wire to grpc

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

.env

@@ -76,7 +76,7 @@
 
 ### Define a list of GRPC Servers for llama-cpp workers to distribute the load
 # https://github.com/ggerganov/llama.cpp/pull/6829
-# https://github.com/ggerganov/llama.cpp/blob/master/tools/rpc/README.md
+# https://github.com/ggerganov/llama.cpp/blob/master/examples/rpc/README.md
 # LLAMACPP_GRPC_SERVERS=""
 
 ### Enable to run parallel requests

.github/workflows/bump_deps.yaml

@@ -12,7 +12,7 @@ jobs:
           - repository: "ggml-org/llama.cpp"
             variable: "CPPLLAMA_VERSION"
             branch: "master"
-          - repository: "ggml-org/whisper.cpp"
+          - repository: "ggerganov/whisper.cpp"
             variable: "WHISPER_CPP_VERSION"
             branch: "master"
           - repository: "PABannier/bark.cpp"

.github/workflows/dependabot_auto.yml

@@ -14,7 +14,7 @@ jobs:
     steps:
       - name: Dependabot metadata
         id: metadata
-        uses: dependabot/fetch-metadata@v2.4.0
+        uses: dependabot/fetch-metadata@v2.3.0
         with:
           github-token: "${{ secrets.GITHUB_TOKEN }}"
           skip-commit-verification: true

.github/workflows/deploy-explorer.yaml

@@ -42,7 +42,7 @@ jobs:
             script: |
                 sudo rm -rf local-ai/ || true
       - name: copy file via ssh
-        uses: appleboy/scp-action@v1.0.0
+        uses: appleboy/scp-action@v0.1.7
         with:
             host: ${{ secrets.EXPLORER_SSH_HOST }}
             username: ${{ secrets.EXPLORER_SSH_USERNAME }}

.github/workflows/image-pr.yml

@@ -33,7 +33,6 @@ jobs:
       # Pushing with all jobs in parallel
       # eats the bandwidth of all the nodes
       max-parallel: ${{ github.event_name != 'pull_request' && 4 || 8 }}
-      fail-fast: false
       matrix:
         include:
           # This is basically covered by the AIO test
@@ -57,35 +56,26 @@ jobs:
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
             makeflags: "--jobs=3 --output-sync=target"
-          - build-type: 'hipblas'
-            platforms: 'linux/amd64'
-            tag-latest: 'false'
-            tag-suffix: '-hipblas'
-            ffmpeg: 'false'
-            image-type: 'extras'
-            base-image: "rocm/dev-ubuntu-22.04:6.1"
-            grpc-base-image: "ubuntu:22.04"
-            runs-on: 'arc-runner-set'
-            makeflags: "--jobs=3 --output-sync=target"
-          - build-type: 'sycl_f16'
-            platforms: 'linux/amd64'
-            tag-latest: 'false'
-            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
-            grpc-base-image: "ubuntu:22.04"
-            tag-suffix: 'sycl-f16-ffmpeg'
-            ffmpeg: 'true'
-            image-type: 'extras'
-            runs-on: 'arc-runner-set'
-            makeflags: "--jobs=3 --output-sync=target"
-          - build-type: 'vulkan'
-            platforms: 'linux/amd64'
-            tag-latest: 'false'
-            tag-suffix: '-vulkan-ffmpeg-core'
-            ffmpeg: 'true'
-            image-type: 'core'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            makeflags: "--jobs=4 --output-sync=target"
+          # - build-type: 'hipblas'
+          #   platforms: 'linux/amd64'
+          #   tag-latest: 'false'
+          #   tag-suffix: '-hipblas'
+          #   ffmpeg: 'false'
+          #   image-type: 'extras'
+          #   base-image: "rocm/dev-ubuntu-22.04:6.1"
+          #   grpc-base-image: "ubuntu:22.04"
+          #   runs-on: 'arc-runner-set'
+          #   makeflags: "--jobs=3 --output-sync=target"
+          # - build-type: 'sycl_f16'
+          #   platforms: 'linux/amd64'
+          #   tag-latest: 'false'
+          #   base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+          #   grpc-base-image: "ubuntu:22.04"
+          #   tag-suffix: 'sycl-f16-ffmpeg'
+          #   ffmpeg: 'true'
+          #   image-type: 'extras'
+          #   runs-on: 'arc-runner-set'
+          #   makeflags: "--jobs=3 --output-sync=target"
   # core-image-build:
   #   uses: ./.github/workflows/image_build.yml
   #   with:

.github/workflows/image.yml

@@ -45,13 +45,13 @@ jobs:
           - build-type: 'hipblas'
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-hipblas-extras'
+            tag-suffix: '-hipblas-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             aio: "-aio-gpu-hipblas"
             base-image: "rocm/dev-ubuntu-22.04:6.1"
             grpc-base-image: "ubuntu:22.04"
-            latest-image: 'latest-gpu-hipblas-extras'
+            latest-image: 'latest-gpu-hipblas'
             latest-image-aio: 'latest-aio-gpu-hipblas'
             runs-on: 'arc-runner-set'
             makeflags: "--jobs=3 --output-sync=target"
@@ -59,13 +59,33 @@ jobs:
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-hipblas'
+            ffmpeg: 'false'
+            image-type: 'extras'
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: 'hipblas'
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-hipblas-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             base-image: "rocm/dev-ubuntu-22.04:6.1"
             grpc-base-image: "ubuntu:22.04"
             runs-on: 'arc-runner-set'
             makeflags: "--jobs=3 --output-sync=target"
-            latest-image: 'latest-gpu-hipblas'
+            latest-image: 'latest-gpu-hipblas-core'
+          - build-type: 'hipblas'
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-hipblas-core'
+            ffmpeg: 'false'
+            image-type: 'core'
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
   self-hosted-jobs:
     uses: ./.github/workflows/image_build.yml
     with:
@@ -95,58 +115,110 @@ jobs:
       max-parallel: ${{ github.event_name != 'pull_request' && 5 || 8 }}
       matrix:
         include:
+          # Extra images
+          - build-type: ''
+            #platforms: 'linux/amd64,linux/arm64'
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: ''
+            ffmpeg: ''
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
+            base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: ''
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-ffmpeg'
+            ffmpeg: 'true'
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
+            base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
             cuda-minor-version: "7"
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            tag-suffix: '-cublas-cuda11-extras'
-            ffmpeg: 'true'
+            tag-suffix: '-cublas-cuda11'
+            ffmpeg: ''
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
-            aio: "-aio-gpu-nvidia-cuda-11"
-            latest-image: 'latest-gpu-nvidia-cuda-11-extras'
-            latest-image-aio: 'latest-aio-gpu-nvidia-cuda-11'
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            tag-suffix: '-cublas-cuda12-extras'
+            tag-suffix: '-cublas-cuda12'
+            ffmpeg: ''
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
+            base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: 'cublas'
+            cuda-major-version: "11"
+            cuda-minor-version: "7"
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-cublas-cuda11-ffmpeg'
+            ffmpeg: 'true'
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
+            base-image: "ubuntu:22.04"
+            aio: "-aio-gpu-nvidia-cuda-11"
+            latest-image: 'latest-gpu-nvidia-cuda-11'
+            latest-image-aio: 'latest-aio-gpu-nvidia-cuda-11'
+            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: 'cublas'
+            cuda-major-version: "12"
+            cuda-minor-version: "0"
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-cublas-cuda12-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
             aio: "-aio-gpu-nvidia-cuda-12"
-            latest-image: 'latest-gpu-nvidia-cuda-12-extras'
+            latest-image: 'latest-gpu-nvidia-cuda-12'
             latest-image-aio: 'latest-aio-gpu-nvidia-cuda-12'
             makeflags: "--jobs=3 --output-sync=target"
+          - build-type: ''
+            #platforms: 'linux/amd64,linux/arm64'
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: ''
+            ffmpeg: ''
+            image-type: 'extras'
+            base-image: "ubuntu:22.04"
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'sycl_f16'
             platforms: 'linux/amd64'
-            tag-latest: 'false'
+            tag-latest: 'auto'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             grpc-base-image: "ubuntu:22.04"
-            tag-suffix: '-sycl-f16-extras'
+            tag-suffix: '-sycl-f16-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             aio: "-aio-gpu-intel-f16"
-            latest-image: 'latest-gpu-intel-f16-extras'
+            latest-image: 'latest-gpu-intel-f16'
             latest-image-aio: 'latest-aio-gpu-intel-f16'
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'sycl_f32'
             platforms: 'linux/amd64'
-            tag-latest: 'false'
+            tag-latest: 'auto'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             grpc-base-image: "ubuntu:22.04"
-            tag-suffix: '-sycl-f32-extras'
+            tag-suffix: '-sycl-f32-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             aio: "-aio-gpu-intel-f32"
-            latest-image: 'latest-gpu-intel-f32-extras'
+            latest-image: 'latest-gpu-intel-f32'
             latest-image-aio: 'latest-aio-gpu-intel-f32'
             makeflags: "--jobs=3 --output-sync=target"
           # Core images
@@ -155,23 +227,43 @@ jobs:
             tag-latest: 'false'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             grpc-base-image: "ubuntu:22.04"
-            tag-suffix: '-sycl-f16'
-            ffmpeg: 'true'
+            tag-suffix: '-sycl-f16-core'
+            ffmpeg: 'false'
             image-type: 'core'
             runs-on: 'arc-runner-set'
             makeflags: "--jobs=3 --output-sync=target"
-            latest-image: 'latest-gpu-intel-f16'
           - build-type: 'sycl_f32'
             platforms: 'linux/amd64'
             tag-latest: 'false'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             grpc-base-image: "ubuntu:22.04"
-            tag-suffix: '-sycl-f32'
+            tag-suffix: '-sycl-f32-core'
+            ffmpeg: 'false'
+            image-type: 'core'
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
+          - build-type: 'sycl_f16'
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
+            tag-suffix: '-sycl-f16-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             runs-on: 'arc-runner-set'
             makeflags: "--jobs=3 --output-sync=target"
-            latest-image: 'latest-gpu-intel-f32'
+            latest-image: 'latest-gpu-intel-f16-core'
+          - build-type: 'sycl_f32'
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
+            tag-suffix: '-sycl-f32-ffmpeg-core'
+            ffmpeg: 'true'
+            image-type: 'core'
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
+            latest-image: 'latest-gpu-intel-f32-core'
 
   core-image-build:
     uses: ./.github/workflows/image_build.yml
@@ -204,7 +296,7 @@ jobs:
           - build-type: ''
             platforms: 'linux/amd64,linux/arm64'
             tag-latest: 'auto'
-            tag-suffix: ''
+            tag-suffix: '-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             base-image: "ubuntu:22.04"
@@ -219,38 +311,62 @@ jobs:
             cuda-minor-version: "7"
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            tag-suffix: '-cublas-cuda11'
-            ffmpeg: 'true'
+            tag-suffix: '-cublas-cuda11-core'
+            ffmpeg: ''
             image-type: 'core'
-            runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            runs-on: 'arc-runner-set'
             makeflags: "--jobs=4 --output-sync=target"
             skip-drivers: 'false'
-            latest-image: 'latest-gpu-nvidia-cuda-12'
           - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            tag-suffix: '-cublas-cuda12'
+            tag-suffix: '-cublas-cuda12-core'
+            ffmpeg: ''
+            image-type: 'core'
+            base-image: "ubuntu:22.04"
+            runs-on: 'arc-runner-set'
+            makeflags: "--jobs=4 --output-sync=target"
+            skip-drivers: 'false'
+          - build-type: 'cublas'
+            cuda-major-version: "11"
+            cuda-minor-version: "7"
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-cublas-cuda11-ffmpeg-core'
+            ffmpeg: 'true'
+            image-type: 'core'
+            runs-on: 'arc-runner-set'
+            base-image: "ubuntu:22.04"
+            makeflags: "--jobs=4 --output-sync=target"
+            skip-drivers: 'false'
+            latest-image: 'latest-gpu-nvidia-cuda-12-core'
+          - build-type: 'cublas'
+            cuda-major-version: "12"
+            cuda-minor-version: "0"
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-cublas-cuda12-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             makeflags: "--jobs=4 --output-sync=target"
-            latest-image: 'latest-gpu-nvidia-cuda-12'
+            latest-image: 'latest-gpu-nvidia-cuda-12-core'
           - build-type: 'vulkan'
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            tag-suffix: '-vulkan'
+            tag-suffix: '-vulkan-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             makeflags: "--jobs=4 --output-sync=target"
-            latest-image: 'latest-gpu-vulkan'
+            latest-image: 'latest-gpu-vulkan-core'
   gh-runner:
     uses: ./.github/workflows/image_build.yml
     with:
@@ -283,8 +399,8 @@ jobs:
             cuda-minor-version: "0"
             platforms: 'linux/arm64'
             tag-latest: 'false'
-            tag-suffix: '-nvidia-l4t-arm64'
-            latest-image: 'latest-nvidia-l4t-arm64'
+            tag-suffix: '-nvidia-l4t-arm64-core'
+            latest-image: 'latest-nvidia-l4t-arm64-core'
             ffmpeg: 'true'
             image-type: 'core'
             base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"

.github/workflows/notify-models.yaml

@@ -79,7 +79,7 @@ jobs:
         args: ${{ steps.summarize.outputs.message }}
     - name: Setup tmate session if fails
       if: ${{ failure() }}
-      uses: mxschmitt/action-tmate@v3.22
+      uses: mxschmitt/action-tmate@v3.19
       with:
         detached: true
         connect-timeout-seconds: 180
@@ -161,7 +161,7 @@ jobs:
         TWITTER_ACCESS_TOKEN_SECRET: ${{ secrets.TWITTER_ACCESS_TOKEN_SECRET }}
     - name: Setup tmate session if fails
       if: ${{ failure() }}
-      uses: mxschmitt/action-tmate@v3.22
+      uses: mxschmitt/action-tmate@v3.19
       with:
         detached: true
         connect-timeout-seconds: 180

.github/workflows/release.yaml

@@ -36,7 +36,6 @@ jobs:
           sudo apt-get update
           sudo apt-get install build-essential ffmpeg protobuf-compiler ccache upx-ucl gawk
           sudo apt-get install -qy binutils-aarch64-linux-gnu gcc-aarch64-linux-gnu g++-aarch64-linux-gnu libgmock-dev
-          make install-go-tools
       - name: Install CUDA Dependencies
         run: |
           curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/cross-linux-aarch64/cuda-keyring_1.1-1_all.deb
@@ -124,7 +123,7 @@ jobs:
             release/*
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.22
+        uses: mxschmitt/action-tmate@v3.19
         with:
           detached: true
           connect-timeout-seconds: 180
@@ -152,7 +151,6 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install -y wget curl build-essential ffmpeg protobuf-compiler ccache upx-ucl gawk cmake libgmock-dev
-          make install-go-tools
       - name: Intel Dependencies
         run: |
           wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB | gpg --dearmor | sudo tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null
@@ -234,7 +232,7 @@ jobs:
             release/*
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.22
+        uses: mxschmitt/action-tmate@v3.19
         with:
           detached: true
           connect-timeout-seconds: 180
@@ -255,7 +253,8 @@ jobs:
       - name: Dependencies
         run: |
           brew install protobuf grpc
-          make install-go-tools
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@8ba23be9613c672d40ae261d2a1335d639bdd59b
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.0
       - name: Build
         id: build
         run: |
@@ -276,7 +275,7 @@ jobs:
             release/*
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.22
+        uses: mxschmitt/action-tmate@v3.19
         with:
           detached: true
           connect-timeout-seconds: 180
@@ -296,7 +295,8 @@ jobs:
       - name: Dependencies
         run: |
           brew install protobuf grpc libomp llvm
-          make install-go-tools
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2
       - name: Build
         id: build
         run: |
@@ -317,7 +317,7 @@ jobs:
             release/*
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.22
+        uses: mxschmitt/action-tmate@v3.19
         with:
           detached: true
           connect-timeout-seconds: 180

.github/workflows/secscan.yaml

@@ -18,7 +18,7 @@ jobs:
         if: ${{ github.actor != 'dependabot[bot]' }}
       - name: Run Gosec Security Scanner
         if: ${{ github.actor != 'dependabot[bot]' }}
-        uses: securego/gosec@v2.22.4
+        uses: securego/gosec@v2.22.3
         with:
           # we let the report trigger content trigger a failure using the GitHub Security features.
           args: '-no-fail -fmt sarif -out results.sarif ./...'

