToolResult ? 'pre-wrap' : undefined,
+                  }}>{entry.text}</p>
+                </div>
+              )
+            })}
             <div ref={transcriptEndRef} />
           </div>
 

core/http/react-ui/src/utils/capabilities.js

@@ -20,3 +20,4 @@ export const CAP_DETECTION = 'FLAG_DETECTION'
 export const CAP_FACE_RECOGNITION = 'FLAG_FACE_RECOGNITION'
 export const CAP_SPEAKER_RECOGNITION = 'FLAG_SPEAKER_RECOGNITION'
 export const CAP_AUDIO_TRANSFORM = 'FLAG_AUDIO_TRANSFORM'
+export const CAP_REALTIME_AUDIO = 'FLAG_REALTIME_AUDIO'

core/http/routes/ui.go

@@ -18,7 +18,11 @@ func RegisterUIRoutes(app *echo.Echo,
 	// SPA routes are handled by the 404 fallback in app.go which serves
 	// index.html for any unmatched HTML request, enabling client-side routing.
 
-	// Pipeline models API (for the Talk page WebRTC interface)
+	// Pipeline models API (for the Talk page WebRTC interface).
+	// A model qualifies when it either declares an explicit VAD+STT+LLM+TTS
+	// pipeline (legacy/composed) or carries the realtime_audio usecase (a
+	// self-contained any-to-any audio backend like liquid-audio that owns the
+	// full loop in a single AudioToAudioStream RPC).
 	app.GET("/api/pipeline-models", func(c echo.Context) error {
 		type pipelineModelInfo struct {
 			Name          string `json:"name"`
@@ -27,9 +31,17 @@ func RegisterUIRoutes(app *echo.Echo,
 			LLM           string `json:"llm"`
 			TTS           string `json:"tts"`
 			Voice         string `json:"voice"`
+			// SelfContained is true for any-to-any audio models — the four
+			// pipeline slots are populated with the model's own name so the
+			// UI can render them, but the Realtime API routes the session
+			// directly to the backend's AudioToAudioStream RPC.
+			SelfContained bool `json:"self_contained,omitempty"`
 		}
 
 		pipelineModels := cl.GetModelConfigsByFilter(func(_ string, cfg *config.ModelConfig) bool {
+			if cfg.HasUsecases(config.FLAG_REALTIME_AUDIO) {
+				return true
+			}
 			p := cfg.Pipeline
 			return p.VAD != "" && p.Transcription != "" && p.LLM != "" && p.TTS != ""
 		})
@@ -38,8 +50,20 @@ func RegisterUIRoutes(app *echo.Echo,
 			return cmp.Compare(a.Name, b.Name)
 		})
 
-		var models []pipelineModelInfo
+		models := make([]pipelineModelInfo, 0, len(pipelineModels))
 		for _, cfg := range pipelineModels {
+			if cfg.HasUsecases(config.FLAG_REALTIME_AUDIO) {
+				models = append(models, pipelineModelInfo{
+					Name:          cfg.Name,
+					VAD:           cfg.Name,
+					Transcription: cfg.Name,
+					LLM:           cfg.Name,
+					TTS:           cfg.Name,
+					Voice:         cfg.TTSConfig.Voice,
+					SelfContained: true,
+				})
+				continue
+			}
 			models = append(models, pipelineModelInfo{
 				Name:          cfg.Name,
 				VAD:           cfg.Pipeline.VAD,

core/http/routes/ui_api.go

@@ -54,6 +54,7 @@ var usecaseFilters = map[string]config.ModelConfigUsecase{
 	config.UsecaseVAD:             config.FLAG_VAD,
 	config.UsecaseAudioTransform:  config.FLAG_AUDIO_TRANSFORM,
 	config.UsecaseDiarization:     config.FLAG_DIARIZATION,
+	config.UsecaseRealtimeAudio:   config.FLAG_REALTIME_AUDIO,
 }
 
 

core/http/routes/ui_pipeline_models_test.go

@@ -0,0 +1,153 @@
+package routes_test
+
+import (
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+
+	"github.com/labstack/echo/v4"
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/http/routes"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("Pipeline models API", func() {
+	var (
+		app          *echo.Echo
+		tempDir      string
+		configLoader *config.ModelConfigLoader
+	)
+
+	BeforeEach(func() {
+		var err error
+		tempDir, err = os.MkdirTemp("", "pipeline-models-test-*")
+		Expect(err).NotTo(HaveOccurred())
+
+		configLoader = config.NewModelConfigLoader(tempDir)
+	})
+
+	AfterEach(func() {
+		Expect(os.RemoveAll(tempDir)).To(Succeed())
+	})
+
+	writeConfig := func(name, body string) {
+		path := filepath.Join(tempDir, name+".yaml")
+		Expect(os.WriteFile(path, []byte(body), 0o644)).To(Succeed())
+	}
+
+	queryPipelineModels := func() []map[string]any {
+		Expect(configLoader.LoadModelConfigsFromPath(tempDir)).To(Succeed())
+
+		app = echo.New()
+		routes.RegisterUIRoutes(app, configLoader, nil, nil, func(next echo.HandlerFunc) echo.HandlerFunc { return next })
+
+		req := httptest.NewRequest(http.MethodGet, "/api/pipeline-models", nil)
+		rec := httptest.NewRecorder()
+		app.ServeHTTP(rec, req)
+		Expect(rec.Code).To(Equal(http.StatusOK))
+		body, err := io.ReadAll(rec.Body)
+		Expect(err).NotTo(HaveOccurred())
+
+		var got []map[string]any
+		Expect(json.Unmarshal(body, &got)).To(Succeed())
+		return got
+	}
+
+	It("returns models with an explicit VAD/STT/LLM/TTS pipeline", func() {
+		writeConfig("legacy-pipeline", `
+name: legacy-pipeline
+backend: llama-cpp
+pipeline:
+  vad: silero
+  transcription: whisper
+  llm: llama
+  tts: piper
+tts:
+  voice: en-amy
+`)
+		// A model with a partial pipeline must not appear.
+		writeConfig("half-pipeline", `
+name: half-pipeline
+backend: llama-cpp
+pipeline:
+  vad: silero
+  transcription: whisper
+`)
+
+		models := queryPipelineModels()
+		Expect(models).To(HaveLen(1))
+		Expect(models[0]["name"]).To(Equal("legacy-pipeline"))
+		Expect(models[0]["vad"]).To(Equal("silero"))
+		Expect(models[0]["llm"]).To(Equal("llama"))
+		Expect(models[0]["voice"]).To(Equal("en-amy"))
+		// self_contained is omitempty — absent for legacy pipelines.
+		_, hasFlag := models[0]["self_contained"]
+		Expect(hasFlag).To(BeFalse())
+	})
+
+	It("surfaces self-contained any-to-any models tagged with realtime_audio", func() {
+		writeConfig("lfm-realtime", `
+name: lfm-realtime
+backend: liquid-audio
+known_usecases:
+  - realtime_audio
+  - chat
+  - tts
+  - transcript
+tts:
+  voice: us_female
+`)
+
+		models := queryPipelineModels()
+		Expect(models).To(HaveLen(1))
+		Expect(models[0]["name"]).To(Equal("lfm-realtime"))
+		// All four pipeline slots are populated with the model's own name so
+		// the Talk page UI has something to render.
+		Expect(models[0]["vad"]).To(Equal("lfm-realtime"))
+		Expect(models[0]["transcription"]).To(Equal("lfm-realtime"))
+		Expect(models[0]["llm"]).To(Equal("lfm-realtime"))
+		Expect(models[0]["tts"]).To(Equal("lfm-realtime"))
+		Expect(models[0]["voice"]).To(Equal("us_female"))
+		Expect(models[0]["self_contained"]).To(BeTrue())
+	})
+
+	It("includes both legacy and self-contained models in the same response", func() {
+		writeConfig("legacy", `
+name: legacy
+backend: llama-cpp
+pipeline:
+  vad: silero
+  transcription: whisper
+  llm: llama
+  tts: piper
+`)
+		writeConfig("realtime", `
+name: realtime
+backend: liquid-audio
+known_usecases:
+  - realtime_audio
+`)
+
+		models := queryPipelineModels()
+		Expect(models).To(HaveLen(2))
+		// Sorted by name → legacy, realtime.
+		Expect(models[0]["name"]).To(Equal("legacy"))
+		Expect(models[1]["name"]).To(Equal("realtime"))
+		Expect(models[1]["self_contained"]).To(BeTrue())
+	})
+
+	It("excludes models that have neither a pipeline nor realtime_audio", func() {
+		writeConfig("plain-chat", `
+name: plain-chat
+backend: llama-cpp
+known_usecases:
+  - chat
+`)
+
+		Expect(queryPipelineModels()).To(BeEmpty())
+	})
+})

core/http/static/chat.js

@@ -1212,6 +1212,9 @@ async function promptGPT(systemPrompt, input) {
 
   // Add stream parameter for both regular chat and MCP (MCP now supports SSE streaming)
   requestBody.stream = true;
+  // include_usage tells LocalAI to emit a trailing chunk with token totals;
+  // the spec-compliant server otherwise drops `usage` from the stream.
+  requestBody.stream_options = { include_usage: true };
   
   // Add generation parameters if they are set (null means use default)
   if (activeChat.temperature !== null && activeChat.temperature !== undefined) {

core/schema/ollama.go

@@ -2,6 +2,8 @@ package schema
 
 import (
 	"context"
+	"encoding/json"
+	"fmt"
 	"time"
 )
 
@@ -18,6 +20,79 @@ type OllamaOptions struct {
 	NumCtx        int      `json:"num_ctx,omitempty"`
 }
 
+// UnmarshalJSON accepts integer parameters encoded as either JSON ints
+// (`8192`) or JSON floats (`8192.0`). Some clients - notably Home Assistant's
+// Ollama integration - serialize ints as floats, which stdlib json refuses
+// to decode into int fields. See https://github.com/mudler/LocalAI/issues/9837.
+func (o *OllamaOptions) UnmarshalJSON(data []byte) error {
+	type aux struct {
+		Temperature   *float64     `json:"temperature,omitempty"`
+		TopP          *float64     `json:"top_p,omitempty"`
+		TopK          *json.Number `json:"top_k,omitempty"`
+		NumPredict    *json.Number `json:"num_predict,omitempty"`
+		RepeatPenalty float64      `json:"repeat_penalty,omitempty"`
+		RepeatLastN   *json.Number `json:"repeat_last_n,omitempty"`
+		Seed          *json.Number `json:"seed,omitempty"`
+		Stop          []string     `json:"stop,omitempty"`
+		NumCtx        *json.Number `json:"num_ctx,omitempty"`
+	}
+	var a aux
+	if err := json.Unmarshal(data, &a); err != nil {
+		return err
+	}
+
+	o.Temperature = a.Temperature
+	o.TopP = a.TopP
+	o.RepeatPenalty = a.RepeatPenalty
+	o.Stop = a.Stop
+
+	var err error
+	if o.TopK, err = jsonNumberToIntPtr(a.TopK); err != nil {
+		return fmt.Errorf("options.top_k: %w", err)
+	}
+	if o.NumPredict, err = jsonNumberToIntPtr(a.NumPredict); err != nil {
+		return fmt.Errorf("options.num_predict: %w", err)
+	}
+	if o.Seed, err = jsonNumberToIntPtr(a.Seed); err != nil {
+		return fmt.Errorf("options.seed: %w", err)
+	}
+	if o.RepeatLastN, err = jsonNumberToInt(a.RepeatLastN); err != nil {
+		return fmt.Errorf("options.repeat_last_n: %w", err)
+	}
+	if o.NumCtx, err = jsonNumberToInt(a.NumCtx); err != nil {
+		return fmt.Errorf("options.num_ctx: %w", err)
+	}
+	return nil
+}
+
+// jsonNumberToInt parses a json.Number literal as an int, tolerating both
+// integer (`8192`) and float (`8192.0`) encodings. A nil pointer or empty
+// string yields 0, matching the zero-value semantics of the int fields.
+func jsonNumberToInt(n *json.Number) (int, error) {
+	if n == nil || *n == "" {
+		return 0, nil
+	}
+	if i, err := n.Int64(); err == nil {
+		return int(i), nil
+	}
+	f, err := n.Float64()
+	if err != nil {
+		return 0, err
+	}
+	return int(f), nil
+}
+
+func jsonNumberToIntPtr(n *json.Number) (*int, error) {
+	if n == nil {
+		return nil, nil
+	}
+	i, err := jsonNumberToInt(n)
+	if err != nil {
+		return nil, err
+	}
+	return &i, nil
+}
+
 // OllamaMessage represents a message in Ollama chat format
 type OllamaMessage struct {
 	Role      string   `json:"role"`