.github/workflows/test-extra.yml

@@ -78,26 +78,6 @@ jobs:
           make --jobs=5 --output-sync=target -C backend/python/diffusers
           make --jobs=5 --output-sync=target -C backend/python/diffusers test
 
-  #tests-vllm:
-  #  runs-on: ubuntu-latest
-  #  steps:
-  #    - name: Clone
-  #      uses: actions/checkout@v4
-  #      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
-  #        sudo apt-get install -y libopencv-dev
-  #        # Install UV
-  #        curl -LsSf https://astral.sh/uv/install.sh | sh
-  #        pip install --user --no-cache-dir grpcio-tools==1.64.1
-  #    - name: Test vllm backend
-  #      run: |
-  #        make --jobs=5 --output-sync=target -C backend/python/vllm
-  #        make --jobs=5 --output-sync=target -C backend/python/vllm test
   # tests-transformers-musicgen:
   #   runs-on: ubuntu-latest
   #   steps:

.github/workflows/test.yml

@@ -71,7 +71,7 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install build-essential ccache upx-ucl curl ffmpeg
-          sudo apt-get install -y libgmock-dev clang
+          sudo apt-get install -y libgmock-dev
           curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
              sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
@@ -96,7 +96,6 @@ jobs:
 
           go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2
           go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
-          go install github.com/GeertJohan/go.rice/rice@latest
 
           # The python3-grpc-tools package in 22.04 is too old
           pip install --user grpcio-tools
@@ -131,7 +130,7 @@ jobs:
           PATH="$PATH:/root/go/bin" GO_TAGS="tts" make --jobs 5 --output-sync=target test
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.22
+        uses: mxschmitt/action-tmate@v3.19
         with:
           detached: true
           connect-timeout-seconds: 180
@@ -184,7 +183,6 @@ jobs:
           rm protoc.zip
           go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2
           go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
-          go install github.com/GeertJohan/go.rice/rice@latest
           PATH="$PATH:$HOME/go/bin" make protogen-go
       - name: Build images
         run: |
@@ -196,7 +194,7 @@ jobs:
             make run-e2e-aio
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.22
+        uses: mxschmitt/action-tmate@v3.19
         with:
           detached: true
           connect-timeout-seconds: 180
@@ -224,7 +222,6 @@ jobs:
         run: |
           brew install protobuf grpc make protoc-gen-go protoc-gen-go-grpc libomp llvm
           pip install --user --no-cache-dir grpcio-tools
-          go install github.com/GeertJohan/go.rice/rice@latest
       - name: Test
         run: |
           export C_INCLUDE_PATH=/usr/local/include
@@ -235,7 +232,7 @@ jobs:
           BUILD_TYPE="GITHUB_CI_HAS_BROKEN_METAL" CMAKE_ARGS="-DGGML_F16C=OFF -DGGML_AVX512=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF" make --jobs 4 --output-sync=target test
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.22
+        uses: mxschmitt/action-tmate@v3.19
         with:
           detached: true
           connect-timeout-seconds: 180

Dockerfile

@@ -46,10 +46,9 @@ EOT
 RUN curl -L -s https://go.dev/dl/go${GO_VERSION}.linux-${TARGETARCH}.tar.gz | tar -C /usr/local -xz
 ENV PATH=$PATH:/root/go/bin:/usr/local/go/bin
 
-# Install grpc compilers and rice
+# Install grpc compilers
 RUN go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2 && \
-    go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af && \
-    go install github.com/GeertJohan/go.rice/rice@latest
+    go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
 
 COPY --chmod=644 custom-ca-certs/* /usr/local/share/ca-certificates/
 RUN update-ca-certificates
@@ -301,9 +300,10 @@ COPY .git .
 RUN make prepare
 
 ## Build the binary
-## If we're on arm64 AND using cublas/hipblas, skip some of the llama-compat backends to save space
-## Otherwise just run the normal build
-RUN if [ "${TARGETARCH}" = "arm64" ] || [ "${BUILD_TYPE}" = "hipblas" ]; then \
+## If it's CUDA or hipblas, we want to skip some of the llama-compat backends to save space
+## We only leave the most CPU-optimized variant and the fallback for the cublas/hipblas build
+## (both will use CUDA or hipblas for the actual computation)
+RUN if [ "${BUILD_TYPE}" = "cublas" ] || [ "${BUILD_TYPE}" = "hipblas" ]; then \
         SKIP_GRPC_BACKEND="backend-assets/grpc/llama-cpp-avx512 backend-assets/grpc/llama-cpp-avx backend-assets/grpc/llama-cpp-avx2" make build; \
     else \
         make build; \

Makefile

@@ -6,11 +6,11 @@ BINARY_NAME=local-ai
 DETECT_LIBS?=true
 
 # llama.cpp versions
-CPPLLAMA_VERSION?=e5c834f718a32b7584f142799bbf508fddb9021c
+CPPLLAMA_VERSION?=6408210082cc0a61b992b487be7e2ff2efbb9e36
 
 # whisper.cpp version
-WHISPER_REPO?=https://github.com/ggml-org/whisper.cpp
-WHISPER_CPP_VERSION?=e41bc5c61ae66af6be2bd7011769bb821a83e8ae
+WHISPER_REPO?=https://github.com/ggerganov/whisper.cpp
+WHISPER_CPP_VERSION?=6266a9f9e56a5b925e9892acf650f3eb1245814d
 
 # go-piper version
 PIPER_REPO?=https://github.com/mudler/go-piper
@@ -24,21 +24,14 @@ BARKCPP_VERSION?=v1.0.0
 STABLEDIFFUSION_GGML_REPO?=https://github.com/richiejp/stable-diffusion.cpp
 STABLEDIFFUSION_GGML_VERSION?=53e3b17eb3d0b5760ced06a1f98320b68b34aaae
 
-# ONEAPI variables for SYCL
-export ONEAPI_VARS?=/opt/intel/oneapi/setvars.sh
-
 ONNX_VERSION?=1.20.0
 ONNX_ARCH?=x64
 ONNX_OS?=linux
 
 export BUILD_TYPE?=
 export STABLE_BUILD_TYPE?=$(BUILD_TYPE)
-export CMAKE_ARGS?=-DBUILD_SHARED_LIBS=OFF
-export WHISPER_CMAKE_ARGS?=-DBUILD_SHARED_LIBS=OFF
+export CMAKE_ARGS?=
 export BACKEND_LIBS?=
-export WHISPER_DIR=$(abspath ./sources/whisper.cpp)
-export WHISPER_INCLUDE_PATH=$(WHISPER_DIR)/include:$(WHISPER_DIR)/ggml/include
-export WHISPER_LIBRARY_PATH=$(WHISPER_DIR)/build/src/:$(WHISPER_DIR)/build/ggml/src
 
 CGO_LDFLAGS?=
 CGO_LDFLAGS_WHISPER?=
@@ -88,7 +81,6 @@ endif
 # IF native is false, we add -DGGML_NATIVE=OFF to CMAKE_ARGS
 ifeq ($(NATIVE),false)
 	CMAKE_ARGS+=-DGGML_NATIVE=OFF
-	WHISPER_CMAKE_ARGS+=-DGGML_NATIVE=OFF
 endif
 
 # Detect if we are running on arm64
@@ -116,31 +108,13 @@ ifeq ($(OS),Darwin)
 	# disable metal if on Darwin and any other value is explicitly passed.
 	else ifneq ($(BUILD_TYPE),metal)
 		CMAKE_ARGS+=-DGGML_METAL=OFF
-		WHISPER_CMAKE_ARGS+=-DGGML_METAL=OFF
 		export GGML_NO_ACCELERATE=1
 		export GGML_NO_METAL=1
-		GO_LDFLAGS_WHISPER+=-lggml-blas
-		export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-blas
 	endif
 
 	ifeq ($(BUILD_TYPE),metal)
+#			-lcblas 	removed: it seems to always be listed as a duplicate flag.
 		CGO_LDFLAGS += -framework Accelerate
-		CGO_LDFLAGS_WHISPER+=-lggml-metal -lggml-blas
-		CMAKE_ARGS+=-DGGML_METAL=ON
-		CMAKE_ARGS+=-DGGML_METAL_USE_BF16=ON
-		CMAKE_ARGS+=-DGGML_METAL_EMBED_LIBRARY=ON
-		CMAKE_ARGS+=-DGGML_OPENMP=OFF
-		WHISPER_CMAKE_ARGS+=-DGGML_METAL=ON
-		WHISPER_CMAKE_ARGS+=-DGGML_METAL_USE_BF16=ON
-		WHISPER_CMAKE_ARGS+=-DGGML_METAL_EMBED_LIBRARY=ON
-		WHISPER_CMAKE_ARGS+=-DWHISPER_BUILD_EXAMPLES=OFF
-		WHISPER_CMAKE_ARGS+=-DWHISPER_BUILD_TESTS=OFF
-		WHISPER_CMAKE_ARGS+=-DWHISPER_BUILD_SERVER=OFF
-		WHISPER_CMAKE_ARGS+=-DGGML_OPENMP=OFF
-		export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-metal/:$(WHISPER_DIR)/build/ggml/src/ggml-blas
-	else
-		CGO_LDFLAGS_WHISPER+=-lggml-blas
-		export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-blas
 	endif
 else
 CGO_LDFLAGS_WHISPER+=-lgomp
@@ -152,29 +126,21 @@ ifeq ($(BUILD_TYPE),openblas)
 endif
 
 ifeq ($(BUILD_TYPE),cublas)
-	CGO_LDFLAGS+=-lcublas -lcudart -L$(CUDA_LIBPATH) -L$(CUDA_LIBPATH)/stubs/ -lcuda
+	CGO_LDFLAGS+=-lcublas -lcudart -L$(CUDA_LIBPATH)
 	export GGML_CUDA=1
-	CMAKE_ARGS+=-DGGML_CUDA=ON
-	WHISPER_CMAKE_ARGS+=-DGGML_CUDA=ON
-	CGO_LDFLAGS_WHISPER+=-lcufft -lggml-cuda
-	export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-cuda/
+	CGO_LDFLAGS_WHISPER+=-L$(CUDA_LIBPATH)/stubs/ -lcuda -lcufft
 endif
 
 ifeq ($(BUILD_TYPE),vulkan)
 	CMAKE_ARGS+=-DGGML_VULKAN=1
-	WHISPER_CMAKE_ARGS+=-DGGML_VULKAN=1
-	CGO_LDFLAGS_WHISPER+=-lggml-vulkan -lvulkan
-	export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-vulkan/
 endif
 
 ifneq (,$(findstring sycl,$(BUILD_TYPE)))
 	export GGML_SYCL=1
-	CMAKE_ARGS+=-DGGML_SYCL=ON
 endif
 
 ifeq ($(BUILD_TYPE),sycl_f16)
 	export GGML_SYCL_F16=1
-	CMAKE_ARGS+=-DGGML_SYCL_F16=ON
 endif
 
 ifeq ($(BUILD_TYPE),hipblas)
@@ -185,7 +151,7 @@ ifeq ($(BUILD_TYPE),hipblas)
 	export CC=$(ROCM_HOME)/llvm/bin/clang
 	export STABLE_BUILD_TYPE=
 	export GGML_HIP=1
-	GPU_TARGETS ?= gfx803,gfx900,gfx906,gfx908,gfx90a,gfx942,gfx1010,gfx1030,gfx1032,gfx1100,gfx1101,gfx1102
+	GPU_TARGETS ?= gfx900,gfx906,gfx908,gfx940,gfx941,gfx942,gfx90a,gfx1030,gfx1031,gfx1100,gfx1101
 	AMDGPU_TARGETS ?= "$(GPU_TARGETS)"
 	CMAKE_ARGS+=-DGGML_HIP=ON -DAMDGPU_TARGETS="$(AMDGPU_TARGETS)" -DGPU_TARGETS="$(GPU_TARGETS)"
 	CGO_LDFLAGS += -O3 --rtlib=compiler-rt -unwindlib=libgcc -lhipblas -lrocblas --hip-link -L${ROCM_HOME}/lib/llvm/lib
@@ -320,9 +286,8 @@ sources/whisper.cpp:
 	git checkout $(WHISPER_CPP_VERSION) && \
 	git submodule update --init --recursive --depth 1 --single-branch
 
-sources/whisper.cpp/build/src/libwhisper.a: sources/whisper.cpp
-	cd sources/whisper.cpp && cmake $(WHISPER_CMAKE_ARGS) . -B ./build
-	cd sources/whisper.cpp/build && cmake --build . --config Release
+sources/whisper.cpp/libwhisper.a: sources/whisper.cpp
+	cd sources/whisper.cpp && $(MAKE) libwhisper.a libggml.a
 
 get-sources: sources/go-piper sources/stablediffusion-ggml.cpp sources/bark.cpp sources/whisper.cpp backend/cpp/llama/llama.cpp
 
@@ -372,14 +337,8 @@ clean-tests:
 clean-dc: clean
 	cp -r /build/backend-assets /workspace/backend-assets
 
-## Install Go tools
-install-go-tools:
-	go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
-	go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2
-	go install github.com/GeertJohan/go.rice/rice@latest
-
 ## Build:
-build: prepare backend-assets grpcs install-go-tools ## Build the project
+build: prepare backend-assets grpcs ## Build the project
 	$(info ${GREEN}I local-ai build info:${RESET})
 	$(info ${GREEN}I BUILD_TYPE: ${YELLOW}$(BUILD_TYPE)${RESET})
 	$(info ${GREEN}I GO_TAGS: ${YELLOW}$(GO_TAGS)${RESET})
@@ -389,9 +348,7 @@ ifneq ($(BACKEND_LIBS),)
 	$(MAKE) backend-assets/lib
 	cp -f $(BACKEND_LIBS) backend-assets/lib/
 endif
-	rm -rf $(BINARY_NAME) || true
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o $(BINARY_NAME) ./
-	rice append --exec $(BINARY_NAME)
 
 build-minimal:
 	BUILD_GRPC_FOR_BACKEND_LLAMA=true GRPC_BACKENDS="backend-assets/grpc/llama-cpp-avx2" GO_TAGS=p2p $(MAKE) build
@@ -463,7 +420,6 @@ prepare-test: grpcs
 	cp -rf backend-assets core/http
 	cp tests/models_fixtures/* test-models
 
-## Test targets
 test: prepare test-models/testmodel.ggml grpcs
 	@echo 'Running tests'
 	export GO_TAGS="tts debug"
@@ -538,7 +494,7 @@ protogen: protogen-go protogen-python
 protogen-clean: protogen-go-clean protogen-python-clean
 
 .PHONY: protogen-go
-protogen-go: install-go-tools
+protogen-go:
 	mkdir -p pkg/grpc/proto
 	protoc --experimental_allow_proto3_optional -Ibackend/ --go_out=pkg/grpc/proto/ --go_opt=paths=source_relative --go-grpc_out=pkg/grpc/proto/ --go-grpc_opt=paths=source_relative \
     backend/backend.proto
@@ -642,12 +598,10 @@ prepare-extra-conda-environments: protogen-python
 prepare-test-extra: protogen-python
 	$(MAKE) -C backend/python/transformers
 	$(MAKE) -C backend/python/diffusers
-	$(MAKE) -C backend/python/vllm
 
 test-extra: prepare-test-extra
 	$(MAKE) -C backend/python/transformers test
 	$(MAKE) -C backend/python/diffusers test
-	$(MAKE) -C backend/python/vllm test
 
 backend-assets:
 	mkdir -p backend-assets
@@ -789,8 +743,8 @@ ifneq ($(UPX),)
 	$(UPX) backend-assets/grpc/silero-vad
 endif
 
-backend-assets/grpc/whisper: sources/whisper.cpp sources/whisper.cpp/build/src/libwhisper.a backend-assets/grpc
-	CGO_LDFLAGS="$(CGO_LDFLAGS) $(CGO_LDFLAGS_WHISPER)" C_INCLUDE_PATH="${WHISPER_INCLUDE_PATH}" LIBRARY_PATH="${WHISPER_LIBRARY_PATH}" LD_LIBRARY_PATH="${WHISPER_LIBRARY_PATH}" \
+backend-assets/grpc/whisper: sources/whisper.cpp sources/whisper.cpp/libwhisper.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS) $(CGO_LDFLAGS_WHISPER)" C_INCLUDE_PATH="$(CURDIR)/sources/whisper.cpp/include:$(CURDIR)/sources/whisper.cpp/ggml/include" LIBRARY_PATH=$(CURDIR)/sources/whisper.cpp \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/whisper ./backend/go/transcribe/whisper
 ifneq ($(UPX),)
 	$(UPX) backend-assets/grpc/whisper

README.md

@@ -30,7 +30,7 @@
 
 <p align="center">
 <a href="https://twitter.com/LocalAI_API" target="blank">
-<img src="https://img.shields.io/badge/X-%23000000.svg?style=for-the-badge&logo=X&logoColor=white&label=LocalAI_API" alt="Follow LocalAI_API"/>
+<img src="https://img.shields.io/twitter/follow/LocalAI_API?label=Follow: LocalAI_API&style=social" alt="Follow LocalAI_API"/>
 </a>
 <a href="https://discord.gg/uJAeKSAGDy" target="blank">
 <img src="https://dcbadge.vercel.app/api/server/uJAeKSAGDy?style=flat-square&theme=default-inverted" alt="Join LocalAI Discord Community"/>
@@ -43,8 +43,7 @@
 
 > :bulb: Get help - [❓FAQ](https://localai.io/faq/) [💭Discussions](https://github.com/go-skynet/LocalAI/discussions) [:speech_balloon: Discord](https://discord.gg/uJAeKSAGDy) [:book: Documentation website](https://localai.io/)
 >
-> [💻 Quickstart](https://localai.io/basics/getting_started/) [🖼️ Models](https://models.localai.io/) [🚀 Roadmap](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap) [🥽 Demo](https://demo.localai.io) [🌍 Explorer](https://explorer.localai.io) [🛫 Examples](https://github.com/mudler/LocalAI-examples) Try on 
-[![Telegram](https://img.shields.io/badge/Telegram-2CA5E0?style=for-the-badge&logo=telegram&logoColor=white)](https://t.me/localaiofficial_bot)
+> [💻 Quickstart](https://localai.io/basics/getting_started/) [🖼️ Models](https://models.localai.io/) [🚀 Roadmap](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap) [🥽 Demo](https://demo.localai.io) [🌍 Explorer](https://explorer.localai.io) [🛫 Examples](https://github.com/mudler/LocalAI-examples) 
 
 [![tests](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml)[![Build and Release](https://github.com/go-skynet/LocalAI/actions/workflows/release.yaml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/release.yaml)[![build container images](https://github.com/go-skynet/LocalAI/actions/workflows/image.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/image.yml)[![Bump dependencies](https://github.com/go-skynet/LocalAI/actions/workflows/bump_deps.yaml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/bump_deps.yaml)[![Artifact Hub](https://img.shields.io/endpoint?url=https://artifacthub.io/badge/repository/localai)](https://artifacthub.io/packages/search?repo=localai)
 
@@ -104,12 +103,9 @@
 Run the installer script:
 
 ```bash
-# Basic installation
 curl https://localai.io/install.sh | sh
 ```
 
-For more installation options, see [Installer Options](https://localai.io/docs/advanced/installer/).
-
 Or run with docker:
 
 ### CPU only image:

aio/gpu-8g/text-to-text.yaml

@@ -48,6 +48,6 @@ template:
     <|im_start|>assistant
 
 download_files:
-- filename: localai-functioncall-qwen2.5-7b-v0.5-q4_k_m.gguf
-  sha256: 4e7b7fe1d54b881f1ef90799219dc6cc285d29db24f559c8998d1addb35713d4
-  uri: huggingface://mudler/LocalAI-functioncall-qwen2.5-7b-v0.5-Q4_K_M-GGUF/localai-functioncall-qwen2.5-7b-v0.5-q4_k_m.gguf
+- filename: localai-functioncall-phi-4-v0.3-q4_k_m.gguf
+  sha256: 23fee048ded2a6e2e1a7b6bbefa6cbf83068f194caa9552aecbaa00fec8a16d5
+  uri: huggingface://mudler/LocalAI-functioncall-phi-4-v0.3-Q4_K_M-GGUF/localai-functioncall-phi-4-v0.3-q4_k_m.gguf

assets.go

@@ -1,15 +1,6 @@
 package main
 
-import (
-	rice "github.com/GeertJohan/go.rice"
-)
+import "embed"
 
-var backendAssets *rice.Box
-
-func init() {
-	var err error
-	backendAssets, err = rice.FindBox("backend-assets")
-	if err != nil {
-		panic(err)
-	}
-}
+//go:embed backend-assets/*
+var backendAssets embed.FS

backend/backend.proto

@@ -14,7 +14,6 @@ service Backend {
   rpc PredictStream(PredictOptions) returns (stream Reply) {}
   rpc Embedding(PredictOptions) returns (EmbeddingResult) {}
   rpc GenerateImage(GenerateImageRequest) returns (Result) {}
-  rpc GenerateVideo(GenerateVideoRequest) returns (Result) {}
   rpc AudioTranscription(TranscriptRequest) returns (TranscriptResult) {}
   rpc TTS(TTSRequest) returns (Result) {}
   rpc SoundGeneration(SoundGenerationRequest) returns (Result) {}
@@ -302,19 +301,6 @@ message GenerateImageRequest {
   int32 CLIPSkip = 11;
 }
 
-message GenerateVideoRequest {
-  string prompt = 1;
-  string start_image = 2;  // Path or base64 encoded image for the start frame
-  string end_image = 3;    // Path or base64 encoded image for the end frame
-  int32 width = 4;
-  int32 height = 5;
-  int32 num_frames = 6;    // Number of frames to generate
-  int32 fps = 7;          // Frames per second
-  int32 seed = 8;
-  float cfg_scale = 9;    // Classifier-free guidance scale
-  string dst = 10;        // Output path for the generated video
-}
-
 message TTSRequest {
   string text = 1;
   string model = 2;

backend/cpp/llama/CMakeLists.txt

@@ -1,17 +1,17 @@
 
 ## XXX: In some versions of CMake clip wasn't being built before llama.
 ## This is an hack for now, but it should be fixed in the future.
-# set(TARGET myclip)
-# add_library(${TARGET} clip.cpp clip.h clip-impl.h llava.cpp llava.h)
-# install(TARGETS ${TARGET} LIBRARY)
-# target_include_directories(myclip PUBLIC .)
-# target_include_directories(myclip PUBLIC ../..)
-# target_include_directories(myclip PUBLIC ../../common)
-# target_link_libraries(${TARGET} PRIVATE common ggml llama ${CMAKE_THREAD_LIBS_INIT})
-# target_compile_features(${TARGET} PRIVATE cxx_std_11)
-# if (NOT MSVC)
-#     target_compile_options(${TARGET} PRIVATE -Wno-cast-qual) # stb_image.h
-# endif()
+set(TARGET myclip)
+add_library(${TARGET} clip.cpp clip.h clip-impl.h llava.cpp llava.h)
+install(TARGETS ${TARGET} LIBRARY)
+target_include_directories(myclip PUBLIC .)
+target_include_directories(myclip PUBLIC ../..)
+target_include_directories(myclip PUBLIC ../../common)
+target_link_libraries(${TARGET} PRIVATE common ggml llama ${CMAKE_THREAD_LIBS_INIT})
+target_compile_features(${TARGET} PRIVATE cxx_std_11)
+if (NOT MSVC)
+    target_compile_options(${TARGET} PRIVATE -Wno-cast-qual) # stb_image.h
+endif()
 # END CLIP hack
 
 
@@ -75,11 +75,7 @@ add_library(hw_grpc_proto
   ${hw_proto_hdrs} )
 
 add_executable(${TARGET} grpc-server.cpp utils.hpp json.hpp)
-
-target_include_directories(${TARGET} PRIVATE ../llava)
-target_include_directories(${TARGET} PRIVATE ${CMAKE_SOURCE_DIR})
-
-target_link_libraries(${TARGET} PRIVATE common llama mtmd ${CMAKE_THREAD_LIBS_INIT} absl::flags hw_grpc_proto
+target_link_libraries(${TARGET} PRIVATE common llama myclip ${CMAKE_THREAD_LIBS_INIT} absl::flags hw_grpc_proto
   absl::flags_parse
   gRPC::${_REFLECTION}
   gRPC::${_GRPC_GRPCPP}

backend/cpp/llama/Makefile

@@ -59,8 +59,8 @@ llama.cpp:
 	git checkout -b build $(LLAMA_VERSION) && \
 	git submodule update --init --recursive --depth 1 --single-branch
 
-llama.cpp/tools/grpc-server: llama.cpp
-	mkdir -p llama.cpp/tools/grpc-server
+llama.cpp/examples/grpc-server: llama.cpp
+	mkdir -p llama.cpp/examples/grpc-server
 	bash prepare.sh
 
 rebuild:
@@ -70,13 +70,13 @@ rebuild:
 
 purge:
 	rm -rf llama.cpp/build
-	rm -rf llama.cpp/tools/grpc-server
+	rm -rf llama.cpp/examples/grpc-server
 	rm -rf grpc-server
 
 clean: purge
 	rm -rf llama.cpp
 
-grpc-server: llama.cpp llama.cpp/tools/grpc-server
+grpc-server: llama.cpp llama.cpp/examples/grpc-server
 	@echo "Building grpc-server with $(BUILD_TYPE) build type and $(CMAKE_ARGS)"
 ifneq (,$(findstring sycl,$(BUILD_TYPE)))
 	+bash -c "source $(ONEAPI_VARS); \

backend/cpp/llama/grpc-server.cpp

@@ -11,7 +11,8 @@
 #include <memory>
 #include <string>
 #include <getopt.h>
-#include "mtmd.h"
+#include "clip.h"
+#include "llava.h"
 #include "log.h"
 #include "stb_image.h"
 #include "common.h"
@@ -51,7 +52,7 @@ struct server_params
 {
     std::string hostname = "127.0.0.1";
     std::vector<std::string> api_keys;
-    std::string public_path = "tools/server/public";
+    std::string public_path = "examples/server/public";
     std::string chat_template = "";
     int32_t port = 8080;
     int32_t read_timeout = 600;
@@ -209,8 +210,6 @@ struct llama_client_slot
     int32_t num_prompt_tokens_processed = 0;
 
     json prompt;
-    json data;
-
     std::string generated_text;
     llama_token sampled;
     std::vector<llama_token> cache_tokens;
@@ -218,6 +217,7 @@ struct llama_client_slot
 
     bool infill = false;
     bool embedding = false;
+    bool reranker = false;
     bool has_next_token = true;
     bool truncated = false;
     bool stopped_eos = false;
@@ -240,7 +240,7 @@ struct llama_client_slot
     int32_t n_past_se = 0; // self-extend
 
     // multimodal
-    mtmd_context * mctx = nullptr;
+    std::vector<slot_image> images;
 
     // stats
     size_t sent_count = 0;
@@ -271,6 +271,17 @@ struct llama_client_slot
         n_past_se              = 0;
 
         generated_token_probs.clear();
+
+        for (slot_image & img : images)
+        {
+            free(img.image_embedding);
+            if (img.img_data) {
+                clip_image_u8_free(img.img_data);
+            }
+            img.prefix_prompt = "";
+        }
+
+        images.clear();
     }
 
     bool has_budget(common_params &global_params) {
@@ -446,9 +457,6 @@ struct llama_server_context
     llama_context *ctx = nullptr;
     const llama_vocab * vocab = nullptr;
 
-    // multimodal
-    mtmd_context * mctx = nullptr;
-
     clip_ctx *clp_ctx = nullptr;
 
     common_params params;
@@ -487,10 +495,6 @@ struct llama_server_context
 
     ~llama_server_context()
     {
-        if (mctx) {
-            mtmd_free(mctx);
-            mctx = nullptr;
-        }
         if (ctx)
         {
             llama_free(ctx);
@@ -509,14 +513,12 @@ struct llama_server_context
         if (!params.mmproj.path.empty()) {
             multimodal = true;
             LOG_INFO("Multi Modal Mode Enabled", {});
-            mtmd_context_params mparams = mtmd_context_params_default();
-            mparams.use_gpu       = has_gpu;
-            mparams.print_timings = false;
-            mparams.n_threads     = params.cpuparams.n_threads;
-            mparams.verbosity     = GGML_LOG_LEVEL_INFO;
-            mctx = mtmd_init_from_file(params.mmproj.path.c_str(), model, mparams);
-            if (mctx == nullptr) {
-                LOG_ERR("failed to load multimodal model, '%s'\n", params.mmproj.path.c_str());
+            clp_ctx = clip_init(params.mmproj.path.c_str(), clip_context_params {
+                /* use_gpu */ has_gpu,
+                /*verbosity=*/ GGML_LOG_LEVEL_INFO,
+            });
+            if(clp_ctx == nullptr) {
+                LOG_ERR("unable to load clip model: %s", params.mmproj.path.c_str());
                 return false;
             }
 
@@ -534,6 +536,12 @@ struct llama_server_context
             return false;
         }
 
+        // Enable reranking if embeddings are enabled - moved after context initialization
+        if (params.embedding) {
+            params.reranking = true;
+            LOG_INFO("Reranking enabled (embeddings are enabled)", {});
+        }
+
         if (multimodal) {
             const int n_embd_clip = clip_n_mmproj_embd(clp_ctx);
             const int n_embd_llm  = llama_model_n_embd(model);
@@ -578,8 +586,6 @@ struct llama_server_context
             slot.id = i;
             slot.n_ctx = n_ctx_slot;
             slot.n_predict = params.n_predict;
-            slot.mctx = mctx;
-            //slot.cache_tokens.has_mtmd = mctx != nullptr;
 
             LOG_INFO("new slot", {
                 {"slot_id",    slot.id},
@@ -617,61 +623,54 @@ struct llama_server_context
         batch = llama_batch_init(n_ctx, 0, params.n_parallel);
     }
 
-    std::vector<server_tokens> tokenize(json &data, const json & json_prompt, bool add_bos) const
+    std::vector<llama_token> tokenize(const json & json_prompt, bool add_bos) const
     {
-        mtmd::bitmaps bitmaps;
-        std::vector<server_tokens> inputs;
+        // TODO: currently, we tokenize using special tokens by default
+        //       this is not always correct (see https://github.com/ggerganov/llama.cpp/pull/4160#issuecomment-1824826216)
+        //       but it's better compared to completely ignoring ChatML and other chat templates
+        const bool TMP_FORCE_SPECIAL = true;
 
-        if (mctx != nullptr)
+        // If `add_bos` is true, we only add BOS, when json_prompt is a string,
+        // or the first element of the json_prompt array is a string.
+        std::vector<llama_token> prompt_tokens;
+
+        if (json_prompt.is_array())
         {
-            const auto &images_data = data.find("image_data");
-            if (images_data != data.end() && images_data->is_array())
+            bool first = true;
+            for (const auto& p : json_prompt)
             {
-                for (const auto &img : *images_data)
+                if (p.is_string())
                 {
-                    const std::vector<uint8_t> image_buffer = base64_decode(img["data"].get<std::string>());
-
-                    mtmd::bitmap bmp(mtmd_helper_bitmap_init_from_buf(image_buffer.data(), image_buffer.size()));
-                    if (!bmp.ptr) {
-                            throw std::runtime_error("Failed to load image");
+                    auto s = p.template get<std::string>();
+                    std::vector<llama_token> p;
+                    if (first)
+                    {
+                        p = common_tokenize(ctx, s, add_bos, TMP_FORCE_SPECIAL);
+                        first = false;
                     }
-                    // calculate bitmap hash (for KV caching)
-                    std::string hash = fnv_hash(bmp.data(), bmp.nx()*bmp.ny()*3);
-                    bmp.set_id(hash.c_str());
-                    bitmaps.entries.push_back(std::move(bmp));
+                    else
+                    {
+                        p = common_tokenize(ctx, s, false, TMP_FORCE_SPECIAL);
+                    }
+                    prompt_tokens.insert(prompt_tokens.end(), p.begin(), p.end());
+                }
+                else
+                {
+                    if (first)
+                    {
+                        first = false;
+                    }
+                    prompt_tokens.push_back(p.template get<llama_token>());
                 }
             }
-
-            // multimodal
-            std::string prompt_str = json_prompt.template get<std::string>();
-            mtmd_input_text inp_txt = {
-                prompt_str.c_str(),
-                /* add_special */   true,
-                /* parse_special */ true,
-            };
-            mtmd::input_chunks chunks(mtmd_input_chunks_init());
-            auto bitmaps_c_ptr = bitmaps.c_ptr();
-            int32_t tokenized = mtmd_tokenize(mctx,
-                                                chunks.ptr.get(),
-                                                &inp_txt,
-                                                bitmaps_c_ptr.data(),
-                                                bitmaps_c_ptr.size());
-            if (tokenized != 0) {
-                throw std::runtime_error("Failed to tokenize prompt");
-            }
-
-            server_tokens tmp(chunks, true);
-            inputs.push_back(std::move(tmp));
-        } else {
-            // non-multimodal version
-            auto tokenized_prompts = tokenize_input_prompts(vocab, json_prompt, true, true);
-            for (auto & p : tokenized_prompts) {
-                auto tmp = server_tokens(p, mctx != nullptr);
-                inputs.push_back(std::move(tmp));
-            }
+        }
+        else
+        {
+            auto s = json_prompt.template get<std::string>();
+            prompt_tokens = common_tokenize(ctx, s, add_bos, TMP_FORCE_SPECIAL);
         }
 
-        return inputs;
+        return prompt_tokens;
     }
 
     llama_client_slot* get_slot(int id) {
@@ -724,8 +723,6 @@ struct llama_server_context
         slot->sparams.grammar_triggers = grammar_triggers;
         slot->sparams.grammar_lazy = grammar_lazy;
 
-        slot->data = data;
-
         if (slot->n_predict > 0 && slot->params.n_predict > slot->n_predict) {
             // Might be better to reject the request with a 400 ?
             LOG_WARNING("Max tokens to predict exceeds server configuration", {
@@ -767,7 +764,43 @@ struct llama_server_context
         if (json_value(data, "ignore_eos", false) && has_eos_token) {
                 slot->sparams.logit_bias.push_back({llama_vocab_eos(vocab), -INFINITY});
         }
-        
+        /*
+        slot->sparams.penalty_prompt_tokens.clear();
+        slot->sparams.use_penalty_prompt_tokens = false;
+        const auto &penalty_prompt = data.find("penalty_prompt");
+        if (penalty_prompt != data.end())
+        {
+            if (penalty_prompt->is_string())
+            {
+                const auto penalty_prompt_string = penalty_prompt->get<std::string>();
+                auto penalty_tokens = llama_tokenize(model, penalty_prompt_string, false);
+                slot->sparams.penalty_prompt_tokens.swap(penalty_tokens);
+                if (slot->params.n_predict > 0)
+                {
+                    slot->sparams.penalty_prompt_tokens.reserve(slot->sparams.penalty_prompt_tokens.size() + slot->params.n_predict);
+                }
+                slot->sparams.use_penalty_prompt_tokens = true;
+            }
+            else if (penalty_prompt->is_array())
+            {
+                const auto n_tokens = penalty_prompt->size();
+                slot->sparams.penalty_prompt_tokens.reserve(n_tokens + std::max(0, slot->params.n_predict));
+                const int n_vocab = llama_n_vocab(model);
+                for (const auto &penalty_token : *penalty_prompt)
+                {
+                    if (penalty_token.is_number_integer())
+                    {
+                        const auto tok = penalty_token.get<llama_token>();
+                        if (tok >= 0 && tok < n_vocab)
+                        {
+                            slot->sparams.penalty_prompt_tokens.push_back(tok);
+                        }
+                    }
+                }
+                slot->sparams.use_penalty_prompt_tokens = true;
+            }
+        }
+      */
         slot->sparams.logit_bias.clear();
 
         const auto &logit_bias = data.find("logit_bias");
@@ -843,6 +876,79 @@ struct llama_server_context
         }
         
 
+        if (multimodal)
+        {
+            const auto &images_data = data.find("image_data");
+            if (images_data != data.end() && images_data->is_array())
+            {
+                for (const auto &img : *images_data)
+                {
+                    const std::vector<uint8_t> image_buffer = base64_decode(img["data"].get<std::string>());
+
+                    slot_image img_sl;
+                    img_sl.id = img.count("id") != 0 ? img["id"].get<int>() : slot->images.size();
+                    img_sl.img_data = clip_image_u8_init();
+                    if (!clip_image_load_from_bytes(image_buffer.data(), image_buffer.size(), img_sl.img_data))
+                    {
+                        LOG_ERR("%s: failed to load image, slot_id: %d, img_sl_id: %d", 
+                             __func__,
+                             slot->id,
+                             img_sl.id
+                        );
+                        return false;
+                    }
+                    LOG_VERBOSE("image loaded", {
+                        {"slot_id",   slot->id},
+                        {"img_sl_id", img_sl.id}
+                    });
+                    img_sl.request_encode_image = true;
+                    slot->images.push_back(img_sl);
+                }
+                // process prompt
+                // example: system prompt [img-102] user [img-103] describe [img-134] -> [{id: 102, prefix: 'system prompt '}, {id: 103, prefix: ' user '}, {id: 134, prefix: ' describe '}]}
+                if (slot->images.size() > 0 && !slot->prompt.is_array())
+                {
+                    std::string prompt = slot->prompt.get<std::string>();
+                    size_t pos = 0, begin_prefix = 0;
+                    std::string pattern = "[img-";
+                    while ((pos = prompt.find(pattern, pos)) != std::string::npos) {
+                        size_t end_prefix = pos;
+                        pos += pattern.length();
+                        size_t end_pos = prompt.find(']', pos);
+                        if (end_pos != std::string::npos)
+                        {
+                            std::string image_id = prompt.substr(pos, end_pos - pos);
+                            try
+                            {
+                                int img_id = std::stoi(image_id);
+                                bool found = false;
+                                for (slot_image &img : slot->images)
+                                {
+                                    if (img.id == img_id) {
+                                        found = true;
+                                        img.prefix_prompt = prompt.substr(begin_prefix, end_prefix - begin_prefix);
+                                        begin_prefix = end_pos + 1;
+                                        break;
+                                    }
+                                }
+                                if (!found) {
+                                    LOG("ERROR: Image with id: %i, not found.\n", img_id);
+                                    slot->images.clear();
+                                    return false;
+                                }
+                            } catch (const std::invalid_argument& e) {
+                                LOG("Invalid image number id in prompt\n");
+                                slot->images.clear();
+                                return false;
+                            }
+                        }
+                    }
+                    slot->prompt = "";
+                    slot->params.input_suffix = prompt.substr(begin_prefix);
+                    slot->params.cache_prompt = false; // multimodal doesn't support cache prompt
+                }
+            }
+        }
 
         if (slot->ctx_sampling != nullptr)
         {
@@ -1090,6 +1196,26 @@ struct llama_server_context
         return slot.has_next_token; // continue
     }
 
+    bool process_images(llama_client_slot &slot) const
+    {
+        for (slot_image &img : slot.images)
+        {
+            if (!img.request_encode_image)
+            {
+                continue;
+            }
+
+            if (!llava_image_embed_make_with_clip_img(clp_ctx, params.cpuparams.n_threads, img.img_data, &img.image_embedding, &img.image_tokens)) {
+                LOG("Error processing the given image");
+                return false;
+            }
+
+            img.request_encode_image = false;
+        }
+
+        return slot.images.size() > 0;
+    }
+
     void send_error(task_server& task, const std::string &error)
     {
         LOG("task %i - error: %s\n", task.id, error.c_str());
@@ -1294,7 +1420,59 @@ struct llama_server_context
         queue_results.send(res);
     }
 
-    void request_completion(int task_id, json data, bool infill, bool embedding, int multitask_id)
+    void send_rerank(llama_client_slot &slot, const llama_batch & batch)
+    {
+        task_result res;
+        res.id = slot.task_id;
+        res.multitask_id = slot.multitask_id;
+        res.error = false;
+        res.stop = true;
+
+        float score = -1e6f; // Default score if we fail to get embeddings
+
+        if (!params.reranking)
+        {
+            LOG_WARNING("reranking disabled", {
+                {"params.reranking", params.reranking},
+            });
+        }
+        else if (ctx == nullptr)
+        {
+            LOG_ERR("context is null, cannot perform reranking");
+            res.error = true;
+        }
+        else
+        {
+            for (int i = 0; i < batch.n_tokens; ++i) {
+                if (!batch.logits[i] || batch.seq_id[i][0] != slot.id) {
+                    continue;
+                }
+
+                const float * embd = llama_get_embeddings_seq(ctx, batch.seq_id[i][0]);
+                if (embd == NULL) {
+                    embd = llama_get_embeddings_ith(ctx, i);
+                }
+
+                if (embd == NULL) {
+                    LOG("failed to get embeddings");
+                    continue;
+                }
+
+                score = embd[0];
+            }
+        }
+
+        // Format result as JSON similar to the embedding function
+        res.result_json = json
+        {
+            {"score", score},
+            {"tokens", slot.num_prompt_tokens}
+        };
+        
+        queue_results.send(res);
+    }
+
+    void request_completion(int task_id, json data, bool infill, bool embedding, bool rerank, int multitask_id)
     {
         task_server task;
         task.id = task_id;
@@ -1302,6 +1480,7 @@ struct llama_server_context
         task.data = std::move(data);
         task.infill_mode = infill;
         task.embedding_mode = embedding;
+        task.reranking_mode = rerank;
         task.type = TASK_TYPE_COMPLETION;
         task.multitask_id = multitask_id;
 