core/schema/ollama_test.go

@@ -84,3 +84,94 @@ var _ = Describe("OllamaEmbedRequest", func() {
 		})
 	})
 })
+
+// Several Ollama clients (notably Home Assistant's Python client) encode
+// integer parameters as JSON floats (`8192.0`). Stdlib json refuses to
+// unmarshal those into `int` fields, so OllamaOptions has a custom
+// UnmarshalJSON that accepts both forms. See
+// https://github.com/mudler/LocalAI/issues/9837.
+var _ = Describe("OllamaOptions JSON unmarshaling", func() {
+	It("accepts integer literals for int fields", func() {
+		body := []byte(`{"num_ctx": 8192, "num_predict": 256, "top_k": 40, "seed": 7, "repeat_last_n": 64}`)
+
+		var opts OllamaOptions
+		Expect(json.Unmarshal(body, &opts)).To(Succeed())
+
+		Expect(opts.NumCtx).To(Equal(8192))
+		Expect(opts.NumPredict).NotTo(BeNil())
+		Expect(*opts.NumPredict).To(Equal(256))
+		Expect(opts.TopK).NotTo(BeNil())
+		Expect(*opts.TopK).To(Equal(40))
+		Expect(opts.Seed).NotTo(BeNil())
+		Expect(*opts.Seed).To(Equal(7))
+		Expect(opts.RepeatLastN).To(Equal(64))
+	})
+
+	It("accepts float literals for int fields (Home Assistant Ollama client)", func() {
+		body := []byte(`{"num_ctx": 8192.0, "num_predict": 256.0, "top_k": 40.0, "seed": 7.0, "repeat_last_n": 64.0}`)
+
+		var opts OllamaOptions
+		Expect(json.Unmarshal(body, &opts)).To(Succeed())
+
+		Expect(opts.NumCtx).To(Equal(8192))
+		Expect(opts.NumPredict).NotTo(BeNil())
+		Expect(*opts.NumPredict).To(Equal(256))
+		Expect(opts.TopK).NotTo(BeNil())
+		Expect(*opts.TopK).To(Equal(40))
+		Expect(opts.Seed).NotTo(BeNil())
+		Expect(*opts.Seed).To(Equal(7))
+		Expect(opts.RepeatLastN).To(Equal(64))
+	})
+
+	It("preserves float fields and stop list", func() {
+		body := []byte(`{"temperature": 0.7, "top_p": 0.9, "repeat_penalty": 1.1, "stop": ["<|end|>", "</s>"]}`)
+
+		var opts OllamaOptions
+		Expect(json.Unmarshal(body, &opts)).To(Succeed())
+
+		Expect(opts.Temperature).NotTo(BeNil())
+		Expect(*opts.Temperature).To(Equal(0.7))
+		Expect(opts.TopP).NotTo(BeNil())
+		Expect(*opts.TopP).To(Equal(0.9))
+		Expect(opts.RepeatPenalty).To(Equal(1.1))
+		Expect(opts.Stop).To(Equal([]string{"<|end|>", "</s>"}))
+	})
+
+	It("leaves optional int fields nil when absent", func() {
+		body := []byte(`{}`)
+
+		var opts OllamaOptions
+		Expect(json.Unmarshal(body, &opts)).To(Succeed())
+
+		Expect(opts.NumPredict).To(BeNil())
+		Expect(opts.TopK).To(BeNil())
+		Expect(opts.Seed).To(BeNil())
+		Expect(opts.NumCtx).To(Equal(0))
+		Expect(opts.RepeatLastN).To(Equal(0))
+	})
+
+	It("accepts nested options on a chat request with float num_ctx", func() {
+		// Mirrors the payload Home Assistant sends; reproduces issue #9837.
+		body := []byte(`{
+			"model": "qwen2",
+			"messages": [{"role": "user", "content": "hi"}],
+			"options": {"num_ctx": 8192.0, "top_k": 40.0}
+		}`)
+
+		var req OllamaChatRequest
+		Expect(json.Unmarshal(body, &req)).To(Succeed())
+
+		Expect(req.Options).NotTo(BeNil())
+		Expect(req.Options.NumCtx).To(Equal(8192))
+		Expect(req.Options.TopK).NotTo(BeNil())
+		Expect(*req.Options.TopK).To(Equal(40))
+	})
+
+	It("rejects non-numeric values with a clear error", func() {
+		body := []byte(`{"num_ctx": "not-a-number"}`)
+
+		var opts OllamaOptions
+		err := json.Unmarshal(body, &opts)
+		Expect(err).To(HaveOccurred())
+	})
+})

core/schema/openai.go

@@ -82,7 +82,21 @@ type OpenAIResponse struct {
 	Choices []Choice `json:"choices,omitempty"`
 	Data    []Item   `json:"data,omitempty"`
 
-	Usage OpenAIUsage `json:"usage"`
+	// Usage is intentionally a pointer with omitempty: per the OpenAI
+	// chat-completion streaming spec, intermediate chunks must not carry
+	// a `usage` field. Marshalling a value-typed usage would emit
+	// `"usage":{"prompt_tokens":0,...}` on every chunk and break
+	// OpenAI-SDK consumers that filter on a truthy `result.usage`
+	// (continuedev/continue, Kilo Code, Roo Code, etc.).
+	Usage *OpenAIUsage `json:"usage,omitempty"`
+}
+
+// StreamOptions mirrors OpenAI's `stream_options` request field. The only
+// member currently honored is IncludeUsage; when true, the streaming
+// chat-completion response emits a trailing chunk with `choices:[]` and a
+// populated `usage` object.
+type StreamOptions struct {
+	IncludeUsage bool `json:"include_usage,omitempty" yaml:"include_usage,omitempty"`
 }
 
 type Choice struct {
@@ -198,6 +212,9 @@ type OpenAIRequest struct {
 
 	Stream bool `json:"stream"`
 
+	// StreamOptions opts into OpenAI streaming extensions, e.g. include_usage.
+	StreamOptions *StreamOptions `json:"stream_options,omitempty" yaml:"stream_options,omitempty"`
+
 	// Image (not supported by OpenAI)
 	Quality string `json:"quality"`
 	Step    int    `json:"step"`

core/services/agentpool/agent_jobs.go

@@ -11,7 +11,6 @@ import (
 	"io"
 	"net"
 	"net/http"
-	"os"
 	"path/filepath"
 	"slices"
 	"strings"
@@ -46,8 +45,6 @@ type AgentJobService struct {
 	tasks     *xsync.SyncedMap[string, schema.Task]
 	jobs      *xsync.SyncedMap[string, schema.Job]
 	persister JobPersister
-	tasksFile string // Path to agent_tasks.json (kept for backward compat)
-	jobsFile  string // Path to agent_jobs.json (kept for backward compat)
 	userID    string // Scoping: empty for global (main service), set for per-user instances
 
 	// Job execution channel
@@ -70,9 +67,6 @@ type AgentJobService struct {
 	// Service lifecycle
 	ctx    context.Context
 	cancel context.CancelFunc
-
-	// Mutex for file operations
-	fileMutex sync.Mutex
 }
 
 // DistributedDispatcher is the interface for distributed job dispatching via NATS.
@@ -220,8 +214,6 @@ func NewAgentJobServiceWithPaths(
 			tasksFile: tasksFile,
 			jobsFile:  jobsFile,
 		},
-		tasksFile:     tasksFile,
-		jobsFile:      jobsFile,
 		jobQueue:      make(chan JobExecution, 100), // Buffer for 100 jobs
 		cancellations: xsync.NewSyncedMap[string, context.CancelFunc](),
 		cronScheduler: cron.New(), // Support seconds in cron
@@ -230,127 +222,51 @@ func NewAgentJobServiceWithPaths(
 	}
 }
 
-// LoadTasksFromFile loads tasks from agent_tasks.json
+// LoadTasksFromFile loads tasks from the persister into the in-memory map
+// and schedules cron entries. Named "FromFile" for backward compat; in DB
+// mode it loads from the database.
 func (s *AgentJobService) LoadTasksFromFile() error {
-	if s.tasksFile == "" {
-		return nil // No file path configured
-	}
-
-	s.fileMutex.Lock()
-	defer s.fileMutex.Unlock()
-
-	if _, err := os.Stat(s.tasksFile); os.IsNotExist(err) {
-		xlog.Debug("agent_tasks.json not found, starting with empty tasks")
-		return nil
-	}
-
-	fileContent, err := os.ReadFile(s.tasksFile)
+	tasks, err := s.persister.LoadTasks(s.userID)
 	if err != nil {
-		return fmt.Errorf("failed to read tasks file: %w", err)
+		return err
 	}
-
-	var tasksFile schema.TasksFile
-	if err := json.Unmarshal(fileContent, &tasksFile); err != nil {
-		return fmt.Errorf("failed to parse tasks file: %w", err)
-	}
-
-	for _, task := range tasksFile.Tasks {
+	for _, task := range tasks {
 		s.tasks.Set(task.ID, task)
-		// Schedule cron if enabled and has cron expression
 		if task.Enabled && task.Cron != "" {
 			if err := s.ScheduleCronTask(task); err != nil {
 				xlog.Warn("Failed to schedule cron task on load", "error", err, "task_id", task.ID)
 			}
 		}
 	}
-
-	xlog.Info("Loaded tasks from file", "count", len(tasksFile.Tasks))
-
 	return nil
 }
 
-// SaveTasksToFile saves tasks to agent_tasks.json
+// SaveTasksToFile flushes the current tasks map via the persister. File
+// persister bulk-writes the JSON file atomically; DB persister no-ops
+// because per-task SaveTask calls already wrote through.
 func (s *AgentJobService) SaveTasksToFile() error {
-	if s.tasksFile == "" {
-		return nil // No file path configured
-	}
-
-	s.fileMutex.Lock()
-	defer s.fileMutex.Unlock()
-
-	tasksFile := schema.TasksFile{
-		Tasks: s.tasks.Values(),
-	}
-
-	fileContent, err := json.MarshalIndent(tasksFile, "", "  ")
-	if err != nil {
-		return fmt.Errorf("failed to marshal tasks: %w", err)
-	}
-
-	if err := os.WriteFile(s.tasksFile, fileContent, 0600); err != nil {
-		return fmt.Errorf("failed to write tasks file: %w", err)
-	}
-
-	return nil
+	return s.persister.FlushTasks()
 }
 