@@ -1332,6 +1511,74 @@ struct llama_server_context
         }
     }
 
+    // for multiple images processing
+    bool ingest_images(llama_client_slot &slot, int n_batch)
+    {
+        int image_idx = 0;
+
+        while (image_idx < (int) slot.images.size())
+        {
+            slot_image &img = slot.images[image_idx];
+
+            // process prefix prompt
+            for (int32_t i = 0; i < (int32_t) batch.n_tokens; i += n_batch)
+            {
+                const int32_t n_tokens = std::min(n_batch, (int32_t) (batch.n_tokens - i));
+                llama_batch batch_view = {
+                    n_tokens,
+                    batch.token    + i,
+                    nullptr,
+                    batch.pos      + i,
+                    batch.n_seq_id + i,
+                    batch.seq_id   + i,
+                    batch.logits   + i,
+                };
+                if (llama_decode(ctx, batch_view))
+                {
+                    LOG("%s : failed to eval\n", __func__);
+                    return false;
+                }
+            }
+
+            // process image with llm
+            for (int i = 0; i < img.image_tokens; i += n_batch)
+            {
+                int n_eval = img.image_tokens - i;
+                if (n_eval > n_batch)
+                {
+                    n_eval = n_batch;
+                }
+
+                const int n_embd = llama_model_n_embd(model);
+                float * embd = img.image_embedding + i * n_embd;
+                llava_embd_batch llava_batch = llava_embd_batch(embd, n_eval, slot.n_past, 0);
+                if (llama_decode(ctx, llava_batch.batch))
+                {
+                    LOG("%s : failed to eval image\n", __func__);
+                    return false;
+                }
+                slot.n_past += n_eval;
+            }
+            image_idx++;
+
+            common_batch_clear(batch);
+
+            // append prefix of next image
+            const auto json_prompt = (image_idx >= (int) slot.images.size()) ?
+                slot.params.input_suffix : // no more images, then process suffix prompt
+                (json)(slot.images[image_idx].prefix_prompt);
+
+            std::vector<llama_token> append_tokens = tokenize(json_prompt, false); // has next image
+            for (int i = 0; i < (int) append_tokens.size(); ++i)
+            {
+                common_batch_add(batch, append_tokens[i], system_tokens.size() + slot.n_past, { slot.id }, true);
+                slot.n_past += 1;
+            }
+        }
+
+        return true;
+    }
+
     void request_cancel(int task_id)
     {
         task_server task;
@@ -1365,7 +1612,7 @@ struct llama_server_context
             subtask_data["prompt"] = subtask_data["prompt"][i];
 
             // subtasks inherit everything else (infill mode, embedding mode, etc.)
-            request_completion(subtask_ids[i], subtask_data, multiprompt_task.infill_mode, multiprompt_task.embedding_mode, multitask_id);
+            request_completion(subtask_ids[i], subtask_data, multiprompt_task.infill_mode, multiprompt_task.embedding_mode, multiprompt_task.reranking_mode, multitask_id);
         }
     }
 
@@ -1404,6 +1651,7 @@ struct llama_server_context
 
                 slot->infill       = task.infill_mode;
                 slot->embedding    = task.embedding_mode;
+                slot->reranker    = task.reranking_mode;
                 slot->task_id      = task.id;
                 slot->multitask_id = task.multitask_id;
 
@@ -1546,7 +1794,7 @@ struct llama_server_context
         {
             for (auto & slot : slots)
             {
-                const bool has_prompt = slot.prompt.is_array() || (slot.prompt.is_string() && !slot.prompt.get<std::string>().empty());
+                const bool has_prompt = slot.prompt.is_array() || (slot.prompt.is_string() && !slot.prompt.get<std::string>().empty()) || !slot.images.empty();
 
                 // empty prompt passed -> release the slot and send empty response
                 // note: infill mode allows empty prompt
@@ -1563,7 +1811,7 @@ struct llama_server_context
                 {
                     slot.state = PROCESSING;
                     slot.command = NONE;
-                    std::vector<server_tokens> prompt_tokens;
+                    std::vector<llama_token> prompt_tokens;
                     slot.t_start_process_prompt = ggml_time_us();
                     slot.t_start_genereration = 0;
 
@@ -1575,41 +1823,24 @@ struct llama_server_context
                             params.input_suffix.erase(0, 1);
                             suff_rm_leading_spc = false;
                         }
-                        auto prefix_tokens = tokenize(slot.data, slot.params.input_prefix, false);
-                        auto suffix_tokens = tokenize(slot.data, slot.params.input_suffix, false);
+                        auto prefix_tokens = tokenize(slot.params.input_prefix, false);
+                        auto suffix_tokens = tokenize(slot.params.input_suffix, false);
 
                         const int space_token = 29871; // TODO: this should not be hardcoded
-                        if (suff_rm_leading_spc && !suffix_tokens.empty() && suffix_tokens[0][0] == space_token) {
+                        if (suff_rm_leading_spc && !suffix_tokens.empty() && suffix_tokens[0] == space_token) {
                             suffix_tokens.erase(suffix_tokens.begin());
                         }
 
-                        // Create llama_tokens vectors for the special tokens
-                        llama_tokens fim_pre_tokens;
-                        fim_pre_tokens.push_back(llama_vocab_fim_pre(vocab));
-                        llama_tokens bos_tokens;
-                        bos_tokens.push_back(llama_vocab_bos(vocab));
-                        llama_tokens fim_suf_tokens;
-                        fim_suf_tokens.push_back(llama_vocab_fim_suf(vocab));
-                        llama_tokens fim_mid_tokens;
-                        fim_mid_tokens.push_back(llama_vocab_fim_mid(vocab));
-
-                        // Create server_tokens objects
-                        server_tokens fim_pre_token(fim_pre_tokens, mctx != nullptr);
-                        server_tokens bos_token(bos_tokens, mctx != nullptr);
-                        server_tokens fim_suf_token(fim_suf_tokens, mctx != nullptr);
-                        server_tokens fim_mid_token(fim_mid_tokens, mctx != nullptr);
-
-                        // Insert tokens in the correct order
-                        prefix_tokens.insert(prefix_tokens.begin(), fim_pre_token);
-                        prefix_tokens.insert(prefix_tokens.begin(), bos_token); // always add BOS
-                        prefix_tokens.insert(prefix_tokens.end(), fim_suf_token);
-                        prefix_tokens.insert(prefix_tokens.end(), suffix_tokens.begin(), suffix_tokens.end());
-                        prefix_tokens.push_back(fim_mid_token);
+                        prefix_tokens.insert(prefix_tokens.begin(), llama_vocab_fim_pre(vocab));
+                        prefix_tokens.insert(prefix_tokens.begin(), llama_vocab_bos(vocab)); // always add BOS
+                        prefix_tokens.insert(prefix_tokens.end(),   llama_vocab_fim_suf(vocab));
+                        prefix_tokens.insert(prefix_tokens.end(),   suffix_tokens.begin(), suffix_tokens.end());
+                        prefix_tokens.push_back(llama_vocab_fim_mid(vocab));
                         prompt_tokens = prefix_tokens;
                     }
                     else
                     {
-                        prompt_tokens = tokenize(slot.data, slot.prompt, system_prompt.empty() && add_bos_token);  // add BOS if there isn't system prompt
+                        prompt_tokens = tokenize(slot.prompt, system_prompt.empty() && add_bos_token);  // add BOS if there isn't system prompt
                     }
 
                     slot.num_prompt_tokens = prompt_tokens.size();
@@ -1637,12 +1868,7 @@ struct llama_server_context
                             {"new_tokens", tokens_to_str(ctx, new_tokens.cbegin(), new_tokens.cend())},
                         });
                         slot.truncated = true;
-                        
-                        // Convert new_tokens to server_tokens
-                        std::vector<server_tokens> new_prompt_tokens;
-                        server_tokens new_server_tokens(new_tokens, mctx != nullptr);
-                        new_prompt_tokens.push_back(std::move(new_server_tokens));
-                        prompt_tokens = std::move(new_prompt_tokens);
+                        prompt_tokens = new_tokens;
 
                         slot.num_prompt_tokens = prompt_tokens.size();
                         GGML_ASSERT(slot.num_prompt_tokens < slot.n_ctx);
@@ -1662,17 +1888,10 @@ struct llama_server_context
                         // push the prompt into the sampling context (do not apply grammar)
                         for (auto &token : prompt_tokens)
                         {
-                            // Convert server_tokens to llama_token for sampling
-                            llama_token tok = token[0];  // Get first token
-                            common_sampler_accept(slot.ctx_sampling, tok, false);
+                            common_sampler_accept(slot.ctx_sampling, token, false);
                         }
 
-                        // Convert server_tokens to llama_tokens for comparison
-                        std::vector<llama_token> prompt_llama_tokens;
-                        for (const auto &token : prompt_tokens) {
-                            prompt_llama_tokens.push_back(token[0]);
-                        }
-                        slot.n_past = common_part(slot.cache_tokens, prompt_llama_tokens);
+                        slot.n_past = common_part(slot.cache_tokens, prompt_tokens);
 
                         // the last token of the cache is not in the KV cache until the next call to llama_decode
                         // (it was sampled, pushed into the "cache_tokens", but not yet put in the context)
@@ -1710,12 +1929,7 @@ struct llama_server_context
                         });
                     }
 
-                    // Convert server_tokens to llama_tokens for cache
-                    std::vector<llama_token> cache_llama_tokens;
-                    for (const auto &token : prompt_tokens) {
-                        cache_llama_tokens.push_back(token[0]);
-                    }
-                    slot.cache_tokens = cache_llama_tokens;
+                    slot.cache_tokens = prompt_tokens;
 
                     if (slot.n_past == slot.num_prompt_tokens && slot.n_past > 0)
                     {
@@ -1739,36 +1953,18 @@ struct llama_server_context
                     });
                     llama_kv_cache_seq_rm(ctx, slot.id, p0, -1);
 
-
-                    // process the prefix of first image
-                    std::vector<server_tokens> prefix_tokens = prompt_tokens;
-
-                    int32_t slot_npast = slot.n_past_se > 0 ? slot.n_past_se : slot.n_past;
-
-                    // check if we should process the image
-                    if (slot.n_past < slot.n_prompt_tokens
-                            && slot.prompt_tokens[slot.n_past] == LLAMA_TOKEN_NULL) {
-                        // process the image
-                        int32_t new_n_past;
-                        int32_t res = prompt_tokens.process_chunk(ctx, mctx, slot.n_past, slot.id, new_n_past);
-                        int32_t n_pos = new_n_past - slot.n_past;
-                        if (res != 0) {
-                            slot.release();
-                            LOG_ERR("failed to process image, res = %d\n", res);
-                            continue;
-                        }
-
-
-                        slot.n_past                    += n_pos;
-                       // slot.n_prompt_tokens_processed += n_pos;
-                    }
-
                     LOG_VERBOSE("prompt ingested", {
                                                     {"n_past",  slot.n_past},
                                                     {"cached",  tokens_to_str(ctx, slot.cache_tokens.cbegin(), slot.cache_tokens.cbegin() + slot.n_past)},
                                                     {"to_eval", tokens_to_str(ctx, slot.cache_tokens.cbegin() + slot.n_past, slot.cache_tokens.cend())},
                                                 });
 
+                    const bool has_images = process_images(slot);
+
+                    // process the prefix of first image
+                    std::vector<llama_token> prefix_tokens = has_images ? tokenize(slot.images[0].prefix_prompt, add_bos_token) : prompt_tokens;
+
+                    int32_t slot_npast = slot.n_past_se > 0 ? slot.n_past_se : slot.n_past;
 
                     int32_t ga_i = slot.ga_i;
                     int32_t ga_n = slot.ga_n;
@@ -1788,6 +1984,19 @@ struct llama_server_context
                         slot_npast++;
                     }
 
+                    if (has_images && !ingest_images(slot, n_batch))
+                    {
+                        LOG_ERR("%s: failed processing images Slot id : %d, Task id: %d", 
+                            __func__,
+                            slot.id,
+                            slot.task_id
+                        );
+                        // FIXME @phymbert: to be properly tested
+                        //  early returning without changing the slot state will block the slot for ever
+                        // no one at the moment is checking the return value
+                        return false;
+                    }
+
                     // extract the logits only for the last token
                     if (batch.n_tokens > 0)
                     {
@@ -1886,6 +2095,14 @@ struct llama_server_context
                     continue;
                 }
 
+                if (slot.reranker)
+                {
+                    send_rerank(slot, batch_view);
+                    slot.release();
+                    slot.i_batch = -1;
+                    continue;
+                }
+
                 completion_token_output result;
                 const llama_token id = common_sampler_sample(slot.ctx_sampling, ctx, slot.i_batch - i);
 
@@ -2016,6 +2233,26 @@ static void start_llama_server() {
 json parse_options(bool streaming, const backend::PredictOptions* predict, llama_server_context &llama)
 {
     
+    // This is for example a slot data from the json data
+    //     slot->params.stream           = json_value(data, "stream",            false);
+    //     slot->params.cache_prompt     = json_value(data, "cache_prompt",      false);
+    //     slot->params.n_predict        = json_value(data, "n_predict",         default_params.n_predict);
+    //     slot->sparams.top_k           = json_value(data, "top_k",             default_sparams.top_k);
+    //     slot->sparams.top_p           = json_value(data, "top_p",             default_sparams.top_p);
+    //     slot->sparams.typical_p       = json_value(data, "typical_p",         default_sparams.typical_p);
+    //     slot->sparams.temp            = json_value(data, "temperature",       default_sparams.temp);
+    //     slot->sparams.penalty_last_n  = json_value(data, "repeat_last_n",     default_sparams.penalty_last_n);
+    //     slot->sparams.penalty_repeat  = json_value(data, "repeat_penalty",    default_sparams.penalty_repeat);
+    //     slot->sparams.penalty_freq    = json_value(data, "frequency_penalty", default_sparams.penalty_freq);
+    //     slot->sparams.penalty_present = json_value(data, "presence_penalty",  default_sparams.penalty_present);
+    //     slot->sparams.mirostat        = json_value(data, "mirostat",          default_sparams.mirostat);
+    //     slot->sparams.mirostat_tau    = json_value(data, "mirostat_tau",      default_sparams.mirostat_tau);
+    //     slot->sparams.mirostat_eta    = json_value(data, "mirostat_eta",      default_sparams.mirostat_eta);
+    //     slot->params.n_keep           = json_value(data, "n_keep",            slot->params.n_keep);
+    //     slot->params.seed             = json_value(data, "seed",              default_params.seed);
+    //     slot->sparams.grammar         = json_value(data, "grammar",           default_sparams.grammar);
+    //     slot->sparams.n_probs         = json_value(data, "n_probs",           default_sparams.n_probs);
+
     // Create now a json data from the prediction options instead
     //
     json data;
@@ -2060,6 +2297,69 @@ json parse_options(bool streaming, const backend::PredictOptions* predict, llama
     return data;
 }
 
+// static void parse_options_completion(bool streaming,const backend::PredictOptions* predict, llama_server_context &llama)
+// {
+//     // https://github.com/ggerganov/llama.cpp/blob/d9b33fe95bd257b36c84ee5769cc048230067d6f/examples/server/server.cpp#L673
+//     gpt_params default_params;
+
+//     llama.stream = streaming;
+//     llama.params.n_predict = predict->tokens() == 0 ? -1 : predict->tokens();
+//     llama.params.sparams.top_k = predict->topk();
+//     llama.params.sparams.top_p = predict->topp();
+//     llama.params.sparams.typical_p = predict->typicalp();
+//     llama.params.sparams.penalty_last_n = predict->repeat();
+//     llama.params.sparams.temp = predict->temperature();
+//     llama.params.sparams.penalty_repeat = predict->penalty();
+//     llama.params.sparams.penalty_present = predict->presencepenalty();
+//     llama.params.sparams.penalty_freq = predict->frequencypenalty();
+//     llama.params.sparams.mirostat = predict->mirostat();
+//     llama.params.sparams.mirostat_tau = predict->mirostattau();
+//     llama.params.sparams.mirostat_eta = predict->mirostateta();
+//     llama.params.n_keep = predict->nkeep();
+//     llama.params.seed = predict->seed();
+//     llama.params.sparams.grammar = predict->grammar();
+//     // llama.params.n_probs = predict->
+//     llama.params.prompt = predict->prompt();
+
+//     llama.params.sparams.logit_bias.clear();
+
+//     if (predict->ignoreeos())
+//     {
+//         llama.params.sparams.logit_bias[llama_token_eos(llama.model)] = -INFINITY;
+//     }
+
+//     // const auto &logit_bias = body.find("logit_bias");
+//     // if (logit_bias != body.end() && logit_bias->is_array())
+//     // {
+//     //     const int n_vocab = llama_n_vocab(llama.model);
+//     //     for (const auto &el : *logit_bias)
+//     //     {
+//     //         if (el.is_array() && el.size() == 2 && el[0].is_number_integer())
+//     //         {
+//     //             llama_token tok = el[0].get<llama_token>();
+//     //             if (tok >= 0 && tok < n_vocab)
+//     //             {
+//     //                 if (el[1].is_number())
+//     //                 {
+//     //                     llama.params.logit_bias[tok] = el[1].get<float>();
+//     //                 }
+//     //                 else if (el[1].is_boolean() && !el[1].get<bool>())
+//     //                 {
+//     //                     llama.params.logit_bias[tok] = -INFINITY;
+//     //                 }
+//     //             }
+//     //         }
+//     //     }
+//     // }
+
+//     llama.params.antiprompt.clear();
+//     for (const std::string& stopPrompt : predict->stopprompts()) {
+//     if (!stopPrompt.empty())
+//             {
+//                 llama.params.antiprompt.push_back(stopPrompt);
+//             }
+//     }
+// }
 
 const std::vector<ggml_type> kv_cache_types = {
     GGML_TYPE_F32,
@@ -2258,7 +2558,7 @@ public:
         json data = parse_options(true, request, llama);
         const int task_id = llama.queue_tasks.get_new_id();
         llama.queue_results.add_waiting_task_id(task_id);
-        llama.request_completion(task_id, data, false, false, -1);
+        llama.request_completion(task_id, data, false, false, false, -1);
         while (true)
         {
             task_result result = llama.queue_results.recv(task_id);
@@ -2312,7 +2612,7 @@ public:
         json data = parse_options(false, request, llama);
         const int task_id = llama.queue_tasks.get_new_id();
         llama.queue_results.add_waiting_task_id(task_id);
-        llama.request_completion(task_id, data, false, false, -1);
+        llama.request_completion(task_id, data, false, false, false, -1);
         std::string completion_text;
         task_result result = llama.queue_results.recv(task_id);
         if (!result.error && result.stop) {
@@ -2349,7 +2649,7 @@ public:
         json data = parse_options(false, request, llama);
         const int task_id = llama.queue_tasks.get_new_id();
         llama.queue_results.add_waiting_task_id(task_id);
-        llama.request_completion(task_id, { {"prompt", data["embeddings"]}, { "n_predict", 0}, {"image_data", ""} }, false, true, -1);
+        llama.request_completion(task_id, { {"prompt", data["embeddings"]}, { "n_predict", 0}, {"image_data", ""} }, false, true, false, -1);
         // get the result
         task_result result = llama.queue_results.recv(task_id);
         //std::cout << "Embedding result JSON" << result.result_json.dump() << std::endl;
@@ -2372,15 +2672,55 @@ public:
     grpc::Status TokenizeString(ServerContext* context, const backend::PredictOptions* request, backend::TokenizationResponse* response){
          json data = parse_options(false, request, llama);
 
-         std::vector<server_tokens> tokens = llama.tokenize(data, data["prompt"],false);
+         std::vector<llama_token> tokens = llama.tokenize(data["prompt"],false);
 
          for (int i=0 ; i< tokens.size(); i++){
-            response->add_tokens(tokens[i].llama_token);
+            response->add_tokens(tokens[i]);
          }
 
         return grpc::Status::OK;
     }
 
+    grpc::Status Rerank(ServerContext* context, const backend::RerankRequest* request, backend::RerankResult* rerankResult) {
+        // Create a JSON object with the query and documents
+        json data = {
+            {"prompt", request->query()},
+            {"documents", request->documents()},
+            {"top_n", request->top_n()}
+        };
+
+        // Generate a new task ID
+        const int task_id = llama.queue_tasks.get_new_id();
+        llama.queue_results.add_waiting_task_id(task_id);
+
+        // Queue the task with reranking mode enabled
+        llama.request_completion(task_id, data, false, false, true, -1);
+
+        // Get the result
+        task_result result = llama.queue_results.recv(task_id);
+        llama.queue_results.remove_waiting_task_id(task_id);
+
+        if (!result.error && result.stop) {
+            // Set usage information
+            backend::Usage* usage = rerankResult->mutable_usage();
+            usage->set_total_tokens(result.result_json.value("tokens", 0));
+            usage->set_prompt_tokens(result.result_json.value("tokens", 0));
+
+            // Get the score from the result
+            float score = result.result_json.value("score", 0.0f);
+
+            // Create document results for each input document
+            for (int i = 0; i < request->documents_size(); i++) {
+                backend::DocumentResult* doc_result = rerankResult->add_results();
+                doc_result->set_index(i);
+                doc_result->set_text(request->documents(i));
+                doc_result->set_relevance_score(score);
+            }
+        }
+
+        return grpc::Status::OK;
+    }
+
     grpc::Status GetMetrics(ServerContext* context, const backend::MetricsRequest* request, backend::MetricsResponse* response) {
         llama_client_slot* active_slot = llama.get_active_slot();
 

backend/cpp/llama/patches/01-llava.patch

@@ -1,7 +1,7 @@
-diff --git a/tools/mtmd/clip.cpp b/tools/mtmd/clip.cpp
+diff --git a/examples/llava/clip.cpp b/examples/llava/clip.cpp
 index 3cd0d2fa..6c5e811a 100644
---- a/tools/mtmd/clip.cpp
-+++ b/tools/mtmd/clip.cpp
+--- a/examples/llava/clip.cpp
++++ b/examples/llava/clip.cpp
 @@ -2608,7 +2608,7 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
                  struct ggml_tensor * patches = ggml_graph_get_tensor(gf, "patches");
                  int* patches_data = (int*)malloc(ggml_nbytes(patches));

backend/cpp/llama/prepare.sh

@@ -7,22 +7,22 @@ for patch in $(ls patches); do
     patch -d llama.cpp/ -p1 < patches/$patch
 done 
 
-cp -r CMakeLists.txt llama.cpp/tools/grpc-server/
-cp -r grpc-server.cpp llama.cpp/tools/grpc-server/
-cp -rfv json.hpp llama.cpp/tools/grpc-server/
-cp -rfv utils.hpp llama.cpp/tools/grpc-server/
+cp -r CMakeLists.txt llama.cpp/examples/grpc-server/
+cp -r grpc-server.cpp llama.cpp/examples/grpc-server/
+cp -rfv json.hpp llama.cpp/examples/grpc-server/
+cp -rfv utils.hpp llama.cpp/examples/grpc-server/
     
-if grep -q "grpc-server" llama.cpp/tools/CMakeLists.txt; then
+if grep -q "grpc-server" llama.cpp/examples/CMakeLists.txt; then
     echo "grpc-server already added"
 else
-    echo "add_subdirectory(grpc-server)" >> llama.cpp/tools/CMakeLists.txt
+    echo "add_subdirectory(grpc-server)" >> llama.cpp/examples/CMakeLists.txt
 fi
 
 ## XXX: In some versions of CMake clip wasn't being built before llama.
 ## This is an hack for now, but it should be fixed in the future.
-# cp -rfv llama.cpp/tools/mtmd/clip.h llama.cpp/tools/grpc-server/clip.h
-# cp -rfv llama.cpp/tools/mtmd/clip-impl.h llama.cpp/tools/grpc-server/clip-impl.h
-# cp -rfv llama.cpp/tools/mtmd/llava.cpp llama.cpp/tools/grpc-server/llava.cpp
-# echo '#include "llama.h"' > llama.cpp/tools/grpc-server/llava.h
-# cat llama.cpp/tools/mtmd/llava.h >> llama.cpp/tools/grpc-server/llava.h
-# cp -rfv llama.cpp/tools/mtmd/clip.cpp llama.cpp/tools/grpc-server/clip.cpp
+cp -rfv llama.cpp/examples/llava/clip.h llama.cpp/examples/grpc-server/clip.h
+cp -rfv llama.cpp/examples/llava/clip-impl.h llama.cpp/examples/grpc-server/clip-impl.h
+cp -rfv llama.cpp/examples/llava/llava.cpp llama.cpp/examples/grpc-server/llava.cpp
+echo '#include "llama.h"' > llama.cpp/examples/grpc-server/llava.h
+cat llama.cpp/examples/llava/llava.h >> llama.cpp/examples/grpc-server/llava.h
+cp -rfv llama.cpp/examples/llava/clip.cpp llama.cpp/examples/grpc-server/clip.cpp

backend/cpp/llama/utils.hpp

@@ -1,4 +1,4 @@
-// https://github.com/ggerganov/llama.cpp/blob/master/tools/server/utils.hpp
+// https://github.com/ggerganov/llama.cpp/blob/master/examples/server/utils.hpp
 
 #pragma once
 
@@ -11,7 +11,7 @@
 
 #include "json.hpp"
 
-#include "../mtmd/clip.h"
+#include "../llava/clip.h"
 
 using json = nlohmann::json;
 
@@ -61,6 +61,7 @@ struct task_server {
     json data;
     bool infill_mode = false;
     bool embedding_mode = false;
+    bool reranking_mode = false;
     int multitask_id = -1;
 };
 
@@ -480,431 +481,4 @@ static inline std::vector<uint8_t> base64_decode(const std::string & encoded_str
     }
 
     return ret;
-
-}
-
-
-
-//
-// tokenizer and input processing utils
-//
-
-static bool json_is_array_of_numbers(const json & data) {
-    if (data.is_array()) {
-        for (const auto & e : data) {
-            if (!e.is_number_integer()) {
-                return false;
-            }
-        }
-        return true;
-    }
-    return false;
-}
-
-// is array having BOTH numbers & strings?
-static bool json_is_array_of_mixed_numbers_strings(const json & data) {
-    bool seen_string = false;
-    bool seen_number = false;
-    if (data.is_array()) {
-        for (const auto & e : data) {
-            seen_string |= e.is_string();
-            seen_number |= e.is_number_integer();
-            if (seen_number && seen_string) {
-                return true;
-            }
-        }
-    }
-    return false;
-}
-
-// get value by path(key1 / key2)
-static json json_get_nested_values(const std::vector<std::string> & paths, const json & js) {
-    json result = json::object();
-
-    for (const std::string & path : paths) {
-        json current = js;
-        const auto keys = string_split<std::string>(path, /*separator*/ '/');
-        bool valid_path = true;
-        for (const std::string & k : keys) {
-            if (valid_path && current.is_object() && current.contains(k)) {
-                current = current[k];
-            } else {
-                valid_path = false;
-            }
-        }
-        if (valid_path) {
-            result[path] = current;
-        }
-    }
-    return result;
-}
-
-
-/**
- * this handles 2 cases:
- * - only string, example: "string"
- * - mixed string and tokens, example: [12, 34, "string", 56, 78]
- */
-static llama_tokens tokenize_mixed(const llama_vocab * vocab, const json & json_prompt, bool add_special, bool parse_special) {
-    // If `add_bos` is true, we only add BOS, when json_prompt is a string,
-    // or the first element of the json_prompt array is a string.
-    llama_tokens prompt_tokens;
-
-    if (json_prompt.is_array()) {
-        bool first = true;
-        for (const auto & p : json_prompt) {
-            if (p.is_string()) {
-                auto s = p.template get<std::string>();
-
-                llama_tokens p;
-                if (first) {
-                    p = common_tokenize(vocab, s, add_special, parse_special);
-                    first = false;
-                } else {
-                    p = common_tokenize(vocab, s, false, parse_special);
-                }
-
-                prompt_tokens.insert(prompt_tokens.end(), p.begin(), p.end());
-            } else {
-                if (first) {
-                    first = false;
-                }
-
-                prompt_tokens.push_back(p.template get<llama_token>());
-            }
-        }
-    } else {
-        auto s = json_prompt.template get<std::string>();
-        prompt_tokens = common_tokenize(vocab, s, add_special, parse_special);
-    }
-
-    return prompt_tokens;
-}
-
-/**
- * break the input "prompt" object into multiple prompt if needed, then tokenize them
- * this supports these cases:
- * - "prompt": "string"
- * - "prompt": [12, 34, 56]
- * - "prompt": [12, 34, "string", 56, 78]
- * and multiple prompts (multi-tasks):
- * - "prompt": ["string1", "string2"]
- * - "prompt": ["string1", [12, 34, 56]]
- * - "prompt": [[12, 34, 56], [78, 90, 12]]
- * - "prompt": [[12, 34, "string", 56, 78], [12, 34, 56]]
- */
-static std::vector<llama_tokens> tokenize_input_prompts(const llama_vocab * vocab, const json & json_prompt, bool add_special, bool parse_special) {
-    std::vector<llama_tokens> result;
-    if (json_prompt.is_string() || json_is_array_of_mixed_numbers_strings(json_prompt)) {
-        // string or mixed
-        result.push_back(tokenize_mixed(vocab, json_prompt, add_special, parse_special));
-    } else if (json_is_array_of_numbers(json_prompt)) {
-        // array of tokens
-        result.push_back(json_prompt.get<llama_tokens>());
-    } else if (json_prompt.is_array()) {
-        // array of prompts
-        result.reserve(json_prompt.size());
-        for (const auto & p : json_prompt) {
-            if (p.is_string() || json_is_array_of_mixed_numbers_strings(p)) {
-                result.push_back(tokenize_mixed(vocab, p, add_special, parse_special));
-            } else if (json_is_array_of_numbers(p)) {
-                // array of tokens
-                result.push_back(p.get<llama_tokens>());
-            } else {
-                throw std::runtime_error("element of \"prompt\" must be a string, an list of tokens, or a list of mixed strings & tokens");
-            }
-        }
-    } else {
-        throw std::runtime_error("\"prompt\" must be a string, an list of tokens, a list of mixed strings & tokens, or a list of prompts");
-    }
-    if (result.empty()) {
-        throw std::runtime_error("\"prompt\" must not be empty");
-    }
-    return result;
-}
-
-
-
-
-//
-// utils for interacting with libmtmd
-// (may need to refactor in near future)
-//
-
-/**
- * server_tokens is a helper to manage the input tokens and image for the server.
- * it is made this way to simplify the logic of KV cache management.
- */
-struct server_tokens {
-    bool has_mtmd = false;
-
-private: // disallow accessing these members directly, risking out-of-sync
-
-    // map a **start** position in tokens to the image chunk
-    std::unordered_map<llama_pos, mtmd::input_chunk_ptr> map_pos_to_image;
-
-    // list of tokens
-    // it can include LLAMA_TOKEN_NULL, which is used to indicate a token that is not a text token
-    // a mtmd_input_chunk can occupy multiple tokens, one llama_token per **position**
-    // important: for models using mrope, an image can contain multiple tokens but will use only one **position**
-    llama_tokens tokens;
-
-    // for ex. with input of 5 text tokens and 2 images:
-    //      [0] [1] [2] [3] [4] [img0] [img0] [img0] [img1] [img1]
-    // pos  0   1   2   3   4   5      6      7      8      9
-    // map_pos_to_image will contain: {5, img0}, {8, img1}
-
-public:
-    server_tokens() = default;
-    ~server_tokens() = default;
-
-    // Prevent copying
-    server_tokens(const server_tokens&) = delete;
-    server_tokens& operator=(const server_tokens&) = delete;
-
-    // Allow moving (usually implicitly generated if members are movable)
-    server_tokens(server_tokens&&) = default;
-    server_tokens& operator=(server_tokens&&) = default;
-
-    // Allow accessing elements using [] operator
-    llama_token operator[](size_t index) { return tokens[index]; }
-    const llama_token& operator[](size_t index) const { return tokens[index]; }
-
-    server_tokens(mtmd::input_chunks & mtmd_chunks, bool has_mtmd) : has_mtmd(has_mtmd) {
-        for (size_t i = 0; i < mtmd_chunks.size(); ++i) {
-            push_back(mtmd_chunks[i]);
-        }
-    }
-
-    server_tokens(llama_tokens & tokens, bool has_mtmd) : has_mtmd(has_mtmd), tokens(tokens) {}
-
-    // for debugging
-    std::string str() const {
-        std::ostringstream oss;
-        oss << "tokens: ";
-        for (const auto & t : tokens) {
-            if (t == LLAMA_TOKEN_NULL) {
-                oss << "<embd> ";
-            } else {
-                oss << t << " ";
-            }
-        }
-        oss << "\n";
-        oss << "image pos: ";
-        for (const auto & it : map_pos_to_image) {
-            oss << it.first << ", ";
-        }
-        return oss.str();
-    }
-
-    const mtmd::input_chunk_ptr & find_chunk(llama_pos pos) const {
-        auto it = map_pos_to_image.find(pos);
-        if (it != map_pos_to_image.end()) {
-            return it->second;
-        } else {
-            throw std::runtime_error("Chunk not found");
-        }
-    }
-
-    void push_back(llama_token tok) {
-        if (tok == LLAMA_TOKEN_NULL) {
-            throw std::runtime_error("Invalid token");
-        }
-        tokens.emplace_back(tok);
-    }
-
-    // will create a copy of the chunk if it contains non-text data
-    void push_back(const mtmd_input_chunk * chunk) {
-        auto type = mtmd_input_chunk_get_type(chunk);
-        if (type == MTMD_INPUT_CHUNK_TYPE_IMAGE) {
-            GGML_ASSERT(has_mtmd);
-            auto img_tokens = mtmd_input_chunk_get_tokens_image(chunk);
-            const int n_pos = mtmd_image_tokens_get_n_pos(img_tokens);
-            llama_pos start_pos = tokens.size();
-            for (int i = 0; i < n_pos; ++i) {
-                tokens.emplace_back(LLAMA_TOKEN_NULL);
-            }
-            mtmd::input_chunk_ptr new_chunk(mtmd_input_chunk_copy(chunk));
-            map_pos_to_image[start_pos] = std::move(new_chunk);
-        } else if (type == MTMD_INPUT_CHUNK_TYPE_TEXT) {
-            size_t n_tokens;
-            auto text_tokens = mtmd_input_chunk_get_tokens_text(chunk, &n_tokens);
-            for (size_t i = 0; i < n_tokens; ++i) {
-                push_back(text_tokens[i]);
-            }
-        } else {
-            GGML_ABORT("Invalid chunk type");
-        }
-    }
-
-    // for compatibility with context shift and prompt truncation
-    void insert(const llama_tokens & inp_tokens) {
-        GGML_ASSERT(!has_mtmd); // only allow this if mtmd is disabled
-        tokens.insert(tokens.end(), inp_tokens.begin(), inp_tokens.end());
-    }
-
-    // for compatibility with speculative decoding, ctx shift, slot save/load
-    const llama_tokens & get_text_tokens() const {
-        GGML_ASSERT(!has_mtmd); // only allow this if mtmd is disabled
-        return tokens;
-    }
-
-    // for compatibility with speculative decoding
-    void set_token(llama_pos pos, llama_token id) {
-        GGML_ASSERT(!has_mtmd); // only allow this if mtmd is disabled
-        tokens[pos] = id;
-    }
-
-    size_t size() const {
-        return tokens.size();
-    }
-
-    bool empty() const {
-        return tokens.empty();
-    }
-
-    void clear() {
-        tokens.clear();
-    }
-
-    void resize(size_t n) {
-        GGML_ASSERT(n <= tokens.size());
-        if (has_mtmd) {
-            // we throw an error if we try to remove a token in the middle of an image
-            // for ex. with input of 5 text tokens and 2 images:
-            //    [0] [1] [2] [3] [4] [img0] [img0] [img0] [img1] [img1]
-            // n  1   2   3   4   5   6      7      8      9      10
-            // allowed to resize      ^                    ^
-            // disallowed to resize          ^      ^             ^
-            if (n > 0) {
-                llama_token last_token = tokens[n - 1];
-                // make sure we never remove tokens in the middle of an image
-                if (last_token == LLAMA_TOKEN_NULL) {
-                    find_chunk(n - 1); // will throw an error if the token is not begin-of-chunk
-                }
-            }
-            // remove all image chunks that are not used anymore
-            for (auto it = map_pos_to_image.begin(); it != map_pos_to_image.end(); ) {
-                llama_pos pos = it->first;
-                if (pos >= (llama_pos)n) {
-                    it = map_pos_to_image.erase(it);
-                } else {
-                    ++it;
-                }
-            }
-        }
-        tokens.resize(n);
-    }
-
-    std::string detokenize(const llama_context * ctx, bool special) const {
-        llama_tokens text_tokens;
-        text_tokens.reserve(tokens.size());
-        for (const auto & t : tokens) {
-            if (t != LLAMA_TOKEN_NULL) {
-                text_tokens.push_back(t);
-            }
-        }
-        return common_detokenize(ctx, text_tokens, special);
-    }
-
-    size_t get_common_prefix(const server_tokens & b) const {
-        size_t max_idx = std::min(tokens.size(), b.tokens.size());
-        for (size_t i = 0; i < max_idx; ++i) {
-            auto & ai =   tokens[i];
-            auto & bi = b.tokens[i];
-
-            if (ai == LLAMA_TOKEN_NULL && bi == LLAMA_TOKEN_NULL) {
-                GGML_ASSERT(has_mtmd);
-                const auto & a_chunk =   find_chunk(i);
-                const auto & b_chunk = b.find_chunk(i);
-                GGML_ASSERT(a_chunk && b_chunk);
-                const auto * a_img = mtmd_input_chunk_get_tokens_image(a_chunk.get());
-                const auto * b_img = mtmd_input_chunk_get_tokens_image(b_chunk.get());
-                std::string ai_id  = mtmd_image_tokens_get_id(a_img);
-                std::string bi_id  = mtmd_image_tokens_get_id(b_img);
-                size_t a_pos       = mtmd_image_tokens_get_n_pos(a_img);
-                size_t b_pos       = mtmd_image_tokens_get_n_pos(b_img);
-                if (ai_id == bi_id && a_pos == b_pos) {
-                    GGML_ASSERT(a_pos > 0 && "Invalid image token"); // should never happen
-                    i += a_pos - 1; // will be +1 by the for loop
-                    continue;
-                } else {
-                    return i;
-                }
-            } else if (ai == bi) {
-                continue;
-            } else {
-                return i;
-            }
-        }
-        return max_idx; // all tokens are equal
-    }
-
-    // make sure all text tokens are within the vocab range
-    bool validate(const struct llama_context * ctx) const {
-        const llama_model * model = llama_get_model(ctx);
-        const llama_vocab * vocab = llama_model_get_vocab(model);
-        const int32_t n_vocab = llama_vocab_n_tokens(vocab);
-
-        for (size_t i = 0; i < tokens.size(); ++i) {
-            auto & t = tokens[i];
-            if (t == LLAMA_TOKEN_NULL) {
-                try {
-                    const auto & chunk = find_chunk(i);
-                    const auto * img_tokens = mtmd_input_chunk_get_tokens_image(chunk.get());
-                    size_t n_pos = mtmd_image_tokens_get_n_pos(img_tokens);
-                    i += n_pos - 1; // will be +1 by the for loop
-                } catch (const std::exception & e) {
-                    return false;
-                }
-            } else if (t < 0 || t >= n_vocab) {
-                return false;
-            }
-        }
-        return true;
-    }
-
-    // encode and decode the image chunk
-    int32_t process_chunk(
-                llama_context * ctx,
-                mtmd_context * mctx,
-                llama_pos n_past,
-                int32_t seq_id,
-                llama_pos & n_pos_out) {
-        auto it = map_pos_to_image.find(n_past);
-        if (it == map_pos_to_image.end()) {
-            throw std::runtime_error("Chunk not found");
-        }
-     //   SRV_INF("%s\n", "processing image...");
-        int32_t n_batch = llama_n_batch(ctx);
-        int64_t t0 = ggml_time_ms();
-        llama_pos new_n_past = n_past;
-        int32_t result = mtmd_helper_eval_chunk_single(mctx, ctx,
-            it->second.get(), // chunk
-            n_past,
-            seq_id,
-            n_batch,
-            true, // logits last
-            &new_n_past);
-        //SRV_INF("image processed in %" PRId64 " ms\n", ggml_time_ms() - t0);
-        if (result != 0) {
-            LOG_ERR("mtmd_helper_eval failed with status %d", result);
-            n_pos_out = n_past;
-            return result;
-        }
-        n_pos_out = new_n_past;
-        return 0;
-    }
-};
-
-// Computes FNV-1a hash of the data
-static std::string fnv_hash(const uint8_t * data, size_t len) {
-    const uint64_t fnv_prime = 0x100000001b3ULL;
-    uint64_t hash = 0xcbf29ce484222325ULL;
-
-    for (size_t i = 0; i < len; ++i) {
-        hash ^= data[i];
-        hash *= fnv_prime;
-    }
-    return std::to_string(hash);
 }