-// LoadJobsFromFile loads jobs from agent_jobs.json
+// LoadJobsFromFile loads jobs from the persister into the in-memory map.
+// Named "FromFile" for backward compat; in DB mode it loads from the
+// database.
 func (s *AgentJobService) LoadJobsFromFile() error {
-	if s.jobsFile == "" {
-		return nil // No file path configured
-	}
-
-	s.fileMutex.Lock()
-	defer s.fileMutex.Unlock()
-
-	if _, err := os.Stat(s.jobsFile); os.IsNotExist(err) {
-		xlog.Debug("agent_jobs.json not found, starting with empty jobs")
-		return nil
-	}
-
-	fileContent, err := os.ReadFile(s.jobsFile)
+	jobs, err := s.persister.LoadJobs(s.userID)
 	if err != nil {
-		return fmt.Errorf("failed to read jobs file: %w", err)
+		return err
 	}
-
-	var jobsFile schema.JobsFile
-	if err := json.Unmarshal(fileContent, &jobsFile); err != nil {
-		return fmt.Errorf("failed to parse jobs file: %w", err)
-	}
-
-	// Load jobs into memory
-	for _, job := range jobsFile.Jobs {
+	for _, job := range jobs {
 		s.jobs.Set(job.ID, job)
 	}
-
-	xlog.Info("Loaded jobs from file", "count", len(jobsFile.Jobs))
 	return nil
 }
 
-// SaveJobsToFile saves jobs to agent_jobs.json
+// SaveJobsToFile flushes the current jobs map via the persister. File
+// persister bulk-writes the JSON file atomically; DB persister no-ops
+// because per-job SaveJob calls already wrote through.
 func (s *AgentJobService) SaveJobsToFile() error {
-	if s.jobsFile == "" {
-		return nil // No file path configured
-	}
-
-	s.fileMutex.Lock()
-	defer s.fileMutex.Unlock()
-
-	jobsFile := schema.JobsFile{
-		Jobs:        s.jobs.Values(),
-		LastCleanup: time.Now(),
-	}
-
-	fileContent, err := json.MarshalIndent(jobsFile, "", "  ")
-	if err != nil {
-		return fmt.Errorf("failed to marshal jobs: %w", err)
-	}
-
-	if err := os.WriteFile(s.jobsFile, fileContent, 0600); err != nil {
-		return fmt.Errorf("failed to write jobs file: %w", err)
-	}
-
-	return nil
+	return s.persister.FlushJobs()
 }
 
 // CreateTask creates a new task

core/services/agentpool/agent_jobs_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"os"
 	"path/filepath"
+	"sync"
 	"time"
 
 	"github.com/mudler/LocalAI/core/config"
@@ -281,6 +282,71 @@ var _ = Describe("AgentJobService", func() {
 			Expect(err).NotTo(HaveOccurred())
 			Expect(retrieved.TaskID).To(Equal(taskID))
 		})
+
+		It("does not surface a partial file when saves and loads race", func() {
+			// Regression for the macOS-only CI flake where a concurrent
+			// LoadJobsFromFile landed between os.WriteFile's open(O_TRUNC)
+			// and write, yielding "unexpected end of JSON input" at offset 0.
+			// Atomic temp+rename in the persister eliminates the window.
+			task := schema.Task{
+				Name:    "Race Task",
+				Model:   "test-model",
+				Prompt:  "Test prompt",
+				Enabled: true,
+			}
+
+			taskID, err := service.CreateTask(task)
+			Expect(err).NotTo(HaveOccurred())
+
+			_, err = service.ExecuteJob(taskID, map[string]string{}, "test", nil)
+			Expect(err).NotTo(HaveOccurred())
+			Expect(service.SaveJobsToFile()).To(Succeed())
+
+			newService := agentpool.NewAgentJobService(
+				appConfig,
+				modelLoader,
+				configLoader,
+				evaluator,
+			)
+
+			var wg sync.WaitGroup
+			deadline := time.Now().Add(500 * time.Millisecond)
+			readerErrs := make(chan error, 1024)
+
+			for range 4 {
+				wg.Add(1)
+				go func() {
+					defer wg.Done()
+					for time.Now().Before(deadline) {
+						_ = service.SaveJobsToFile()
+					}
+				}()
+			}
+
+			for range 4 {
+				wg.Add(1)
+				go func() {
+					defer wg.Done()
+					for time.Now().Before(deadline) {
+						if err := newService.LoadJobsFromFile(); err != nil {
+							readerErrs <- err
+							return
+						}
+					}
+				}()
+			}
+
+			wg.Wait()
+			close(readerErrs)
+
+			var firstErr error
+			for err := range readerErrs {
+				if firstErr == nil {
+					firstErr = err
+				}
+			}
+			Expect(firstErr).NotTo(HaveOccurred(), "concurrent load saw a partial/empty file")
+		})
 	})
 
 	Describe("Prompt templating", func() {

core/services/agentpool/job_persister.go

@@ -16,6 +16,12 @@ type JobPersister interface {
 	SaveJob(userID string, job schema.Job) error
 	DeleteJob(jobID string) error
 
+	// Bulk flush of the current in-memory state. File-backed persister
+	// rewrites the whole JSON file; DB-backed persister no-ops because
+	// SaveTask/SaveJob are already write-through.
+	FlushTasks() error
+	FlushJobs() error
+
 	// Authoritative reads — DB returns fresh data; file returns nil, nil
 	GetJob(jobID string) (*schema.Job, error)
 	ListJobs(userID, taskID, status string, limit int) ([]schema.Job, error)

core/services/agentpool/job_persister_db.go

@@ -32,6 +32,12 @@ func (p *dbJobPersister) DeleteJob(jobID string) error {
 	return p.store.DeleteJob(jobID)
 }
 
+// FlushTasks is a no-op: SaveTask already writes through to the DB.
+func (p *dbJobPersister) FlushTasks() error { return nil }
+
+// FlushJobs is a no-op: SaveJob already writes through to the DB.
+func (p *dbJobPersister) FlushJobs() error { return nil }
+
 func (p *dbJobPersister) GetJob(jobID string) (*schema.Job, error) {
 	rec, err := p.store.GetJob(jobID)
 	if err != nil {

core/services/agentpool/job_persister_file.go

@@ -4,6 +4,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"os"
+	"path/filepath"
 	"sync"
 	"time"
 
@@ -41,6 +42,14 @@ func (p *fileJobPersister) DeleteJob(_ string) error {
 	return p.saveJobsToFile()
 }
 
+func (p *fileJobPersister) FlushTasks() error {
+	return p.saveTasksToFile()
+}
+
+func (p *fileJobPersister) FlushJobs() error {
+	return p.saveJobsToFile()
+}
+
 // GetJob returns nil — file persister has no authoritative reads.
 func (p *fileJobPersister) GetJob(_ string) (*schema.Job, error) {
 	return nil, nil
@@ -127,7 +136,7 @@ func (p *fileJobPersister) saveTasksToFile() error {
 		return fmt.Errorf("failed to marshal tasks: %w", err)
 	}
 
-	return os.WriteFile(p.tasksFile, data, 0600)
+	return writeFileAtomic(p.tasksFile, data, 0600)
 }
 
 // saveJobsToFile serializes the entire jobs map to the JSON file.
@@ -149,5 +158,45 @@ func (p *fileJobPersister) saveJobsToFile() error {
 		return fmt.Errorf("failed to marshal jobs: %w", err)
 	}
 
-	return os.WriteFile(p.jobsFile, data, 0600)
+	return writeFileAtomic(p.jobsFile, data, 0600)
+}
+
+// writeFileAtomic writes data to path via a same-directory temp file + rename.
+// os.WriteFile opens with O_TRUNC, so a concurrent reader can land between the
+// truncate and the write and see an empty file ("unexpected end of JSON input").
+// rename(2) is atomic on POSIX, so readers see either the prior contents or the
+// new contents and never a zero-byte window.
+func writeFileAtomic(path string, data []byte, perm os.FileMode) error {
+	dir := filepath.Dir(path)
+	tmp, err := os.CreateTemp(dir, filepath.Base(path)+".tmp-*")
+	if err != nil {
+		return fmt.Errorf("failed to create temp file: %w", err)
+	}
+	tmpPath := tmp.Name()
+	removeTmp := func() { _ = os.Remove(tmpPath) }
+
+	if _, err := tmp.Write(data); err != nil {
+		_ = tmp.Close()
+		removeTmp()
+		return fmt.Errorf("failed to write temp file: %w", err)
+	}
+	if err := tmp.Chmod(perm); err != nil {
+		_ = tmp.Close()
+		removeTmp()
+		return fmt.Errorf("failed to chmod temp file: %w", err)
+	}
+	if err := tmp.Sync(); err != nil {
+		_ = tmp.Close()
+		removeTmp()
+		return fmt.Errorf("failed to sync temp file: %w", err)
+	}
+	if err := tmp.Close(); err != nil {
+		removeTmp()
+		return fmt.Errorf("failed to close temp file: %w", err)
+	}
+	if err := os.Rename(tmpPath, path); err != nil {
+		removeTmp()
+		return fmt.Errorf("failed to rename temp file: %w", err)
+	}
+	return nil
 }

core/services/galleryop/list_models.go

@@ -16,6 +16,14 @@ const (
 
 func ListModels(bcl *config.ModelConfigLoader, ml *model.ModelLoader, filter config.ModelConfigFilterFn, looseFilePolicy LooseFilePolicy) ([]string, error) {
 
+	// Callers (e.g. the Ollama /api/tags handler) pass nil to mean "no
+	// filtering". Without this guard the loose-file loop below dereferences
+	// filter and panics, which Echo surfaces to clients as a dropped
+	// connection (see issue #9817).
+	if filter == nil {
+		filter = config.NoFilterFn
+	}
+
 	skipMap := map[string]struct{}{}
 
 	dataModels := []string{}

core/services/galleryop/list_models_test.go

@@ -0,0 +1,64 @@
+package galleryop_test
+
+import (
+	"os"
+	"path/filepath"
+
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/services/galleryop"
+	"github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/system"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+// Regression test for issue #9817: the Ollama /api/tags handler calls
+// ListModels with a nil filter, which used to panic as soon as a loose file
+// existed under ModelsPath. The panic surfaced to Ollama clients (e.g. Home
+// Assistant) as "Server disconnected without sending a response".
+var _ = Describe("ListModels", func() {
+	var (
+		tempDir     string
+		bcl         *config.ModelConfigLoader
+		ml          *model.ModelLoader
+		systemState *system.SystemState
+	)
+
+	BeforeEach(func() {
+		var err error
+		tempDir, err = os.MkdirTemp("", "list-models-test-*")
+		Expect(err).NotTo(HaveOccurred())
+
+		systemState, err = system.GetSystemState(system.WithModelPath(tempDir))
+		Expect(err).NotTo(HaveOccurred())
+		ml = model.NewModelLoader(systemState)
+		bcl = config.NewModelConfigLoader(tempDir)
+	})
+
+	AfterEach(func() {
+		os.RemoveAll(tempDir)
+	})
+
+	It("does not panic with a nil filter when loose files exist", func() {
+		// ListFilesInModelPath skips well-known weight-file extensions
+		// (.gguf, .bin, ...) so use an extension-less file to ensure the
+		// filter path is exercised.
+		Expect(os.WriteFile(filepath.Join(tempDir, "loose-model"), []byte("x"), 0o644)).To(Succeed())
+
+		var names []string
+		var err error
+		Expect(func() {
+			names, err = galleryop.ListModels(bcl, ml, nil, galleryop.SKIP_IF_CONFIGURED)
+		}).ToNot(Panic())
+		Expect(err).ToNot(HaveOccurred())
+		Expect(names).To(ContainElement("loose-model"))
+	})
+
+	It("does not panic with a nil filter when ModelsPath is empty", func() {
+		Expect(func() {
+			_, err := galleryop.ListModels(bcl, ml, nil, galleryop.SKIP_IF_CONFIGURED)
+			Expect(err).ToNot(HaveOccurred())
+		}).ToNot(Panic())
+	})
+})

core/services/nodes/health.go

@@ -12,6 +12,24 @@ import (
 	"gorm.io/gorm"
 )
 
+// perModelMissThreshold is the number of consecutive failed gRPC probes
+// against a model's backend before the model is removed from the registry.
+// A single failure can be transient (network blip, brief GC pause on the
+// worker, a long-running request hogging the gRPC server thread); requiring
+// N consecutive misses avoids deleting healthy rows over noise. At the
+// default 15s tick this means a model has to be unreachable for ~45s before
+// it gets reaped.
+const perModelMissThreshold = 3
+
+// modelKey identifies a specific (node, model, replica) tuple. We track miss
+// counts per tuple because the same model name can be loaded on multiple
+// replicas on the same node.
+type modelKey struct {
+	NodeID       string
+	ModelName    string
+	ReplicaIndex int
+}
+
 // HealthMonitor periodically checks the health of registered backend nodes.
 type HealthMonitor struct {
 	registry            NodeHealthStore
@@ -21,6 +39,8 @@ type HealthMonitor struct {
 	autoOffline         bool                 // mark stale nodes as offline (preserves approval status)
 	clientFactory       BackendClientFactory // creates gRPC backend clients
 	perModelHealthCheck bool                 // check each model's backend process individually
+	missesMu            sync.Mutex
+	misses              map[modelKey]int // consecutive failed-probe counts; reset on success or model removal
 	cancel              context.CancelFunc
 	cancelMu            sync.Mutex
 }
@@ -46,6 +66,7 @@ func NewHealthMonitor(registry NodeHealthStore, db *gorm.DB, checkInterval, stal
 		autoOffline:         true,
 		clientFactory:       factory,
 		perModelHealthCheck: perModelHealthCheck,
+		misses:              make(map[modelKey]int),
 	}
 }
 
@@ -152,9 +173,11 @@ func (hm *HealthMonitor) doCheckAll(ctx context.Context) {
 			}
 		}
 
-		// Per-model backend health check (opt-in): probe each model's gRPC address
-		// and remove stale model records. This does NOT affect the node's status —
-		// a crashed backend process is a model-level issue, not a node-level one.
+		// Per-model backend health check: probe each model's gRPC address and
+		// remove stale model records. This does NOT affect the node's status —
+		// a crashed backend process is a model-level issue, not a node-level
+		// one. A model is only removed after perModelMissThreshold consecutive
+		// failed probes so a single network/GC blip doesn't force a reload.
 		if hm.perModelHealthCheck {
 			models, _ := hm.registry.GetNodeModels(ctx, node.ID)
 			for _, m := range models {
@@ -163,15 +186,43 @@ func (hm *HealthMonitor) doCheckAll(ctx context.Context) {
 				}
 				mClient := hm.clientFactory.NewClient(m.Address, false)
 				mCheckCtx, mCancel := context.WithTimeout(ctx, 5*time.Second)
-				if ok, _ := mClient.HealthCheck(mCheckCtx); !ok {
-					xlog.Warn("Model backend unhealthy, removing from registry",
-						"node", node.ID, "model", m.ModelName, "replica", m.ReplicaIndex, "address", m.Address)
-					hm.registry.RemoveNodeModel(ctx, node.ID, m.ModelName, m.ReplicaIndex)
-				}
+				ok, _ := mClient.HealthCheck(mCheckCtx)
 				mCancel()
 				if closer, ok := mClient.(io.Closer); ok {
 					closer.Close()
 				}
+
+				key := modelKey{NodeID: node.ID, ModelName: m.ModelName, ReplicaIndex: m.ReplicaIndex}
+				hm.missesMu.Lock()
+				if ok {
+					// Probe succeeded — wipe any previous miss streak.
+					delete(hm.misses, key)
+					hm.missesMu.Unlock()
+					continue
+				}
+				hm.misses[key]++
+				misses := hm.misses[key]
+				hm.missesMu.Unlock()
+
+				if misses < perModelMissThreshold {
+					xlog.Debug("Model backend probe failed, awaiting threshold before removal",
+						"node", node.ID, "model", m.ModelName, "replica", m.ReplicaIndex,
+						"address", m.Address, "misses", misses, "threshold", perModelMissThreshold)
+					continue
+				}
+				xlog.Warn("Model backend unhealthy after consecutive misses, removing from registry",
+					"node", node.ID, "model", m.ModelName, "replica", m.ReplicaIndex,
+					"address", m.Address, "misses", misses)
+				if err := hm.registry.RemoveNodeModel(ctx, node.ID, m.ModelName, m.ReplicaIndex); err != nil {
+					xlog.Warn("Failed to remove unhealthy model from registry",
+						"node", node.ID, "model", m.ModelName, "replica", m.ReplicaIndex, "error", err)
+					// Leave the miss counter in place so the next tick retries
+					// the removal rather than starting the streak over.
+					continue
+				}
+				hm.missesMu.Lock()
+				delete(hm.misses, key)
+				hm.missesMu.Unlock()
 			}
 		}
 	}

core/services/nodes/health_mock_test.go

@@ -232,6 +232,9 @@ func (c *fakeBackendClient) AudioTransform(_ context.Context, _ *pb.AudioTransfo
 func (c *fakeBackendClient) AudioTransformStream(_ context.Context, _ ...ggrpc.CallOption) (grpc.AudioTransformStreamClient, error) {
 	return nil, nil
 }
+func (c *fakeBackendClient) AudioToAudioStream(_ context.Context, _ ...ggrpc.CallOption) (grpc.AudioToAudioStreamClient, error) {
+	return nil, nil
+}
 func (c *fakeBackendClient) ModelMetadata(_ context.Context, _ *pb.ModelOptions, _ ...ggrpc.CallOption) (*pb.ModelMetadataResponse, error) {
 	return nil, nil
 }
@@ -321,6 +324,7 @@ func newTestHealthMonitor(store NodeHealthStore, factory BackendClientFactory, a
 		staleThreshold: staleThreshold,
 		autoOffline:    autoOffline,
 		clientFactory:  factory,
+		misses:         make(map[modelKey]int),
 	}
 }
 

core/services/nodes/health_test.go

@@ -255,7 +255,7 @@ var _ = Describe("HealthMonitor (mock-based)", func() {
 			Expect(calls).NotTo(ContainElement(ContainSubstring("MarkUnhealthy")))
 		})
 
-		It("removes stale model via per-model health check without affecting node status", func() {
+		It("removes stale model via per-model health check after consecutive failures", func() {
 			store := newFakeNodeHealthStore()
 			factory := newFakeBackendClientFactory()
 			hm := newTestHealthMonitor(store, factory, true, staleThreshold)
@@ -268,6 +268,15 @@ var _ = Describe("HealthMonitor (mock-based)", func() {
 			// Model backend is dead
 			factory.setClient("10.0.0.10:50053", &fakeBackendClient{healthy: false, err: fmt.Errorf("connection refused")})
 
+			// First (perModelMissThreshold-1) probes must NOT remove the row —
+			// a single failure could be a transient blip.
+			for i := 0; i < perModelMissThreshold-1; i++ {
+				hm.doCheckAll(context.Background())
+				Expect(store.getCalls()).NotTo(ContainElement(ContainSubstring("RemoveNodeModel")),
+					"removed too early at miss %d", i+1)
+			}
+
+			// Threshold-th consecutive miss triggers removal.
 			hm.doCheckAll(context.Background())
 
 			// Node should remain healthy — only the specific replica record is removed.
@@ -275,5 +284,36 @@ var _ = Describe("HealthMonitor (mock-based)", func() {
 			Expect(store.getCalls()).To(ContainElement("RemoveNodeModel:node-model:piper-model:0"))
 			Expect(store.getCalls()).NotTo(ContainElement(ContainSubstring("MarkUnhealthy")))
 		})
+
+		It("preserves model row when an intermittent failure is followed by a success", func() {
+			store := newFakeNodeHealthStore()
+			factory := newFakeBackendClientFactory()
+			hm := newTestHealthMonitor(store, factory, true, staleThreshold)
+			hm.perModelHealthCheck = true
+
+			node := makeTestNode("node-flap", "flap-worker", "10.0.0.11:50051", StatusHealthy, freshTime())
+			store.addNode(node)
+			store.addNodeModel("node-flap", NodeModel{NodeID: "node-flap", ModelName: "piper-model", Address: "10.0.0.11:50053"})
+
+			deadClient := &fakeBackendClient{healthy: false, err: fmt.Errorf("connection refused")}
+			liveClient := &fakeBackendClient{healthy: true}
+
+			// Two failing probes then a recovery — should NOT remove the row,
+			// and should reset the miss counter so two more failures don't tip
+			// it over.
+			factory.setClient("10.0.0.11:50053", deadClient)
+			hm.doCheckAll(context.Background())
+			hm.doCheckAll(context.Background())
+			factory.setClient("10.0.0.11:50053", liveClient)
+			hm.doCheckAll(context.Background())
+
+			Expect(store.getCalls()).NotTo(ContainElement(ContainSubstring("RemoveNodeModel")))
+
+			// Counter is reset; two more failures must not be enough to remove.
+			factory.setClient("10.0.0.11:50053", deadClient)
+			hm.doCheckAll(context.Background())
+			hm.doCheckAll(context.Background())
+			Expect(store.getCalls()).NotTo(ContainElement(ContainSubstring("RemoveNodeModel")))
+		})
 	})
 })

core/services/nodes/inflight_test.go

@@ -176,6 +176,10 @@ func (f *fakeGRPCBackend) AudioTransformStream(_ context.Context, _ ...ggrpc.Cal
 	return nil, nil
 }
 
+func (f *fakeGRPCBackend) AudioToAudioStream(_ context.Context, _ ...ggrpc.CallOption) (grpc.AudioToAudioStreamClient, error) {
+	return nil, nil
+}
+
 func (f *fakeGRPCBackend) ModelMetadata(_ context.Context, _ *pb.ModelOptions, _ ...ggrpc.CallOption) (*pb.ModelMetadataResponse, error) {
 	return &pb.ModelMetadataResponse{}, nil
 }

core/services/nodes/registry.go

@@ -546,7 +546,13 @@ func (r *NodeRegistry) GetByName(ctx context.Context, name string) (*BackendNode
 	return &node, nil
 }
 