backend/go/image/stablediffusion-ggml/Makefile

@@ -20,7 +20,7 @@ CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
 
 # If build type is cublas, then we set -DGGML_CUDA=ON to CMAKE_ARGS automatically
 ifeq ($(BUILD_TYPE),cublas)
-	CMAKE_ARGS+=-DSD_CUDA=ON
+	CMAKE_ARGS+=-DGGML_CUDA=ON
 # If build type is openblas then we set -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
 # to CMAKE_ARGS automatically
 else ifeq ($(BUILD_TYPE),openblas)
@@ -30,14 +30,14 @@ else ifeq ($(BUILD_TYPE),clblas)
 	CMAKE_ARGS+=-DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
 # If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++
 else ifeq ($(BUILD_TYPE),hipblas)
-	CMAKE_ARGS+=-DSD_HIPBLAS=ON
+	CMAKE_ARGS+=-DGGML_HIP=ON
 # If it's OSX, DO NOT embed the metal library - -DGGML_METAL_EMBED_LIBRARY=ON requires further investigation
 # But if it's OSX without metal, disable it here
 else ifeq ($(OS),Darwin)
 	ifneq ($(BUILD_TYPE),metal)
-		CMAKE_ARGS+=-DSD_METAL=OFF
+		CMAKE_ARGS+=-DGGML_METAL=OFF
 	else
-		CMAKE_ARGS+=-DSD_METAL=ON
+		CMAKE_ARGS+=-DGGML_METAL=ON
 		CMAKE_ARGS+=-DGGML_METAL_EMBED_LIBRARY=ON
 		TARGET+=--target ggml-metal
 	endif

backend/go/transcribe/whisper/whisper.go

@@ -74,7 +74,7 @@ func (sd *Whisper) AudioTranscription(opts *pb.TranscriptRequest) (pb.Transcript
 		context.SetTranslate(true)
 	}
 
-	if err := context.Process(data, nil, nil, nil); err != nil {
+	if err := context.Process(data, nil, nil); err != nil {
 		return pb.TranscriptResult{}, err
 	}
 

backend/python/bark/requirements.txt

@@ -1,4 +1,4 @@
 bark==0.1.5
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 certifi

backend/python/common/template/requirements.txt

@@ -1,3 +1,3 @@
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 grpcio-tools

backend/python/coqui/requirements.txt

@@ -1,4 +1,4 @@
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 certifi
 packaging==24.1

backend/python/diffusers/backend.py

@@ -168,13 +168,9 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             # We are storing all the options in a dict so we can use it later when
             # generating the images
             for opt in options:
-                if ":" not in opt:
-                    continue
                 key, value = opt.split(":")
                 self.options[key] = value
 
-            print(f"Options: {self.options}", file=sys.stderr)
-
             local = False
             modelFile = request.Model
 

backend/python/diffusers/requirements.txt

@@ -1,5 +1,5 @@
 setuptools
-grpcio==1.72.0
+grpcio==1.71.0
 pillow
 protobuf
 certifi

backend/python/exllama2/requirements.txt

@@ -1,4 +1,4 @@
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 certifi
 wheel

backend/python/faster-whisper/requirements.txt

@@ -1,3 +1,3 @@
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 grpcio-tools

backend/python/kokoro/requirements.txt

@@ -1,4 +1,4 @@
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 phonemizer
 scipy

backend/python/rerankers/requirements.txt

@@ -1,3 +1,3 @@
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 certifi

backend/python/transformers/requirements.txt

@@ -1,4 +1,4 @@
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 certifi
 setuptools

backend/python/vllm/backend.py

@@ -194,40 +194,27 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             await iterations.aclose()
 
     async def _predict(self, request, context, streaming=False):
-        # Build the sampling parameters
-        # NOTE: this must stay in sync with the vllm backend
-        request_to_sampling_params = {
-            "N": "n",
-            "PresencePenalty": "presence_penalty",
-            "FrequencyPenalty": "frequency_penalty",
-            "RepetitionPenalty": "repetition_penalty",
-            "Temperature": "temperature",
-            "TopP": "top_p",
-            "TopK": "top_k",
-            "MinP": "min_p",
-            "Seed": "seed",
-            "StopPrompts": "stop",
-            "StopTokenIds": "stop_token_ids",
-            "BadWords": "bad_words",
-            "IncludeStopStrInOutput": "include_stop_str_in_output",
-            "IgnoreEOS": "ignore_eos",
-            "Tokens": "max_tokens",
-            "MinTokens": "min_tokens",
-            "Logprobs": "logprobs",
-            "PromptLogprobs": "prompt_logprobs",
-            "SkipSpecialTokens": "skip_special_tokens",
-            "SpacesBetweenSpecialTokens": "spaces_between_special_tokens",
-            "TruncatePromptTokens": "truncate_prompt_tokens",
-            "GuidedDecoding": "guided_decoding",
-        }
 
+        # Build sampling parameters
         sampling_params = SamplingParams(top_p=0.9, max_tokens=200)
-
-        for request_field, param_field in request_to_sampling_params.items():
-            if hasattr(request, request_field):
-                value = getattr(request, request_field)
-                if value not in (None, 0, [], False):
-                    setattr(sampling_params, param_field, value)
+        if request.TopP != 0:
+            sampling_params.top_p = request.TopP
+        if request.Tokens > 0:
+            sampling_params.max_tokens = request.Tokens
+        if request.Temperature != 0:
+            sampling_params.temperature = request.Temperature
+        if request.TopK != 0:
+            sampling_params.top_k = request.TopK
+        if request.PresencePenalty != 0:
+            sampling_params.presence_penalty = request.PresencePenalty
+        if request.FrequencyPenalty != 0:
+            sampling_params.frequency_penalty = request.FrequencyPenalty
+        if request.StopPrompts:
+            sampling_params.stop = request.StopPrompts
+        if request.IgnoreEOS:
+            sampling_params.ignore_eos = request.IgnoreEOS
+        if request.Seed != 0:
+            sampling_params.seed = request.Seed
 
         # Extract image paths and process images
         prompt = request.Prompt

backend/python/vllm/requirements.txt

@@ -1,4 +1,4 @@
-grpcio==1.72.0
+grpcio==1.71.0
 protobuf
 certifi
 setuptools

backend/python/vllm/test.py

@@ -75,53 +75,6 @@ class TestBackendServicer(unittest.TestCase):
         finally:
             self.tearDown()
 
-    def test_sampling_params(self):
-        """
-        This method tests if all sampling parameters are correctly processed
-        NOTE: this does NOT test for correctness, just that we received a compatible response
-        """
-        try:
-            self.setUp()
-            with grpc.insecure_channel("localhost:50051") as channel:
-                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(Model="facebook/opt-125m"))
-                self.assertTrue(response.success)
-
-                req = backend_pb2.PredictOptions(
-                    Prompt="The capital of France is",
-                    TopP=0.8,
-                    Tokens=50,
-                    Temperature=0.7,
-                    TopK=40,
-                    PresencePenalty=0.1,
-                    FrequencyPenalty=0.2,
-                    RepetitionPenalty=1.1,
-                    MinP=0.05,
-                    Seed=42,
-                    StopPrompts=["\n"],
-                    StopTokenIds=[50256],
-                    BadWords=["badword"],
-                    IncludeStopStrInOutput=True,
-                    IgnoreEOS=True,
-                    MinTokens=5,
-                    Logprobs=5,
-                    PromptLogprobs=5,
-                    SkipSpecialTokens=True,
-                    SpacesBetweenSpecialTokens=True,
-                    TruncatePromptTokens=10,
-                    GuidedDecoding=True,
-                    N=2,
-                )
-                resp = stub.Predict(req)
-                self.assertIsNotNone(resp.message)
-                self.assertIsNotNone(resp.logprobs)
-        except Exception as err:
-            print(err)
-            self.fail("sampling params service failed")
-        finally:
-            self.tearDown()
-
-
     def test_embedding(self):
         """
         This method tests if the embeddings are generated successfully

core/application/startup.go

@@ -43,12 +43,18 @@ func New(opts ...config.AppOption) (*Application, error) {
 	if err != nil {
 		return nil, fmt.Errorf("unable to create ModelPath: %q", err)
 	}
-	if options.GeneratedContentDir != "" {
-		err := os.MkdirAll(options.GeneratedContentDir, 0750)
+	if options.ImageDir != "" {
+		err := os.MkdirAll(options.ImageDir, 0750)
 		if err != nil {
 			return nil, fmt.Errorf("unable to create ImageDir: %q", err)
 		}
 	}
+	if options.AudioDir != "" {
+		err := os.MkdirAll(options.AudioDir, 0750)
+		if err != nil {
+			return nil, fmt.Errorf("unable to create AudioDir: %q", err)
+		}
+	}
 	if options.UploadDir != "" {
 		err := os.MkdirAll(options.UploadDir, 0750)
 		if err != nil {

core/backend/options.go

@@ -99,7 +99,7 @@ func grpcModelOpts(c config.BackendConfig) *pb.ModelOptions {
 		mmap = *c.MMap
 	}
 
-	ctxSize := 4096
+	ctxSize := 1024
 	if c.ContextSize != nil {
 		ctxSize = *c.ContextSize
 	}

core/backend/soundgeneration.go

@@ -35,17 +35,12 @@ func SoundGeneration(
 		return "", nil, fmt.Errorf("could not load sound generation model")
 	}
 
-	if err := os.MkdirAll(appConfig.GeneratedContentDir, 0750); err != nil {
+	if err := os.MkdirAll(appConfig.AudioDir, 0750); err != nil {
 		return "", nil, fmt.Errorf("failed creating audio directory: %s", err)
 	}
 
-	audioDir := filepath.Join(appConfig.GeneratedContentDir, "audio")
-	if err := os.MkdirAll(audioDir, 0750); err != nil {
-		return "", nil, fmt.Errorf("failed creating audio directory: %s", err)
-	}
-
-	fileName := utils.GenerateUniqueFileName(audioDir, "sound_generation", ".wav")
-	filePath := filepath.Join(audioDir, fileName)
+	fileName := utils.GenerateUniqueFileName(appConfig.AudioDir, "sound_generation", ".wav")
+	filePath := filepath.Join(appConfig.AudioDir, fileName)
 
 	res, err := soundGenModel.SoundGeneration(context.Background(), &proto.SoundGenerationRequest{
 		Text:        text,

core/backend/tts.go

@@ -32,13 +32,12 @@ func ModelTTS(
 		return "", nil, fmt.Errorf("could not load tts model %q", backendConfig.Model)
 	}
 
-	audioDir := filepath.Join(appConfig.GeneratedContentDir, "audio")
-	if err := os.MkdirAll(audioDir, 0750); err != nil {
+	if err := os.MkdirAll(appConfig.AudioDir, 0750); err != nil {
 		return "", nil, fmt.Errorf("failed creating audio directory: %s", err)
 	}
 
-	fileName := utils.GenerateUniqueFileName(audioDir, "tts", ".wav")
-	filePath := filepath.Join(audioDir, fileName)
+	fileName := utils.GenerateUniqueFileName(appConfig.AudioDir, "tts", ".wav")
+	filePath := filepath.Join(appConfig.AudioDir, fileName)
 
 	// We join the model name to the model path here. This seems to only be done for TTS and is HIGHLY suspect.
 	// This should be addressed in a follow up PR soon.

core/backend/video.go

@@ -1,36 +0,0 @@
-package backend
-
-import (
-	"github.com/mudler/LocalAI/core/config"
-
-	"github.com/mudler/LocalAI/pkg/grpc/proto"
-	model "github.com/mudler/LocalAI/pkg/model"
-)
-
-func VideoGeneration(height, width int32, prompt, startImage, endImage, dst string, loader *model.ModelLoader, backendConfig config.BackendConfig, appConfig *config.ApplicationConfig) (func() error, error) {
-
-	opts := ModelOptions(backendConfig, appConfig)
-	inferenceModel, err := loader.Load(
-		opts...,
-	)
-	if err != nil {
-		return nil, err
-	}
-	defer loader.Close()
-
-	fn := func() error {
-		_, err := inferenceModel.GenerateVideo(
-			appConfig.Context,
-			&proto.GenerateVideoRequest{
-				Height:     height,
-				Width:      width,
-				Prompt:     prompt,
-				StartImage: startImage,
-				EndImage:   endImage,
-				Dst:        dst,
-			})
-		return err
-	}
-
-	return fn, nil
-}

core/cli/context/context.go

@@ -1,13 +1,11 @@
 package cliContext
 
-import (
-	rice "github.com/GeertJohan/go.rice"
-)
+import "embed"
 
 type Context struct {
 	Debug    bool    `env:"LOCALAI_DEBUG,DEBUG" default:"false" hidden:"" help:"DEPRECATED, use --log-level=debug instead. Enable debug logging"`
 	LogLevel *string `env:"LOCALAI_LOG_LEVEL" enum:"error,warn,info,debug,trace" help:"Set the level of logs to output [${enum}]"`
 
 	// This field is not a command line argument/flag, the struct tag excludes it from the parsed CLI
-	BackendAssets *rice.Box `kong:"-"`
+	BackendAssets embed.FS `kong:"-"`
 }

core/cli/run.go

@@ -21,7 +21,8 @@ type RunCMD struct {
 
 	ModelsPath                   string        `env:"LOCALAI_MODELS_PATH,MODELS_PATH" type:"path" default:"${basepath}/models" help:"Path containing models used for inferencing" group:"storage"`
 	BackendAssetsPath            string        `env:"LOCALAI_BACKEND_ASSETS_PATH,BACKEND_ASSETS_PATH" type:"path" default:"/tmp/localai/backend_data" help:"Path used to extract libraries that are required by some of the backends in runtime" group:"storage"`
-	GeneratedContentPath         string        `env:"LOCALAI_GENERATED_CONTENT_PATH,GENERATED_CONTENT_PATH" type:"path" default:"/tmp/generated/content" help:"Location for generated content (e.g. images, audio, videos)" group:"storage"`
+	ImagePath                    string        `env:"LOCALAI_IMAGE_PATH,IMAGE_PATH" type:"path" default:"/tmp/generated/images" help:"Location for images generated by backends (e.g. stablediffusion)" group:"storage"`
+	AudioPath                    string        `env:"LOCALAI_AUDIO_PATH,AUDIO_PATH" type:"path" default:"/tmp/generated/audio" help:"Location for audio generated by backends (e.g. piper)" group:"storage"`
 	UploadPath                   string        `env:"LOCALAI_UPLOAD_PATH,UPLOAD_PATH" type:"path" default:"/tmp/localai/upload" help:"Path to store uploads from files api" group:"storage"`
 	ConfigPath                   string        `env:"LOCALAI_CONFIG_PATH,CONFIG_PATH" default:"/tmp/localai/config" group:"storage"`
 	LocalaiConfigDir             string        `env:"LOCALAI_CONFIG_DIR" type:"path" default:"${basepath}/configuration" help:"Directory for dynamic loading of certain configuration files (currently api_keys.json and external_backends.json)" group:"storage"`
@@ -46,7 +47,7 @@ type RunCMD struct {
 	CSRF                               bool     `env:"LOCALAI_CSRF" help:"Enables fiber CSRF middleware" group:"api"`
 	UploadLimit                        int      `env:"LOCALAI_UPLOAD_LIMIT,UPLOAD_LIMIT" default:"15" help:"Default upload-limit in MB" group:"api"`
 	APIKeys                            []string `env:"LOCALAI_API_KEY,API_KEY" help:"List of API Keys to enable API authentication. When this is set, all the requests must be authenticated with one of these API keys" group:"api"`
-	DisableWebUI                       bool     `env:"LOCALAI_DISABLE_WEBUI,DISABLE_WEBUI" default:"false" help:"Disables the web user interface. When set to true, the server will only expose API endpoints without serving the web interface" group:"api"`
+	DisableWebUI                       bool     `env:"LOCALAI_DISABLE_WEBUI,DISABLE_WEBUI" default:"false" help:"Disable webui" group:"api"`
 	DisablePredownloadScan             bool     `env:"LOCALAI_DISABLE_PREDOWNLOAD_SCAN" help:"If true, disables the best-effort security scanner before downloading any files." group:"hardening" default:"false"`
 	OpaqueErrors                       bool     `env:"LOCALAI_OPAQUE_ERRORS" default:"false" help:"If true, all error responses are replaced with blank 500 errors. This is intended only for hardening against information leaks and is normally not recommended." group:"hardening"`
 	UseSubtleKeyComparison             bool     `env:"LOCALAI_SUBTLE_KEY_COMPARISON" default:"false" help:"If true, API Key validation comparisons will be performed using constant-time comparisons rather than simple equality. This trades off performance on each request for resiliancy against timing attacks." group:"hardening"`
@@ -80,7 +81,8 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 		config.WithModelPath(r.ModelsPath),
 		config.WithContextSize(r.ContextSize),
 		config.WithDebug(zerolog.GlobalLevel() <= zerolog.DebugLevel),
-		config.WithGeneratedContentDir(r.GeneratedContentPath),
+		config.WithImageDir(r.ImagePath),
+		config.WithAudioDir(r.AudioPath),
 		config.WithUploadDir(r.UploadPath),
 		config.WithConfigsDir(r.ConfigPath),
 		config.WithDynamicConfigDir(r.LocalaiConfigDir),

core/cli/soundgeneration.go

@@ -70,7 +70,7 @@ func (t *SoundGenerationCMD) Run(ctx *cliContext.Context) error {
 	opts := &config.ApplicationConfig{
 		ModelPath:            t.ModelsPath,
 		Context:              context.Background(),
-		GeneratedContentDir:  outputDir,
+		AudioDir:             outputDir,
 		AssetsDestination:    t.BackendAssetsPath,
 		ExternalGRPCBackends: externalBackends,
 	}

core/cli/tts.go

@@ -36,10 +36,10 @@ func (t *TTSCMD) Run(ctx *cliContext.Context) error {
 	text := strings.Join(t.Text, " ")
 
 	opts := &config.ApplicationConfig{
-		ModelPath:           t.ModelsPath,
-		Context:             context.Background(),
-		GeneratedContentDir: outputDir,
-		AssetsDestination:   t.BackendAssetsPath,
+		ModelPath:         t.ModelsPath,
+		Context:           context.Background(),
+		AudioDir:          outputDir,
+		AssetsDestination: t.BackendAssetsPath,
 	}
 	ml := model.NewModelLoader(opts.ModelPath, opts.SingleBackend)
 

core/cli/util.go

@@ -7,11 +7,11 @@ import (
 
 	"github.com/rs/zerolog/log"
 