-// MarkUnhealthy sets a node status to unhealthy.
+// MarkUnhealthy sets a node status to unhealthy. Deliberately status-only:
+// callers fire this on transient triggers (a single nats.ErrNoResponders from
+// managers_distributed / reconciler) where the next heartbeat is expected to
+// flip the node back to healthy, and cascade-deleting node_models here would
+// force a full model reload on every brief NATS hiccup. Stale rows are reaped
+// by the per-model health probe (on by default; see HealthMonitor) and by
+// MarkOffline when the heartbeat really has gone away.
 func (r *NodeRegistry) MarkUnhealthy(ctx context.Context, nodeID string) error {
 	return r.setStatus(ctx, nodeID, StatusUnhealthy)
 }
@@ -556,9 +562,23 @@ func (r *NodeRegistry) MarkHealthy(ctx context.Context, nodeID string) error {
 	return r.setStatus(ctx, nodeID, StatusHealthy)
 }
 
-// MarkDraining sets a node status to draining (no new requests).
+// MarkDraining sets a node status to draining (no new requests) and clears its
+// model records. Routing already filters out non-healthy nodes, so removing
+// the rows on drain doesn't change new-request behavior — but it does stop the
+// Models UI from showing the node's models as "running" while the box has been
+// taken out of rotation, and it prevents stale rows from being selected if
+// (re)scheduling logic gets relaxed elsewhere. In-flight requests already hold
+// their gRPC client through Route() and will finish normally; the only
+// observable effect is that the per-call IncrementInFlight bookkeeping logs a
+// non-fatal warning, which is acceptable for a drain.
 func (r *NodeRegistry) MarkDraining(ctx context.Context, nodeID string) error {
-	return r.setStatus(ctx, nodeID, StatusDraining)
+	if err := r.setStatus(ctx, nodeID, StatusDraining); err != nil {
+		return err
+	}
+	if err := r.db.WithContext(ctx).Where("node_id = ?", nodeID).Delete(&NodeModel{}).Error; err != nil {
+		xlog.Warn("Failed to clear model records on draining", "node", nodeID, "error", err)
+	}
+	return nil
 }
 
 // FindStaleNodes returns nodes that haven't sent a heartbeat within the given threshold.
@@ -673,9 +693,18 @@ func (r *NodeRegistry) FindAndLockNodeWithModel(ctx context.Context, modelName s
 		// to moderate concurrency where requests don't overlap) collapses to
 		// "biggest GPU wins every time" and one node ends up taking nearly all
 		// the load while replicas on other nodes sit idle.
+		// Filter on backend_nodes.status = healthy in the inner JOIN itself,
+		// not only in the later node-fetch step. The previous version picked
+		// a (node_id, replica) pair purely on node_models state, then bailed
+		// out when the second query couldn't find a healthy node row — but
+		// any concurrent reader of node_models could still pick the same
+		// stale row in the same window, and other helpers that mirror this
+		// JOIN need the same invariant. Belt-and-braces: status filter here
+		// AND the status-checked node fetch below.
 		q := tx.Clauses(clause.Locking{Strength: "UPDATE"}).
 			Joins("JOIN backend_nodes ON backend_nodes.id = node_models.node_id").
-			Where("node_models.model_name = ? AND node_models.state = ?", modelName, "loaded")
+			Where("node_models.model_name = ? AND node_models.state = ? AND backend_nodes.status = ?",
+				modelName, "loaded", StatusHealthy)
 		if len(candidateNodeIDs) > 0 {
 			q = q.Where("node_models.node_id IN ?", candidateNodeIDs)
 		}

docs/content/advanced/model-configuration.md

@@ -316,23 +316,66 @@ These are set via the `options:` array in the model configuration (format: `key:
 
 #### Speculative Type Values
 
-| Type | Description |
-|------|-------------|
-| `none` | No speculative decoding (default) |
-| `draft` | Draft model-based speculation (auto-set when `draft_model` is configured) |
-| `eagle3` | EAGLE3 draft model architecture |
-| `ngram_simple` | Simple self-speculative using token history |
-| `ngram_map_k` | N-gram with key-only map |
-| `ngram_map_k4v` | N-gram with keys and 4 m-gram values |
-| `ngram_mod` | Modified n-gram speculation |
-| `ngram_cache` | 3-level n-gram cache |
+The canonical names match upstream llama.cpp (dash-separated). For backward compatibility LocalAI also accepts the underscore-separated forms and the bare `draft` / `eagle3` aliases.
 
-Multiple types can be chained by passing a comma-separated list to `spec_type` (e.g. `spec_type:ngram_simple,ngram_mod`). The runtime tries them in order and accepts the first proposal that meets the acceptance criteria.
+| Type | Aliases accepted | Description |
+|------|------------------|-------------|
+| `none` | | No speculative decoding (default) |
+| `draft-simple` | `draft`, `draft_simple` | Draft model-based speculation (auto-set when `draft_model` is configured) |
+| `draft-eagle3` | `eagle3`, `draft_eagle3` | EAGLE3 draft model architecture |
+| `ngram-simple` | `ngram_simple` | Simple self-speculative using token history |
+| `ngram-map-k` | `ngram_map_k` | N-gram with key-only map |
+| `ngram-map-k4v` | `ngram_map_k4v` | N-gram with keys and 4 m-gram values |
+| `ngram-mod` | `ngram_mod` | Modified n-gram speculation |
+| `ngram-cache` | `ngram_cache` | 3-level n-gram cache |
+
+Multiple types can be chained by passing a comma-separated list to `spec_type` (e.g. `spec_type:ngram-simple,ngram-mod`). The runtime tries them in order and accepts the first proposal that meets the acceptance criteria.
 
 {{% notice note %}}
 Speculative decoding is automatically disabled when multimodal models (with `mmproj`) are active. The `n_draft` parameter can also be overridden per-request.
 {{% /notice %}}
 
+### Reasoning Models (DeepSeek-R1, Qwen3, etc.)
+
+These load-time options control how the backend parses `<think>` reasoning blocks and how much budget the model is allowed for thinking. They are set per model via the `options:` array.
+
+| Option | Type | Default | Description |
+|--------|------|---------|-------------|
+| `reasoning_format` | string | `deepseek` | Parser for reasoning/thinking blocks. One of `none`, `auto`, `deepseek`, `deepseek-legacy` (alias `deepseek_legacy`). |
+| `enable_reasoning` / `reasoning_budget` | int | `-1` | Reasoning budget in tokens: `-1` unlimited, `0` disabled, `>0` token cap for the thinking section. |
+| `prefill_assistant` | bool | `true` | When `false`, the trailing assistant message is not pre-filled by the chat template. |
+
+{{% notice note %}}
+This is the load-time reasoning configuration. The orthogonal per-request `enable_thinking` chat-template kwarg (set via the YAML `reasoning.disable` field) toggles thinking on/off per call without restarting the model.
+{{% /notice %}}
+
+### Multimodal Backend Options
+
+| Option | Type | Default | Description |
+|--------|------|---------|-------------|
+| `mmproj_use_gpu` / `mmproj_offload` | bool | `true` | Set `false` to keep the multimodal projector on CPU (saves VRAM at cost of speed). |
+| `image_min_tokens` | int | `-1` | Minimum vision tokens per image. `-1` keeps the model default. |
+| `image_max_tokens` | int | `-1` | Maximum vision tokens per image. `-1` keeps the model default. |
+
+### Embedding & Reranking Backend Options
+
+| Option | Type | Default | Description |
+|--------|------|---------|-------------|
+| `pooling_type` / `pooling` | string | auto | Pooling strategy for embeddings: `none`, `mean`, `cls`, `last`, `rank`. Reranking automatically uses `rank`. |
+| `embd_normalize` / `embedding_normalize` | int | `2` | Normalization: `-1` none, `0` max-abs, `1` taxicab, `2` Euclidean (L2), `>2` p-norm. |
+
+### Other Backend Tuning Options
+
+These llama.cpp options are passed through the `options:` array.
+
+| Option | Type | Default | Description |
+|--------|------|---------|-------------|
+| `n_ubatch` / `ubatch` | int | same as `batch` | Physical batch size. Decouple from `n_batch` when an embedding/rerank workload needs a different value. |
+| `threads_batch` / `n_threads_batch` | int | same as `threads` | Threads used during prompt processing. `<= 0` means `hardware_concurrency()`. |
+| `direct_io` / `use_direct_io` | bool | `false` | Open the model with `O_DIRECT` (faster cold loads on NVMe; ignored if not supported). |
+| `verbosity` | int | `3` | llama.cpp internal log verbosity threshold. Higher = more verbose. |
+| `override_tensor` / `tensor_buft_overrides` | string | "" | Per-tensor buffer-type overrides for the main model. Format: `<tensor regex>=<buffer type>,<tensor regex>=<buffer type>,...`. Mirrors the existing `draft_override_tensor` syntax for the draft model. |
+
 ### Prompt Caching
 
 | Field | Type | Description |

docs/data/version.json

@@ -1,3 +1,3 @@
 {
-  "version": "v4.2.2"
+  "version": "v4.2.4"
 }

gallery/index.yaml

@@ -3197,6 +3197,110 @@
     - filename: llama-cpp/models/LFM2.5-1.2B-Nova-Function-Calling.Q4_K_M.gguf
       sha256: 5d039ad4195447cf4b6dbee8f7fe11f985c01d671a18153084c869077e431fbf
       uri: https://huggingface.co/NovachronoAI/LFM2.5-1.2B-Nova-Function-Calling-GGUF/resolve/main/LFM2.5-1.2B-Nova-Function-Calling.Q4_K_M.gguf
+- name: lfm2.5-audio-1.5b-realtime
+  url: github:mudler/LocalAI/gallery/liquid-audio.yaml@master
+  urls:
+    - https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B
+  description: |
+    LFM2.5-Audio-1.5B is LiquidAI's any-to-any audio foundation model. The
+    1.2B LFM2.5 backbone plus a FastConformer audio encoder and an LFM2-based
+    audio detokenizer give real-time speech-to-speech with text + audio output
+    interleaved at 12.5 Hz / 24 kHz. This entry runs in S2S (speech-to-speech)
+    mode and is the model the LocalAI realtime API any-to-any path consumes.
+    Switch to ASR, TTS, or chat by picking the sibling gallery entries.
+  license: LFM-Open-License-v1.0
+  icon: https://cdn-avatars.huggingface.co/v1/production/uploads/61b8e2ba285851687028d395/7_6D7rWrLxp2hb6OHSV1p.png
+  tags:
+    - lfm2
+    - liquid
+    - audio
+    - speech-to-speech
+    - any-to-any
+    - realtime
+    - 1.5b
+  last_checked: "2026-05-11"
+  overrides:
+    backend: liquid-audio
+    # realtime_audio drives the Talk-page filter; the rest let the model
+    # also surface on the chat / transcribe / speech endpoints when called
+    # directly (the backend implements all three RPCs).
+    known_usecases:
+      - realtime_audio
+      - chat
+      - transcript
+      - tts
+      - vad
+    options:
+      - mode:s2s
+- name: lfm2.5-audio-1.5b-chat
+  url: github:mudler/LocalAI/gallery/liquid-audio.yaml@master
+  urls:
+    - https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B
+  description: |
+    LFM2.5-Audio-1.5B in text-only chat mode. The model runs `generate_sequential`
+    with no audio modality, behaving like a small LFM2 chat model. Pick this
+    entry for tool-calling experiments without the audio overhead.
+  license: LFM-Open-License-v1.0
+  tags:
+    - lfm2
+    - liquid
+    - audio
+    - chat
+    - 1.5b
+  last_checked: "2026-05-11"
+  overrides:
+    backend: liquid-audio
+    known_usecases:
+      - chat
+    options:
+      - mode:chat
+- name: lfm2.5-audio-1.5b-asr
+  url: github:mudler/LocalAI/gallery/liquid-audio.yaml@master
+  urls:
+    - https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B
+  description: |
+    LFM2.5-Audio-1.5B in ASR mode. System prompt `Perform ASR.` is prepended;
+    output is capitalised and punctuated. Wire this entry as a transcription
+    model on the /v1/audio/transcriptions endpoint.
+  license: LFM-Open-License-v1.0
+  tags:
+    - lfm2
+    - liquid
+    - audio
+    - asr
+    - speech-to-text
+    - 1.5b
+  last_checked: "2026-05-11"
+  overrides:
+    backend: liquid-audio
+    known_usecases:
+      - transcript
+    options:
+      - mode:asr
+- name: lfm2.5-audio-1.5b-tts
+  url: github:mudler/LocalAI/gallery/liquid-audio.yaml@master
+  urls:
+    - https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B
+  description: |
+    LFM2.5-Audio-1.5B in TTS mode. Four baked voices: us_male, us_female,
+    uk_male, uk_female — pick the default at load time via `voice:` option,
+    or override per-request via the OpenAI `/v1/audio/speech` `voice` field.
+  license: LFM-Open-License-v1.0
+  tags:
+    - lfm2
+    - liquid
+    - audio
+    - tts
+    - text-to-speech
+    - 1.5b
+  last_checked: "2026-05-11"
+  overrides:
+    backend: liquid-audio
+    known_usecases:
+      - tts
+    options:
+      - mode:tts
+      - voice:us_female
 - name: mistral-nemo-instruct-2407-12b-thinking-m-claude-opus-high-reasoning-i1
   url: github:mudler/LocalAI/gallery/virtual.yaml@master
   urls:

gallery/lfm.yaml

@@ -10,6 +10,16 @@ config_file: |
         - <dummy32000>
         - </s>
         - <|endoftext|>
+    function:
+        # LFM2 Pythonic tool-call syntax: <|tool_call_start|>[name(k="v", ...)]<|tool_call_end|>
+        # Mirrors common_chat_params_init_lfm2 in llama.cpp/common/chat.cpp.
+        response_regex:
+            - '<\|tool_call_start\|>\[(?P<name>\w+)\((?P<arguments>.*?)\)\]<\|tool_call_end\|>'
+        argument_regex:
+            - '(?P<key>\w+)\s*=\s*"(?P<value>[^"]*)"'
+            - '(?P<key>\w+)\s*=\s*(?P<value>-?\d+(?:\.\d+)?|true|false|null)'
+        argument_regex_key_name: key
+        argument_regex_value_name: value
     template:
         chat: |
             {{.Input -}}

gallery/liquid-audio.yaml

@@ -0,0 +1,40 @@
+---
+name: "liquid-audio"
+
+description: |
+  LiquidAI LFM2 / LFM2.5 Audio models served by the Python `liquid-audio` backend.
+  Supports four roles via the `mode:` option:
+    - chat       text-only chat completion (generate_sequential, no audio)
+    - asr        speech-to-text (Perform ASR. system prompt)
+    - tts        text-to-speech in 4 baked voices (us_male/us_female/uk_male/uk_female)
+    - s2s        interleaved speech-to-speech (the realtime any-to-any path)
+
+license: "LFM Open License v1.0"
+
+urls:
+  - https://huggingface.co/LiquidAI/LFM2.5-Audio-1.5B
+  - https://github.com/Liquid4All/liquid-audio
+
+config_file: |
+  backend: liquid-audio
+  context_size: 32768
+  f16: true
+  mmap: true
+  # realtime_audio surfaces the model on the Talk page; chat/tts/transcript
+  # let it also serve the standalone /v1/chat/completions, /v1/audio/speech,
+  # and /v1/audio/transcriptions endpoints (backend implements all three).
+  known_usecases:
+    - realtime_audio
+    - chat
+    - tts
+    - transcript
+    - vad
+  parameters:
+    model: LiquidAI/LFM2.5-Audio-1.5B
+  # Special tokens emitted in the text track during interleaved generation.
+  # Included so a future client-side parser can spot them; the LFM2 tool-call
+  # format itself is auto-detected by the upstream llama.cpp parser when the
+  # model loads under that backend.
+  stopwords:
+    - <|im_end|>
+    - <|endoftext|>

go.mod

@@ -163,7 +163,7 @@ require (
 	github.com/gocolly/colly v1.2.0 // indirect
 	github.com/gofiber/fiber/v2 v2.52.13 // indirect
 	github.com/golang/protobuf v1.5.4 // indirect
-	github.com/gomarkdown/markdown v0.0.0-20250311123330-531bef5e742b // indirect
+	github.com/gomarkdown/markdown v0.0.0-20260411013819-759bbc3e3207 // indirect
 	github.com/google/go-github/v69 v69.2.0 // indirect
 	github.com/google/go-querystring v1.1.0 // indirect
 	github.com/jackc/pgpassfile v1.0.0 // indirect
@@ -359,7 +359,7 @@ require (
 	github.com/jaypipes/pcidb v1.1.1 // indirect
 	github.com/jbenet/go-temp-err-catcher v0.1.0 // indirect
 	github.com/josharian/intern v1.0.0 // indirect
-	github.com/klauspost/compress v1.18.5 // indirect
+	github.com/klauspost/compress v1.18.5
 	github.com/klauspost/pgzip v1.2.5 // indirect
 	github.com/koron/go-ssdp v0.0.6 // indirect
 	github.com/libp2p/go-buffer-pool v0.1.0 // indirect

go.sum

@@ -472,8 +472,8 @@ github.com/golang/protobuf v1.5.4/go.mod h1:lnTiLA8Wa4RWRcIUkrtSVa5nRhsEGBg48fD6
 github.com/golang/snappy v0.0.2/go.mod h1:/XxbfmMg8lxefKM7IXC3fBNl/7bRcc72aCRzEWrmP2Q=
 github.com/golang/snappy v0.0.5-0.20231225225746-43d5d4cd4e0e h1:4bw4WeyTYPp0smaXiJZCNnLrvVBqirQVreixayXezGc=
 github.com/golang/snappy v0.0.5-0.20231225225746-43d5d4cd4e0e/go.mod h1:/XxbfmMg8lxefKM7IXC3fBNl/7bRcc72aCRzEWrmP2Q=
-github.com/gomarkdown/markdown v0.0.0-20250311123330-531bef5e742b h1:EY/KpStFl60qA17CptGXhwfZ+k1sFNJIUNR8DdbcuUk=
-github.com/gomarkdown/markdown v0.0.0-20250311123330-531bef5e742b/go.mod h1:JDGcbDT52eL4fju3sZ4TeHGsQwhG9nbDV21aMyhwPoA=
+github.com/gomarkdown/markdown v0.0.0-20260411013819-759bbc3e3207 h1:p7t34F7K4OCRQblcDhNJnP46Uaarz3z2cLcvOZYxWn8=
+github.com/gomarkdown/markdown v0.0.0-20260411013819-759bbc3e3207/go.mod h1:JDGcbDT52eL4fju3sZ4TeHGsQwhG9nbDV21aMyhwPoA=
 github.com/google/btree v0.0.0-20180813153112-4030bb1f1f0c/go.mod h1:lNA+9X1NB3Zf8V7Ke586lFgjr2dZNuvo3lPJSGZ5JPQ=
 github.com/google/btree v1.0.0/go.mod h1:lNA+9X1NB3Zf8V7Ke586lFgjr2dZNuvo3lPJSGZ5JPQ=
 github.com/google/btree v1.1.3 h1:CVpQJjYgC4VbzxeGVHfvZrv1ctoYCAI8vbl07Fcxlyg=

pkg/downloader/huggingface.go

@@ -33,6 +33,7 @@ func HuggingFaceScan(uri URI) (*HuggingFaceScanResult, error) {
 	if err != nil {
 		return nil, err
 	}
+	defer results.Body.Close()
 	if results.StatusCode != 200 {
 		return nil, fmt.Errorf("unexpected status code during HuggingFaceScan: %d", results.StatusCode)
 	}

pkg/functions/parse_lfm2_test.go

@@ -0,0 +1,106 @@
+package functions_test
+
+import (
+	. "github.com/mudler/LocalAI/pkg/functions"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+// LFM2 / LFM2.5 emit tool calls in a Pythonic syntax wrapped in special tokens:
+//
+//	<|tool_call_start|>[func_name(arg1="value1", arg2="value2")]<|tool_call_end|>
+//
+// See backend/cpp/llama-cpp/llama.cpp/common/chat.cpp:1277 (common_chat_params_init_lfm2)
+// and https://docs.liquid.ai/lfm/key-concepts/tool-use. The format is auto-detected
+// by upstream llama.cpp when the chat template contains <|tool_list_start|>/<|tool_list_end|>.
+//
+// The tests below pin the LocalAI-side parser config (response_regex + argument_regex)
+// that the lfm gallery template ships, so configurations relying on the gRPC backend
+// returning raw text (rather than pre-parsed tool_calls via use_jinja) still work.
+var _ = Describe("LFM2 Pythonic tool-call parsing", func() {
+	// Matches the markers exactly; non-greedy `arguments` so the closing `)]` of one
+	// call doesn't swallow trailing content that happens to share characters.
+	const lfm2ResponseRegex = `<\|tool_call_start\|>\[(?P<name>\w+)\((?P<arguments>.*?)\)\]<\|tool_call_end\|>`
+
+	// Two argument extractors: quoted strings and bare scalars (numbers / true / false / null).
+	// ParseFunctionCallArgs runs every regex in order, so later matches with the same key
+	// would overwrite earlier ones — which is fine here because the patterns are disjoint.
+	var lfm2ArgRegex = []string{
+		`(?P<key>\w+)\s*=\s*"(?P<value>[^"]*)"`,
+		`(?P<key>\w+)\s*=\s*(?P<value>-?\d+(?:\.\d+)?|true|false|null)`,
+	}
+
+	cfg := func() FunctionsConfig {
+		return FunctionsConfig{
+			ResponseRegex:      []string{lfm2ResponseRegex},
+			ArgumentRegex:      lfm2ArgRegex,
+			ArgumentRegexKey:   "key",
+			ArgumentRegexValue: "value",
+		}
+	}
+
+	It("parses a single string-arg call", func() {
+		input := `<|tool_call_start|>[get_weather(city="Berlin")]<|tool_call_end|>`
+		results := ParseFunctionCall(input, cfg())
+		Expect(results).To(HaveLen(1))
+		Expect(results[0].Name).To(Equal("get_weather"))
+		Expect(results[0].Arguments).To(Equal(`{"city":"Berlin"}`))
+	})
+
+	It("parses multiple string args", func() {
+		input := `<|tool_call_start|>[search(query="hello world", source="web")]<|tool_call_end|>`
+		results := ParseFunctionCall(input, cfg())
+		Expect(results).To(HaveLen(1))
+		Expect(results[0].Name).To(Equal("search"))
+		// argument map ordering is not stable; check content as JSON
+		Expect(results[0].Arguments).To(SatisfyAny(
+			Equal(`{"query":"hello world","source":"web"}`),
+			Equal(`{"source":"web","query":"hello world"}`),
+		))
+	})
+
+	It("parses numeric and boolean args", func() {
+		input := `<|tool_call_start|>[set_volume(level=42, mute=false)]<|tool_call_end|>`
+		results := ParseFunctionCall(input, cfg())
+		Expect(results).To(HaveLen(1))
+		Expect(results[0].Name).To(Equal("set_volume"))
+		// ArgumentRegex always emits string values; the JSON we produce represents
+		// them as strings. A typed parser is a future enhancement (PEG parser).
+		Expect(results[0].Arguments).To(SatisfyAny(
+			Equal(`{"level":"42","mute":"false"}`),
+			Equal(`{"mute":"false","level":"42"}`),
+		))
+	})
+
+	It("parses a no-args call", func() {
+		input := `<|tool_call_start|>[get_time()]<|tool_call_end|>`
+		results := ParseFunctionCall(input, cfg())
+		Expect(results).To(HaveLen(1))
+		Expect(results[0].Name).To(Equal("get_time"))
+		Expect(results[0].Arguments).To(Equal(`{}`))
+	})
+
+	It("ignores surrounding text", func() {
+		input := `Sure, let me check.
+<|tool_call_start|>[get_weather(city="Paris")]<|tool_call_end|>
+Standby.`
+		results := ParseFunctionCall(input, cfg())
+		Expect(results).To(HaveLen(1))
+		Expect(results[0].Name).To(Equal("get_weather"))
+		Expect(results[0].Arguments).To(Equal(`{"city":"Paris"}`))
+	})
+
+	It("returns no results when the markers are absent", func() {
+		input := `Plain text response with no tool call.`
+		results := ParseFunctionCall(input, cfg())
+		Expect(results).To(BeEmpty())
+	})
+
+	It("preserves quoted argument values that contain spaces and equals signs", func() {
+		input := `<|tool_call_start|>[search(query="x = y + 1")]<|tool_call_end|>`
+		results := ParseFunctionCall(input, cfg())
+		Expect(results).To(HaveLen(1))
+		Expect(results[0].Name).To(Equal("search"))
+		Expect(results[0].Arguments).To(Equal(`{"query":"x = y + 1"}`))
+	})
+})