-	gguf "github.com/gpustack/gguf-parser-go"
 	cliContext "github.com/mudler/LocalAI/core/cli/context"
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/gallery"
 	"github.com/mudler/LocalAI/pkg/downloader"
+	gguf "github.com/thxcode/gguf-parser-go"
 )
 
 type UtilCMD struct {
@@ -51,7 +51,7 @@ func (u *GGUFInfoCMD) Run(ctx *cliContext.Context) error {
 	log.Info().
 		Any("eosTokenID", f.Tokenizer().EOSTokenID).
 		Any("bosTokenID", f.Tokenizer().BOSTokenID).
-		Any("modelName", f.Metadata().Name).
+		Any("modelName", f.Model().Name).
 		Any("architecture", f.Architecture().Architecture).Msgf("GGUF file loaded: %s", u.Args[0])
 
 	log.Info().Any("tokenizer", fmt.Sprintf("%+v", f.Tokenizer())).Msg("Tokenizer")

core/config/application_config.go

@@ -2,11 +2,11 @@ package config
 
 import (
 	"context"
+	"embed"
 	"encoding/json"
 	"regexp"
 	"time"
 
-	rice "github.com/GeertJohan/go.rice"
 	"github.com/mudler/LocalAI/pkg/xsysinfo"
 	"github.com/rs/zerolog/log"
 )
@@ -19,21 +19,20 @@ type ApplicationConfig struct {
 	UploadLimitMB, Threads, ContextSize int
 	F16                                 bool
 	Debug                               bool
-	GeneratedContentDir                 string
-
-	ConfigsDir string
-	UploadDir  string
-
-	DynamicConfigsDir             string
-	DynamicConfigsDirPollInterval time.Duration
-	CORS                          bool
-	CSRF                          bool
-	PreloadJSONModels             string
-	PreloadModelsFromPath         string
-	CORSAllowOrigins              string
-	ApiKeys                       []string
-	P2PToken                      string
-	P2PNetworkID                  string
+	ImageDir                            string
+	AudioDir                            string
+	UploadDir                           string
+	ConfigsDir                          string
+	DynamicConfigsDir                   string
+	DynamicConfigsDirPollInterval       time.Duration
+	CORS                                bool
+	CSRF                                bool
+	PreloadJSONModels                   string
+	PreloadModelsFromPath               string
+	CORSAllowOrigins                    string
+	ApiKeys                             []string
+	P2PToken                            string
+	P2PNetworkID                        string
 
 	DisableWebUI                       bool
 	EnforcePredownloadScans            bool
@@ -47,7 +46,7 @@ type ApplicationConfig struct {
 
 	Galleries []Gallery
 
-	BackendAssets     *rice.Box
+	BackendAssets     embed.FS
 	AssetsDestination string
 
 	ExternalGRPCBackends map[string]string
@@ -198,7 +197,7 @@ func WithBackendAssetsOutput(out string) AppOption {
 	}
 }
 
-func WithBackendAssets(f *rice.Box) AppOption {
+func WithBackendAssets(f embed.FS) AppOption {
 	return func(o *ApplicationConfig) {
 		o.BackendAssets = f
 	}
@@ -280,9 +279,15 @@ func WithDebug(debug bool) AppOption {
 	}
 }
 
-func WithGeneratedContentDir(generatedContentDir string) AppOption {
+func WithAudioDir(audioDir string) AppOption {
 	return func(o *ApplicationConfig) {
-		o.GeneratedContentDir = generatedContentDir
+		o.AudioDir = audioDir
+	}
+}
+
+func WithImageDir(imageDir string) AppOption {
+	return func(o *ApplicationConfig) {
+		o.ImageDir = imageDir
 	}
 }
 

core/config/backend_config.go

@@ -304,6 +304,9 @@ func (cfg *BackendConfig) SetDefaults(opts ...ConfigLoaderOption) {
 	defaultTFZ := 1.0
 	defaultZero := 0
 
+	// Try to offload all GPU layers (if GPU is found)
+	defaultHigh := 99999999
+
 	trueV := true
 	falseV := false
 
@@ -363,6 +366,9 @@ func (cfg *BackendConfig) SetDefaults(opts ...ConfigLoaderOption) {
 	if cfg.MirostatTAU == nil {
 		cfg.MirostatTAU = &defaultMirostatTAU
 	}
+	if cfg.NGPULayers == nil {
+		cfg.NGPULayers = &defaultHigh
+	}
 
 	if cfg.LowVRAM == nil {
 		cfg.LowVRAM = &falseV
@@ -430,19 +436,18 @@ func (c *BackendConfig) HasTemplate() bool {
 type BackendConfigUsecases int
 
 const (
-	FLAG_ANY              BackendConfigUsecases = 0b000000000000
-	FLAG_CHAT             BackendConfigUsecases = 0b000000000001
-	FLAG_COMPLETION       BackendConfigUsecases = 0b000000000010
-	FLAG_EDIT             BackendConfigUsecases = 0b000000000100
-	FLAG_EMBEDDINGS       BackendConfigUsecases = 0b000000001000
-	FLAG_RERANK           BackendConfigUsecases = 0b000000010000
-	FLAG_IMAGE            BackendConfigUsecases = 0b000000100000
-	FLAG_TRANSCRIPT       BackendConfigUsecases = 0b000001000000
-	FLAG_TTS              BackendConfigUsecases = 0b000010000000
-	FLAG_SOUND_GENERATION BackendConfigUsecases = 0b000100000000
-	FLAG_TOKENIZE         BackendConfigUsecases = 0b001000000000
-	FLAG_VAD              BackendConfigUsecases = 0b010000000000
-	FLAG_VIDEO            BackendConfigUsecases = 0b100000000000
+	FLAG_ANY              BackendConfigUsecases = 0b00000000000
+	FLAG_CHAT             BackendConfigUsecases = 0b00000000001
+	FLAG_COMPLETION       BackendConfigUsecases = 0b00000000010
+	FLAG_EDIT             BackendConfigUsecases = 0b00000000100
+	FLAG_EMBEDDINGS       BackendConfigUsecases = 0b00000001000
+	FLAG_RERANK           BackendConfigUsecases = 0b00000010000
+	FLAG_IMAGE            BackendConfigUsecases = 0b00000100000
+	FLAG_TRANSCRIPT       BackendConfigUsecases = 0b00001000000
+	FLAG_TTS              BackendConfigUsecases = 0b00010000000
+	FLAG_SOUND_GENERATION BackendConfigUsecases = 0b00100000000
+	FLAG_TOKENIZE         BackendConfigUsecases = 0b01000000000
+	FLAG_VAD              BackendConfigUsecases = 0b10000000000
 
 	// Common Subsets
 	FLAG_LLM BackendConfigUsecases = FLAG_CHAT | FLAG_COMPLETION | FLAG_EDIT
@@ -463,7 +468,6 @@ func GetAllBackendConfigUsecases() map[string]BackendConfigUsecases {
 		"FLAG_TOKENIZE":         FLAG_TOKENIZE,
 		"FLAG_VAD":              FLAG_VAD,
 		"FLAG_LLM":              FLAG_LLM,
-		"FLAG_VIDEO":            FLAG_VIDEO,
 	}
 }
 
@@ -528,17 +532,6 @@ func (c *BackendConfig) GuessUsecases(u BackendConfigUsecases) bool {
 			return false
 		}
 
-	}
-	if (u & FLAG_VIDEO) == FLAG_VIDEO {
-		videoBackends := []string{"diffusers", "stablediffusion"}
-		if !slices.Contains(videoBackends, c.Backend) {
-			return false
-		}
-
-		if c.Backend == "diffusers" && c.Diffusers.PipelineType == "" {
-			return false
-		}
-
 	}
 	if (u & FLAG_RERANK) == FLAG_RERANK {
 		if c.Backend != "rerankers" {

core/config/gguf.go

@@ -3,10 +3,9 @@ package config
 import (
 	"strings"
 
-	"github.com/mudler/LocalAI/pkg/xsysinfo"
 	"github.com/rs/zerolog/log"
 
-	gguf "github.com/gpustack/gguf-parser-go"
+	gguf "github.com/thxcode/gguf-parser-go"
 )
 
 type familyType uint8
@@ -24,7 +23,6 @@ const (
 
 const (
 	defaultContextSize = 1024
-	defaultNGPULayers  = 99999999
 )
 
 type settingsConfig struct {
@@ -149,7 +147,7 @@ var knownTemplates = map[string]familyType{
 func guessGGUFFromFile(cfg *BackendConfig, f *gguf.GGUFFile, defaultCtx int) {
 
 	if defaultCtx == 0 && cfg.ContextSize == nil {
-		ctxSize := f.EstimateLLaMACppRun().ContextSize
+		ctxSize := f.EstimateLLaMACppUsage().ContextSize
 		if ctxSize > 0 {
 			cSize := int(ctxSize)
 			cfg.ContextSize = &cSize
@@ -159,46 +157,6 @@ func guessGGUFFromFile(cfg *BackendConfig, f *gguf.GGUFFile, defaultCtx int) {
 		}
 	}
 
-	// GPU options
-	if cfg.Options == nil {
-		if xsysinfo.HasGPU("nvidia") || xsysinfo.HasGPU("amd") {
-			cfg.Options = []string{"gpu"}
-		}
-	}
-
-	// vram estimation
-	vram, err := xsysinfo.TotalAvailableVRAM()
-	if err != nil {
-		log.Error().Msgf("guessDefaultsFromFile(TotalAvailableVRAM): %s", err)
-	} else if vram > 0 {
-		estimate, err := xsysinfo.EstimateGGUFVRAMUsage(f, vram)
-		if err != nil {
-			log.Error().Msgf("guessDefaultsFromFile(EstimateGGUFVRAMUsage): %s", err)
-		} else {
-			if estimate.IsFullOffload {
-				log.Warn().Msgf("guessDefaultsFromFile: %s", "full offload is recommended")
-			}
-
-			if estimate.EstimatedVRAM > vram {
-				log.Warn().Msgf("guessDefaultsFromFile: %s", "estimated VRAM usage is greater than available VRAM")
-			}
-
-			if cfg.NGPULayers == nil && estimate.EstimatedLayers > 0 {
-				log.Debug().Msgf("guessDefaultsFromFile: %d layers estimated", estimate.EstimatedLayers)
-				cfg.NGPULayers = &estimate.EstimatedLayers
-			}
-		}
-	}
-
-	if cfg.NGPULayers == nil {
-		// we assume we want to offload all layers
-		defaultHigh := defaultNGPULayers
-		cfg.NGPULayers = &defaultHigh
-	}
-
-	log.Debug().Any("NGPULayers", cfg.NGPULayers).Msgf("guessDefaultsFromFile: %s", "NGPULayers set")
-
-	// template estimations
 	if cfg.HasTemplate() {
 		// nothing to guess here
 		log.Debug().Any("name", cfg.Name).Msgf("guessDefaultsFromFile: %s", "template already set")
@@ -208,12 +166,12 @@ func guessGGUFFromFile(cfg *BackendConfig, f *gguf.GGUFFile, defaultCtx int) {
 	log.Debug().
 		Any("eosTokenID", f.Tokenizer().EOSTokenID).
 		Any("bosTokenID", f.Tokenizer().BOSTokenID).
-		Any("modelName", f.Metadata().Name).
+		Any("modelName", f.Model().Name).
 		Any("architecture", f.Architecture().Architecture).Msgf("Model file loaded: %s", cfg.ModelFileName())
 
 	// guess the name
 	if cfg.Name == "" {
-		cfg.Name = f.Metadata().Name
+		cfg.Name = f.Model().Name
 	}
 
 	family := identifyFamily(f)
@@ -249,7 +207,6 @@ func guessGGUFFromFile(cfg *BackendConfig, f *gguf.GGUFFile, defaultCtx int) {
 		cfg.TemplateConfig.JinjaTemplate = true
 		cfg.TemplateConfig.ChatMessage = chatTemplate.ValueString()
 	}
-
 }
 
 func identifyFamily(f *gguf.GGUFFile) familyType {
@@ -274,7 +231,7 @@ func identifyFamily(f *gguf.GGUFFile) familyType {
 	commandR := arch == "command-r" && eosTokenID == 255001
 	qwen2 := arch == "qwen2"
 	phi3 := arch == "phi-3"
-	gemma := strings.HasPrefix(arch, "gemma") || strings.Contains(strings.ToLower(f.Metadata().Name), "gemma")
+	gemma := strings.HasPrefix(arch, "gemma") || strings.Contains(strings.ToLower(f.Model().Name), "gemma")
 	deepseek2 := arch == "deepseek2"
 
 	switch {

core/config/guesser.go

@@ -4,8 +4,8 @@ import (
 	"os"
 	"path/filepath"
 
-	gguf "github.com/gpustack/gguf-parser-go"
 	"github.com/rs/zerolog/log"
+	gguf "github.com/thxcode/gguf-parser-go"
 )
 
 func guessDefaultsFromFile(cfg *BackendConfig, modelPath string, defaultCtx int) {

core/http/app.go

@@ -5,8 +5,6 @@ import (
 	"errors"
 	"fmt"
 	"net/http"
-	"os"
-	"path/filepath"
 
 	"github.com/dave-gray101/v2keyauth"
 	"github.com/mudler/LocalAI/pkg/utils"
@@ -155,19 +153,12 @@ func API(application *application.Application) (*fiber.App, error) {
 		Browse:     true,
 	}))
 
-	if application.ApplicationConfig().GeneratedContentDir != "" {
-		os.MkdirAll(application.ApplicationConfig().GeneratedContentDir, 0750)
-		audioPath := filepath.Join(application.ApplicationConfig().GeneratedContentDir, "audio")
-		imagePath := filepath.Join(application.ApplicationConfig().GeneratedContentDir, "images")
-		videoPath := filepath.Join(application.ApplicationConfig().GeneratedContentDir, "videos")
+	if application.ApplicationConfig().ImageDir != "" {
+		router.Static("/generated-images", application.ApplicationConfig().ImageDir)
+	}
 
-		os.MkdirAll(audioPath, 0750)
-		os.MkdirAll(imagePath, 0750)
-		os.MkdirAll(videoPath, 0750)
-
-		router.Static("/generated-audio", audioPath)
-		router.Static("/generated-images", imagePath)
-		router.Static("/generated-videos", videoPath)
+	if application.ApplicationConfig().AudioDir != "" {
+		router.Static("/generated-audio", application.ApplicationConfig().AudioDir)
 	}
 
 	// Auth is applied to _all_ endpoints. No exceptions. Filtering out endpoints to bypass is the role of the Filter property of the KeyAuth Configuration

core/http/app_test.go

@@ -3,6 +3,7 @@ package http_test
 import (
 	"bytes"
 	"context"
+	"embed"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -23,7 +24,6 @@ import (
 	. "github.com/onsi/gomega"
 	"gopkg.in/yaml.v3"
 
-	rice "github.com/GeertJohan/go.rice"
 	openaigo "github.com/otiai10/openaigo"
 	"github.com/sashabaranov/go-openai"
 	"github.com/sashabaranov/go-openai/jsonschema"
@@ -264,15 +264,8 @@ func getRequest(url string, header http.Header) (error, int, []byte) {
 
 const bertEmbeddingsURL = `https://gist.githubusercontent.com/mudler/0a080b166b87640e8644b09c2aee6e3b/raw/f0e8c26bb72edc16d9fbafbfd6638072126ff225/bert-embeddings-gallery.yaml`
 