pkg/grpc/backend.go

@@ -82,6 +82,7 @@ type Backend interface {
 
 	AudioTransform(ctx context.Context, in *pb.AudioTransformRequest, opts ...grpc.CallOption) (*pb.AudioTransformResult, error)
 	AudioTransformStream(ctx context.Context, opts ...grpc.CallOption) (AudioTransformStreamClient, error)
+	AudioToAudioStream(ctx context.Context, opts ...grpc.CallOption) (AudioToAudioStreamClient, error)
 
 	ModelMetadata(ctx context.Context, in *pb.ModelOptions, opts ...grpc.CallOption) (*pb.ModelMetadataResponse, error)
 

pkg/grpc/base/base.go

@@ -158,6 +158,11 @@ func (llm *Base) AudioTransformStream(in <-chan *pb.AudioTransformFrameRequest,
 	return fmt.Errorf("unimplemented")
 }
 
+func (llm *Base) AudioToAudioStream(in <-chan *pb.AudioToAudioRequest, out chan<- *pb.AudioToAudioResponse) error {
+	close(out)
+	return fmt.Errorf("unimplemented")
+}
+
 func (llm *Base) StartFineTune(*pb.FineTuneRequest) (*pb.FineTuneJobResult, error) {
 	return nil, fmt.Errorf("unimplemented")
 }

pkg/grpc/client.go

@@ -805,6 +805,67 @@ func (c *Client) AudioTransformStream(ctx context.Context, opts ...grpc.CallOpti
 	}, nil
 }
 
+// AudioToAudioStreamClient is the duplex interface returned by
+// (*Client).AudioToAudioStream. Mirrors AudioTransformStreamClient's
+// shape so realtime-API callers can plug in interchangeable backends.
+type AudioToAudioStreamClient interface {
+	Send(*pb.AudioToAudioRequest) error
+	Recv() (*pb.AudioToAudioResponse, error)
+	CloseSend() error
+	Context() context.Context
+}
+
+type audioToAudioStreamClient struct {
+	pb.Backend_AudioToAudioStreamClient
+	conn   *grpc.ClientConn
+	closer func()
+}
+
+func (s *audioToAudioStreamClient) CloseSend() error {
+	err := s.Backend_AudioToAudioStreamClient.CloseSend()
+	if s.closer != nil {
+		s.closer()
+	}
+	return err
+}
+
+func (c *Client) AudioToAudioStream(ctx context.Context, opts ...grpc.CallOption) (AudioToAudioStreamClient, error) {
+	if !c.parallel {
+		c.opMutex.Lock()
+	}
+	c.setBusy(true)
+	c.wdMark()
+
+	cleanup := func() {
+		c.wdUnMark()
+		c.setBusy(false)
+		if !c.parallel {
+			c.opMutex.Unlock()
+		}
+	}
+
+	conn, err := c.dial()
+	if err != nil {
+		cleanup()
+		return nil, err
+	}
+	client := pb.NewBackendClient(conn)
+	stream, err := client.AudioToAudioStream(ctx, opts...)
+	if err != nil {
+		_ = conn.Close()
+		cleanup()
+		return nil, err
+	}
+	return &audioToAudioStreamClient{
+		Backend_AudioToAudioStreamClient: stream,
+		conn:                             conn,
+		closer: func() {
+			_ = conn.Close()
+			cleanup()
+		},
+	}, nil
+}
+
 func (c *Client) StartFineTune(ctx context.Context, in *pb.FineTuneRequest, opts ...grpc.CallOption) (*pb.FineTuneJobResult, error) {
 	if !c.parallel {
 		c.opMutex.Lock()

pkg/grpc/embed.go

@@ -181,6 +181,31 @@ func (e *embedBackend) AudioTransformStream(ctx context.Context, opts ...grpc.Ca
 	}, nil
 }
 
+func (e *embedBackend) AudioToAudioStream(ctx context.Context, opts ...grpc.CallOption) (AudioToAudioStreamClient, error) {
+	reqs := make(chan *pb.AudioToAudioRequest, 8)
+	resps := make(chan *pb.AudioToAudioResponse, 8)
+	srvDone := make(chan error, 1)
+
+	server := &embedBackendAudioToAudioStream{
+		ctx:   ctx,
+		reqs:  reqs,
+		resps: resps,
+	}
+
+	go func() {
+		err := e.s.AudioToAudioStream(server)
+		close(resps)
+		srvDone <- err
+	}()
+
+	return &embedBackendAudioToAudioStreamClient{
+		ctx:     ctx,
+		reqs:    reqs,
+		resps:   resps,
+		srvDone: srvDone,
+	}, nil
+}
+
 func (e *embedBackend) ModelMetadata(ctx context.Context, in *pb.ModelOptions, opts ...grpc.CallOption) (*pb.ModelMetadataResponse, error) {
 	return e.s.ModelMetadata(ctx, in)
 }
@@ -236,6 +261,8 @@ func (e *embedBackend) Free(ctx context.Context) error {
 
 var _ pb.Backend_AudioTransformStreamServer = new(embedBackendAudioTransformStream)
 var _ AudioTransformStreamClient = new(embedBackendAudioTransformStreamClient)
+var _ pb.Backend_AudioToAudioStreamServer = new(embedBackendAudioToAudioStream)
+var _ AudioToAudioStreamClient = new(embedBackendAudioToAudioStreamClient)
 
 // embedBackendAudioTransformStream is the server side of an in-process bidi
 // stream. The hosted server reads requests from `reqs` (closed by client when
@@ -332,6 +359,99 @@ func (e *embedBackendAudioTransformStreamClient) CloseSend() error {
 
 func (e *embedBackendAudioTransformStreamClient) Context() context.Context { return e.ctx }
 
+// embedBackendAudioToAudioStream is the in-process server-side handle for
+// the bidirectional any-to-any audio RPC. Mirrors embedBackendAudioTransform
+// Stream — the hosted server reads requests from `reqs` (closed by client
+// when done sending) and writes responses to `resps`.
+type embedBackendAudioToAudioStream struct {
+	ctx   context.Context
+	reqs  <-chan *pb.AudioToAudioRequest
+	resps chan<- *pb.AudioToAudioResponse
+}
+
+func (e *embedBackendAudioToAudioStream) Send(resp *pb.AudioToAudioResponse) error {
+	select {
+	case e.resps <- resp:
+		return nil
+	case <-e.ctx.Done():
+		return e.ctx.Err()
+	}
+}
+
+func (e *embedBackendAudioToAudioStream) Recv() (*pb.AudioToAudioRequest, error) {
+	select {
+	case req, ok := <-e.reqs:
+		if !ok {
+			return nil, io.EOF
+		}
+		return req, nil
+	case <-e.ctx.Done():
+		return nil, e.ctx.Err()
+	}
+}
+
+func (e *embedBackendAudioToAudioStream) SetHeader(md metadata.MD) error  { return nil }
+func (e *embedBackendAudioToAudioStream) SendHeader(md metadata.MD) error { return nil }
+func (e *embedBackendAudioToAudioStream) SetTrailer(md metadata.MD)       {}
+func (e *embedBackendAudioToAudioStream) Context() context.Context        { return e.ctx }
+func (e *embedBackendAudioToAudioStream) SendMsg(m any) error {
+	if x, ok := m.(*pb.AudioToAudioResponse); ok {
+		return e.Send(x)
+	}
+	return nil
+}
+func (e *embedBackendAudioToAudioStream) RecvMsg(m any) error { return nil }
+
+type embedBackendAudioToAudioStreamClient struct {
+	ctx       context.Context
+	reqs      chan<- *pb.AudioToAudioRequest
+	resps     <-chan *pb.AudioToAudioResponse
+	srvDone   <-chan error
+	closeOnce bool
+}
+
+func (e *embedBackendAudioToAudioStreamClient) Send(req *pb.AudioToAudioRequest) error {
+	select {
+	case e.reqs <- req:
+		return nil
+	case <-e.ctx.Done():
+		return e.ctx.Err()
+	}
+}
+
+func (e *embedBackendAudioToAudioStreamClient) Recv() (*pb.AudioToAudioResponse, error) {
+	select {
+	case resp, ok := <-e.resps:
+		if !ok {
+			// Server goroutine writes to srvDone immediately after closing
+			// resps; block (cap with ctx) so we don't race past a real error.
+			select {
+			case err := <-e.srvDone:
+				if err != nil {
+					return nil, err
+				}
+			case <-e.ctx.Done():
+				return nil, e.ctx.Err()
+			}
+			return nil, io.EOF
+		}
+		return resp, nil
+	case <-e.ctx.Done():
+		return nil, e.ctx.Err()
+	}
+}
+
+func (e *embedBackendAudioToAudioStreamClient) CloseSend() error {
+	if e.closeOnce {
+		return nil
+	}
+	e.closeOnce = true
+	close(e.reqs)
+	return nil
+}
+
+func (e *embedBackendAudioToAudioStreamClient) Context() context.Context { return e.ctx }
+
 var _ pb.Backend_AudioTranscriptionStreamServer = new(embedBackendAudioTranscriptionStream)
 
 type embedBackendAudioTranscriptionStream struct {

pkg/grpc/interface.go

@@ -45,6 +45,7 @@ type AIModel interface {
 
 	AudioTransform(*pb.AudioTransformRequest) (*pb.AudioTransformResult, error)
 	AudioTransformStream(in <-chan *pb.AudioTransformFrameRequest, out chan<- *pb.AudioTransformFrameResponse) error
+	AudioToAudioStream(in <-chan *pb.AudioToAudioRequest, out chan<- *pb.AudioToAudioResponse) error
 
 	ModelMetadata(*pb.ModelOptions) (*pb.ModelMetadataResponse, error)
 

pkg/grpc/server.go

@@ -487,6 +487,66 @@ func (s *server) AudioTransformStream(stream pb.Backend_AudioTransformStreamServ
 	return recvErr
 }
 
+// AudioToAudioStream is the bidirectional any-to-any S2S handler. The
+// shape mirrors AudioTransformStream exactly (recv → in chan, out chan →
+// send) so backends can implement either via the same goroutine idiom.
+func (s *server) AudioToAudioStream(stream pb.Backend_AudioToAudioStreamServer) error {
+	if s.llm.Locking() {
+		s.llm.Lock()
+		defer s.llm.Unlock()
+	}
+
+	in := make(chan *pb.AudioToAudioRequest, 8)
+	out := make(chan *pb.AudioToAudioResponse, 8)
+
+	recvErrCh := make(chan error, 1)
+	go func() {
+		defer close(in)
+		for {
+			req, err := stream.Recv()
+			if err != nil {
+				if errors.Is(err, io.EOF) {
+					recvErrCh <- nil
+					return
+				}
+				recvErrCh <- err
+				return
+			}
+			select {
+			case in <- req:
+			case <-stream.Context().Done():
+				recvErrCh <- stream.Context().Err()
+				return
+			}
+		}
+	}()
+
+	sendDone := make(chan error, 1)
+	go func() {
+		for resp := range out {
+			if err := stream.Send(resp); err != nil {
+				sendDone <- err
+				for range out {
+				}
+				return
+			}
+		}
+		sendDone <- nil
+	}()
+
+	backendErr := s.llm.AudioToAudioStream(in, out)
+	sendErr := <-sendDone
+	recvErr := <-recvErrCh
+
+	if backendErr != nil {
+		return backendErr
+	}
+	if sendErr != nil {
+		return sendErr
+	}
+	return recvErr
+}
+
 func (s *server) StartFineTune(ctx context.Context, in *pb.FineTuneRequest) (*pb.FineTuneJobResult, error) {
 	if s.llm.Locking() {
 		s.llm.Lock()

pkg/utils/untar.go

@@ -1,9 +1,13 @@
 package utils
 
 import (
+	"archive/tar"
 	"fmt"
 	"os"
+	"path/filepath"
+	"strings"
 
+	"github.com/klauspost/compress/zip"
 	"github.com/mholt/archiver/v3"
 )
 
@@ -54,7 +58,15 @@ func ExtractArchive(archive, dst string) error {
 		v.Tar = mytar
 	}
 
+	extractRoot, err := filepath.Abs(dst)
+	if err != nil {
+		return err
+	}
+
 	err = archiver.Walk(archive, func(f archiver.File) error {
+		if err := validateArchiveMemberPath(extractRoot, archiveMemberName(f)); err != nil {
+			return err
+		}
 		if f.FileInfo.Mode()&os.ModeSymlink != 0 {
 			return fmt.Errorf("archive contains a symlink")
 		}
@@ -67,3 +79,41 @@ func ExtractArchive(archive, dst string) error {
 
 	return un.Unarchive(archive, dst)
 }
+
+func archiveMemberName(f archiver.File) string {
+	switch h := f.Header.(type) {
+	case tar.Header:
+		return h.Name
+	case *tar.Header:
+		return h.Name
+	case zip.FileHeader:
+		return h.Name
+	case *zip.FileHeader:
+		return h.Name
+	default:
+		return f.Name()
+	}
+}
+
+func validateArchiveMemberPath(root, name string) error {
+	if name == "" {
+		return fmt.Errorf("archive contains an empty path")
+	}
+
+	normalizedName := filepath.FromSlash(strings.ReplaceAll(name, "\\", "/"))
+	cleanedName := filepath.Clean(normalizedName)
+	if filepath.IsAbs(cleanedName) || cleanedName == ".." || strings.HasPrefix(cleanedName, ".."+string(os.PathSeparator)) {
+		return fmt.Errorf("archive contains an unsafe path: %s", name)
+	}
+
+	targetPath := filepath.Join(root, cleanedName)
+	relativePath, err := filepath.Rel(root, targetPath)
+	if err != nil {
+		return err
+	}
+	if relativePath == ".." || strings.HasPrefix(relativePath, ".."+string(os.PathSeparator)) || filepath.IsAbs(relativePath) {
+		return fmt.Errorf("archive contains an unsafe path: %s", name)
+	}
+
+	return nil
+}

pkg/utils/untar_test.go

@@ -0,0 +1,128 @@
+package utils_test
+
+import (
+	"archive/tar"
+	"archive/zip"
+	"os"
+	"path/filepath"
+
+	. "github.com/mudler/LocalAI/pkg/utils"
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+var _ = Describe("utils/archive tests", func() {
+	It("extracts regular nested zip members", func() {
+		tmpDir := GinkgoT().TempDir()
+		archivePath := filepath.Join(tmpDir, "model.zip")
+		extractPath := filepath.Join(tmpDir, "models")
+
+		Expect(writeZipArchive(archivePath, map[string]string{
+			"nested/model.yaml": "name: test",
+		})).To(Succeed())
+
+		Expect(ExtractArchive(archivePath, extractPath)).To(Succeed())
+
+		extracted, err := os.ReadFile(filepath.Join(extractPath, "nested", "model.yaml"))
+		Expect(err).ToNot(HaveOccurred())
+		Expect(string(extracted)).To(Equal("name: test"))
+	})
+
+	It("rejects zip members that escape the destination", func() {
+		tmpDir := GinkgoT().TempDir()
+		archivePath := filepath.Join(tmpDir, "model.zip")
+		extractPath := filepath.Join(tmpDir, "models")
+
+		Expect(writeZipArchive(archivePath, map[string]string{
+			"../escaped.txt": "escaped",
+		})).To(Succeed())
+
+		err := ExtractArchive(archivePath, extractPath)
+
+		Expect(err).To(HaveOccurred())
+		Expect(err.Error()).To(ContainSubstring("unsafe path"))
+		Expect(filepath.Join(tmpDir, "escaped.txt")).ToNot(BeAnExistingFile())
+	})
+
+	It("rejects tar members that escape the destination", func() {
+		tmpDir := GinkgoT().TempDir()
+		archivePath := filepath.Join(tmpDir, "model.tar")
+		extractPath := filepath.Join(tmpDir, "models")
+
+		Expect(writeTarArchive(archivePath, map[string]string{
+			"../escaped.txt": "escaped",
+		})).To(Succeed())
+
+		err := ExtractArchive(archivePath, extractPath)
+
+		Expect(err).To(HaveOccurred())
+		Expect(err.Error()).To(ContainSubstring("unsafe path"))
+		Expect(filepath.Join(tmpDir, "escaped.txt")).ToNot(BeAnExistingFile())
+	})
+})
+
+func writeZipArchive(path string, files map[string]string) (err error) {
+	out, err := os.Create(path)
+	if err != nil {
+		return err
+	}
+	defer func() {
+		if closeErr := out.Close(); err == nil {
+			err = closeErr
+		}
+	}()
+
+	writer := zip.NewWriter(out)
+	defer func() {
+		if closeErr := writer.Close(); err == nil {
+			err = closeErr
+		}
+	}()
+
+	for name, contents := range files {
+		fileWriter, err := writer.Create(name)
+		if err != nil {
+			return err
+		}
+		if _, err := fileWriter.Write([]byte(contents)); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func writeTarArchive(path string, files map[string]string) (err error) {
+	out, err := os.Create(path)
+	if err != nil {
+		return err
+	}
+	defer func() {
+		if closeErr := out.Close(); err == nil {
+			err = closeErr
+		}
+	}()
+
+	writer := tar.NewWriter(out)
+	defer func() {
+		if closeErr := writer.Close(); err == nil {
+			err = closeErr
+		}
+	}()
+
+	for name, contents := range files {
+		data := []byte(contents)
+		if err := writer.WriteHeader(&tar.Header{
+			Name: name,
+			Mode: 0o600,
+			Size: int64(len(data)),
+		}); err != nil {
+			return err
+		}
+		if _, err := writer.Write(data); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}

pkg/xsysinfo/gpu.go

@@ -827,10 +827,12 @@ func getVulkanGPUMemory() []GPUMemoryInfo {
 }
 
 type vulkanGPUTextInfo struct {
-	index      int
-	name       string
-	deviceType string
-	totalVRAM  uint64
+	index        int
+	name         string
+	deviceType   string
+	totalVRAM    uint64
+	budgetVRAM   uint64
+	usageVRAM    uint64
 }
 
 func parseVulkanGPUMemoryText(r io.Reader) []GPUMemoryInfo {
@@ -841,13 +843,19 @@ func parseVulkanGPUMemoryText(r io.Reader) []GPUMemoryInfo {
 	inMemoryHeaps := false
 	inHeap := false
 	heapSize := uint64(0)
+	heapBudget := uint64(0)
+	heapUsage := uint64(0)
 	heapDeviceLocal := false
 
 	flushHeap := func() {
 		if current != nil && inHeap && heapDeviceLocal {
 			current.totalVRAM += heapSize
+			current.usageVRAM += heapUsage
+			current.budgetVRAM += heapBudget
 		}
 		heapSize = 0
+		heapBudget = 0
+		heapUsage = 0
 		heapDeviceLocal = false
 		inHeap = false
 	}
@@ -857,14 +865,25 @@ func parseVulkanGPUMemoryText(r io.Reader) []GPUMemoryInfo {
 			return
 		}
 
+		if current.usageVRAM == 0 && current.budgetVRAM != 0 {
+			current.usageVRAM = current.totalVRAM - current.budgetVRAM
+		} else if current.usageVRAM != 0 && current.budgetVRAM == 0 {
+			current.budgetVRAM = current.totalVRAM - current.usageVRAM
+		} else if current.usageVRAM == 0 && current.budgetVRAM == 0 {
+			current.usageVRAM  = 0
+			current.budgetVRAM = current.totalVRAM
+		}
+
+		usagePercent := float64(current.usageVRAM) / float64(current.totalVRAM) * float64(100.0)
+
 		gpus = append(gpus, GPUMemoryInfo{
 			Index:        current.index,
 			Name:         current.name,
 			Vendor:       VendorVulkan,
 			TotalVRAM:    current.totalVRAM,
-			UsedVRAM:     0, // Vulkan heap size is capacity, not real-time usage.
-			FreeVRAM:     current.totalVRAM,
-			UsagePercent: 0,
+			UsedVRAM:     current.usageVRAM,
+			FreeVRAM:     current.budgetVRAM,
+			UsagePercent: usagePercent,
 		})
 	}
 
@@ -942,6 +961,20 @@ func parseVulkanGPUMemoryText(r io.Reader) []GPUMemoryInfo {
 			continue
 		}
 
+		if strings.HasPrefix(line, "budget") {
+			if budget, ok := parseVulkanUintValue(line); ok {
+				heapBudget = budget
+			}
+			continue
+		}
+
+		if strings.HasPrefix(line, "usage") {
+			if usage, ok := parseVulkanUintValue(line); ok {
+				heapUsage = usage
+			}
+			continue
+		}
+
 		if strings.Contains(line, "MEMORY_HEAP_DEVICE_LOCAL_BIT") {
 			heapDeviceLocal = true
 		}