-var backendAssets *rice.Box
-
-func init() {
-	var err error
-	backendAssets, err = rice.FindBox("backend-assets")
-	if err != nil {
-		panic(err)
-	}
-}
+//go:embed backend-assets/*
+var backendAssets embed.FS
 
 var _ = Describe("API test", func() {
 
@@ -636,7 +629,8 @@ var _ = Describe("API test", func() {
 			application, err := application.New(
 				append(commonOpts,
 					config.WithContext(c),
-					config.WithGeneratedContentDir(tmpdir),
+					config.WithAudioDir(tmpdir),
+					config.WithImageDir(tmpdir),
 					config.WithGalleries(galleries),
 					config.WithModelPath(modelDir),
 					config.WithBackendAssets(backendAssets),

core/http/endpoints/localai/video.go

@@ -1,205 +0,0 @@
-package localai
-
-import (
-	"bufio"
-	"encoding/base64"
-	"encoding/json"
-	"fmt"
-	"io"
-	"net/http"
-	"os"
-	"path/filepath"
-	"strings"
-	"time"
-
-	"github.com/google/uuid"
-	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/core/http/middleware"
-	"github.com/mudler/LocalAI/core/schema"
-
-	"github.com/mudler/LocalAI/core/backend"
-
-	"github.com/gofiber/fiber/v2"
-	model "github.com/mudler/LocalAI/pkg/model"
-	"github.com/rs/zerolog/log"
-)
-
-func downloadFile(url string) (string, error) {
-	// Get the data
-	resp, err := http.Get(url)
-	if err != nil {
-		return "", err
-	}
-	defer resp.Body.Close()
-
-	// Create the file
-	out, err := os.CreateTemp("", "video")
-	if err != nil {
-		return "", err
-	}
-	defer out.Close()
-
-	// Write the body to file
-	_, err = io.Copy(out, resp.Body)
-	return out.Name(), err
-}
-
-//
-
-/*
-*
-
-	curl http://localhost:8080/v1/images/generations \
-	  -H "Content-Type: application/json" \
-	  -d '{
-	    "prompt": "A cute baby sea otter",
-	    "n": 1,
-	    "size": "512x512"
-	  }'
-
-*
-*/
-// VideoEndpoint
-// @Summary Creates a video given a prompt.
-// @Param request body schema.OpenAIRequest true "query params"
-// @Success 200 {object} schema.OpenAIResponse "Response"
-// @Router /video [post]
-func VideoEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
-	return func(c *fiber.Ctx) error {
-		input, ok := c.Locals(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST).(*schema.VideoRequest)
-		if !ok || input.Model == "" {
-			log.Error().Msg("Video Endpoint - Invalid Input")
-			return fiber.ErrBadRequest
-		}
-
-		config, ok := c.Locals(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG).(*config.BackendConfig)
-		if !ok || config == nil {
-			log.Error().Msg("Video Endpoint - Invalid Config")
-			return fiber.ErrBadRequest
-		}
-
-		src := ""
-		if input.StartImage != "" {
-
-			var fileData []byte
-			var err error
-			// check if input.File is an URL, if so download it and save it
-			// to a temporary file
-			if strings.HasPrefix(input.StartImage, "http://") || strings.HasPrefix(input.StartImage, "https://") {
-				out, err := downloadFile(input.StartImage)
-				if err != nil {
-					return fmt.Errorf("failed downloading file:%w", err)
-				}
-				defer os.RemoveAll(out)
-
-				fileData, err = os.ReadFile(out)
-				if err != nil {
-					return fmt.Errorf("failed reading file:%w", err)
-				}
-
-			} else {
-				// base 64 decode the file and write it somewhere
-				// that we will cleanup
-				fileData, err = base64.StdEncoding.DecodeString(input.StartImage)
-				if err != nil {
-					return err
-				}
-			}
-
-			// Create a temporary file
-			outputFile, err := os.CreateTemp(appConfig.GeneratedContentDir, "b64")
-			if err != nil {
-				return err
-			}
-			// write the base64 result
-			writer := bufio.NewWriter(outputFile)
-			_, err = writer.Write(fileData)
-			if err != nil {
-				outputFile.Close()
-				return err
-			}
-			outputFile.Close()
-			src = outputFile.Name()
-			defer os.RemoveAll(src)
-		}
-
-		log.Debug().Msgf("Parameter Config: %+v", config)
-
-		switch config.Backend {
-		case "stablediffusion":
-			config.Backend = model.StableDiffusionGGMLBackend
-		case "":
-			config.Backend = model.StableDiffusionGGMLBackend
-		}
-
-		width := input.Width
-		height := input.Height
-
-		if width == 0 {
-			width = 512
-		}
-		if height == 0 {
-			height = 512
-		}
-
-		b64JSON := input.ResponseFormat == "b64_json"
-
-		tempDir := ""
-		if !b64JSON {
-			tempDir = filepath.Join(appConfig.GeneratedContentDir, "videos")
-		}
-		// Create a temporary file
-		outputFile, err := os.CreateTemp(tempDir, "b64")
-		if err != nil {
-			return err
-		}
-		outputFile.Close()
-
-		// TODO: use mime type to determine the extension
-		output := outputFile.Name() + ".mp4"
-
-		// Rename the temporary file
-		err = os.Rename(outputFile.Name(), output)
-		if err != nil {
-			return err
-		}
-
-		baseURL := c.BaseURL()
-
-		fn, err := backend.VideoGeneration(height, width, input.Prompt, src, input.EndImage, output, ml, *config, appConfig)
-		if err != nil {
-			return err
-		}
-		if err := fn(); err != nil {
-			return err
-		}
-
-		item := &schema.Item{}
-
-		if b64JSON {
-			defer os.RemoveAll(output)
-			data, err := os.ReadFile(output)
-			if err != nil {
-				return err
-			}
-			item.B64JSON = base64.StdEncoding.EncodeToString(data)
-		} else {
-			base := filepath.Base(output)
-			item.URL = baseURL + "/generated-videos/" + base
-		}
-
-		id := uuid.New().String()
-		created := int(time.Now().Unix())
-		resp := &schema.OpenAIResponse{
-			ID:      id,
-			Created: created,
-			Data:    []schema.Item{*item},
-		}
-
-		jsonResult, _ := json.Marshal(resp)
-		log.Debug().Msgf("Response: %s", jsonResult)
-
-		// Return the prediction in the response body
-		return c.JSON(resp)
-	}
-}

core/http/endpoints/openai/image.go

@@ -72,7 +72,7 @@ func ImageEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appCon
 			log.Error().Msg("Image Endpoint - Invalid Input")
 			return fiber.ErrBadRequest
 		}
-
+		
 		config, ok := c.Locals(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG).(*config.BackendConfig)
 		if !ok || config == nil {
 			log.Error().Msg("Image Endpoint - Invalid Config")
@@ -108,7 +108,7 @@ func ImageEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appCon
 			}
 
 			// Create a temporary file
-			outputFile, err := os.CreateTemp(appConfig.GeneratedContentDir, "b64")
+			outputFile, err := os.CreateTemp(appConfig.ImageDir, "b64")
 			if err != nil {
 				return err
 			}
@@ -184,7 +184,7 @@ func ImageEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appCon
 
 				tempDir := ""
 				if !b64JSON {
-					tempDir = filepath.Join(appConfig.GeneratedContentDir, "images")
+					tempDir = appConfig.ImageDir
 				}
 				// Create a temporary file
 				outputFile, err := os.CreateTemp(tempDir, "b64")
@@ -192,7 +192,6 @@ func ImageEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appCon
 					return err
 				}
 				outputFile.Close()
-
 				output := outputFile.Name() + ".png"
 				// Rename the temporary file
 				err = os.Rename(outputFile.Name(), output)

core/http/routes/localai.go

@@ -59,11 +59,6 @@ func RegisterLocalAIRoutes(router *fiber.App,
 		router.Get("/metrics", localai.LocalAIMetricsEndpoint())
 	}
 
-	router.Post("/video",
-		requestExtractor.BuildFilteredFirstAvailableDefaultModel(config.BuildUsecaseFilterFn(config.FLAG_VIDEO)),
-		requestExtractor.SetModelAndConfig(func() schema.LocalAIRequest { return new(schema.VideoRequest) }),
-		localai.VideoEndpoint(cl, ml, appConfig))
-
 	// Backend Statistics Module
 	// TODO: Should these use standard middlewares? Refactor later, they are extremely simple.
 	backendMonitorService := services.NewBackendMonitorService(ml, cl, appConfig) // Split out for now

core/schema/localai.go

@@ -24,20 +24,6 @@ type GalleryResponse struct {
 	StatusURL string `json:"status"`
 }
 
-type VideoRequest struct {
-	BasicModelRequest
-	Prompt         string  `json:"prompt" yaml:"prompt"`
-	StartImage     string  `json:"start_image" yaml:"start_image"`
-	EndImage       string  `json:"end_image" yaml:"end_image"`
-	Width          int32   `json:"width" yaml:"width"`
-	Height         int32   `json:"height" yaml:"height"`
-	NumFrames      int32   `json:"num_frames" yaml:"num_frames"`
-	FPS            int32   `json:"fps" yaml:"fps"`
-	Seed           int32   `json:"seed" yaml:"seed"`
-	CFGScale       float32 `json:"cfg_scale" yaml:"cfg_scale"`
-	ResponseFormat string  `json:"response_format" yaml:"response_format"`
-}
-
 // @Description TTS request body
 type TTSRequest struct {
 	BasicModelRequest

docs/content/docs/advanced/advanced-usage.md

@@ -481,7 +481,8 @@ In the help text below, BASEPATH is the location that local-ai is being executed
 |-----------|---------|-------------|----------------------|
 | --models-path | BASEPATH/models | Path containing models used for inferencing  | $LOCALAI_MODELS_PATH |
 | --backend-assets-path |/tmp/localai/backend_data | Path used to extract libraries that are required by some of the backends in runtime | $LOCALAI_BACKEND_ASSETS_PATH |
-| --generated-content-path | /tmp/generated/content | Location for assets generated by backends (e.g. stablediffusion) | $LOCALAI_GENERATED_CONTENT_PATH |
+| --image-path | /tmp/generated/images | Location for images generated by backends (e.g. stablediffusion) | $LOCALAI_IMAGE_PATH |
+| --audio-path | /tmp/generated/audio | Location for audio generated by backends (e.g. piper) | $LOCALAI_AUDIO_PATH |
 | --upload-path | /tmp/localai/upload | Path to store uploads from files api | $LOCALAI_UPLOAD_PATH |
 | --config-path | /tmp/localai/config | | $LOCALAI_CONFIG_PATH |
 | --localai-config-dir | BASEPATH/configuration | Directory for dynamic loading of certain configuration files (currently api_keys.json and external_backends.json) | $LOCALAI_CONFIG_DIR |
@@ -514,7 +515,6 @@ In the help text below, BASEPATH is the location that local-ai is being executed
 | --upload-limit | 15 | Default upload-limit in MB | $LOCALAI_UPLOAD_LIMIT |
 | --api-keys | API-KEYS,... | List of API Keys to enable API authentication. When this is set, all the requests must be authenticated with one of these API keys | $LOCALAI_API_KEY |
 | --disable-welcome |  | Disable welcome pages | $LOCALAI_DISABLE_WELCOME |
-| --disable-webui | false | Disables the web user interface. When set to true, the server will only expose API endpoints without serving the web interface | $LOCALAI_DISABLE_WEBUI |
 | --machine-tag |  | If not empty - put that string to Machine-Tag header in each response. Useful to track response from different machines using multiple P2P federated nodes | $LOCALAI_MACHINE_TAG |
 
 #### Backend Flags

docs/content/docs/advanced/installer.md

@@ -34,12 +34,4 @@ List of the Environment Variables:
 | **FEDERATED** | Set to "true" to share the instance with the federation (p2p token is required see [documentation]({{%relref "docs/features/distributed_inferencing" %}}))  |
 | **FEDERATED_SERVER** | Set to "true" to run the instance as a federation server which forwards requests to the federation (p2p token is required see [documentation]({{%relref "docs/features/distributed_inferencing" %}}))  |
 
-## Uninstallation
-
-To uninstall, run:
-
-```
-curl https://localai.io/install.sh | sh -s -- --uninstall
-```
-
 We are looking into improving the installer, and as this is a first iteration any feedback is welcome! Open up an [issue](https://github.com/mudler/LocalAI/issues/new/choose) if something doesn't work for you!

docs/content/docs/features/GPU-acceleration.md

@@ -57,14 +57,12 @@ diffusers:
 
 Requirement: nvidia-container-toolkit (installation instructions [1](https://www.server-world.info/en/note?os=Ubuntu_22.04&p=nvidia&f=2) [2](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html))
 
-If using a system with SELinux, ensure you have the policies installed, such as those [provided by nvidia](https://github.com/NVIDIA/dgx-selinux/)
-
-To check what CUDA version do you need, you can either run `nvidia-smi` or `nvcc --version`.
+To check what CUDA version do you need, you can either run `nvidia-smi` or `nvcc --version`. 
 
 Alternatively, you can also check nvidia-smi with docker:
 
 ```
-docker run --runtime=nvidia --rm nvidia/cuda:12.8.0-base-ubuntu24.04 nvidia-smi
+docker run --runtime=nvidia --rm nvidia/cuda nvidia-smi
 ```
 
 To use CUDA, use the images with the `cublas` tag, for example.
@@ -114,7 +112,7 @@ llama_init_from_file: kv self size  =  512.00 MB
 
 ## ROCM(AMD) acceleration
 
-There are a limited number of tested configurations for ROCm systems however most newer deditated GPU consumer grade devices seem to be supported under the current ROCm6 implementation.
+There are a limited number of tested configurations for ROCm systems however most newer deditated GPU consumer grade devices seem to be supported under the current ROCm6 implementation. 
 
 Due to the nature of ROCm it is best to run all implementations in containers as this limits the number of packages required for installation on host system, compatability and package versions for dependencies across all variations of OS must be tested independently if disired, please refer to the [build]({{%relref "docs/getting-started/build#Acceleration" %}}) documentation.
 
@@ -139,7 +137,7 @@ LocalAI hipblas images are built against the following targets: gfx900,gfx906,gf
 
 If your device is not one of these you must specify the corresponding `GPU_TARGETS` and specify `REBUILD=true`. Otherwise you don't need to specify these in the commands below.
 
-### Verified
+### Verified 
 
 The devices in the following list have been tested with `hipblas` images running `ROCm 6.0.0`
 
@@ -167,7 +165,7 @@ The devices in the following list have been tested with `hipblas` images running
 1. Check your GPU LLVM target is compatible with the version of ROCm. This can be found in the [LLVM Docs](https://llvm.org/docs/AMDGPUUsage.html).
 2. Check which ROCm version is compatible with your LLVM target and your chosen OS (pay special attention to supported kernel versions). See the following for compatability for ([ROCm 6.0.0](https://rocm.docs.amd.com/projects/install-on-linux/en/docs-6.0.0/reference/system-requirements.html)) or ([ROCm 6.0.2](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/reference/system-requirements.html))
 3. Install you chosen version of the `dkms` and `rocm` (it is recommended that the native package manager be used for this process for any OS as version changes are executed more easily via this method if updates are required). Take care to restart after installing `amdgpu-dkms` and before installing `rocm`, for details regarding this see the installation documentation for your chosen OS ([6.0.2](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/how-to/native-install/index.html) or [6.0.0](https://rocm.docs.amd.com/projects/install-on-linux/en/docs-6.0.0/how-to/native-install/index.html))
-4. Deploy. Yes it's that easy.
+4. Deploy. Yes it's that easy. 
 
 #### Setup Example (Docker/containerd)
 
@@ -249,7 +247,7 @@ This configuration has been tested on a 'custom' cluster managed by SUSE Rancher
 
 - When installing the ROCM kernel driver on your system ensure that you are installing an equal or newer version that that which is currently implemented in LocalAI (6.0.0 at time of writing).
 - AMD documentation indicates that this will ensure functionality however your milage may vary depending on the GPU and distro you are using.
-- If you encounter an `Error 413` on attempting to upload an audio file or image for whisper or llava/bakllava on a k8s deployment, note that the ingress for your deployment may require the annontation `nginx.ingress.kubernetes.io/proxy-body-size: "25m"` to allow larger uploads. This may be included in future versions of the helm chart.
+- If you encounter an `Error 413` on attempting to upload an audio file or image for whisper or llava/bakllava on a k8s deployment, note that the ingress for your deployment may require the annontation `nginx.ingress.kubernetes.io/proxy-body-size: "25m"` to allow larger uploads. This may be included in future versions of the helm chart. 
 
 ## Intel acceleration (sycl)
 
@@ -280,36 +278,3 @@ docker run --rm -ti --device /dev/dri -p 8080:8080 -e DEBUG=true -e MODELS_PATH=
 ```
 
 Note also that sycl does have a known issue to hang with `mmap: true`. You have to disable it in the model configuration if explicitly enabled.
-
-## Vulkan acceleration
-
-### Requirements
-
-If using nvidia, follow the steps in the [CUDA](#cudanvidia-acceleration) section to configure your docker runtime to allow access to the GPU.
-
-### Container images
-
-To use Vulkan, use the images with the `vulkan` tag, for example `{{< version >}}-vulkan-ffmpeg-core`.
-
-#### Example
-
-To run LocalAI with Docker and Vulkan, you can use the following command as an example:
-
-```bash
-docker run -p 8080:8080 -e DEBUG=true -v $PWD/models:/build/models localai/localai:latest-vulkan-ffmpeg-core
-```
-
-### Notes
-
-In addition to the commands to run LocalAI normally, you need to specify additonal flags to pass the GPU hardware to the container.
-
-These flags are the same as the sections above, depending on the hardware, for [nvidia](#cudanvidia-acceleration), [AMD](#rocmamd-acceleration) or [Intel](#intel-acceleration-sycl).
-
-If you have mixed hardware, you can pass flags for multiple GPUs, for example:
-
-```bash
-docker run -p 8080:8080 -e DEBUG=true -v $PWD/models:/build/models \
---gpus=all \ # nvidia passthrough
---device /dev/dri --device /dev/kfd \ # AMD/Intel passthrough
-localai/localai:latest-vulkan-ffmpeg-core
-```

docs/content/docs/getting-started/quickstart.md

@@ -18,15 +18,12 @@ If you are exposing LocalAI remotely, make sure you protect the API endpoints ad
 
 ## Quickstart
 
-### Using the Bash Installer
 
+### Using the Bash Installer
 ```bash
-# Basic installation
 curl https://localai.io/install.sh | sh
 ```
 
-See [Installer]({{% relref "docs/advanced/installer" %}}) for all the supported options
-
 ### Run with docker:
 ```bash
 # CPU only image:
@@ -101,57 +98,6 @@ The AIO images come pre-configured with the following features:
 
 For instructions on using AIO images, see [Using container images]({{% relref "docs/getting-started/container-images#all-in-one-images" %}}).
 
-## Using LocalAI and the full stack with LocalAGI
-
-LocalAI is part of the Local family stack, along with LocalAGI and LocalRecall.
-
-[LocalAGI](https://github.com/mudler/LocalAGI) is a powerful, self-hostable AI Agent platform designed for maximum privacy and flexibility which encompassess and uses all the softwre stack. It provides a complete drop-in replacement for OpenAI's Responses APIs with advanced agentic capabilities, working entirely locally on consumer-grade hardware (CPU and GPU).
-
-### Quick Start
-
-```bash
-# Clone the repository
-git clone https://github.com/mudler/LocalAGI
-cd LocalAGI
-
-# CPU setup (default)
-docker compose up
-
-# NVIDIA GPU setup
-docker compose -f docker-compose.nvidia.yaml up
-
-# Intel GPU setup (for Intel Arc and integrated GPUs)
-docker compose -f docker-compose.intel.yaml up
-
-# Start with a specific model (see available models in models.localai.io, or localai.io to use any model in huggingface)
-MODEL_NAME=gemma-3-12b-it docker compose up
-
-# NVIDIA GPU setup with custom multimodal and image models
-MODEL_NAME=gemma-3-12b-it \
-MULTIMODAL_MODEL=minicpm-v-2_6 \
-IMAGE_MODEL=flux.1-dev-ggml \
-docker compose -f docker-compose.nvidia.yaml up
-```
-
-### Key Features
-
-- **Privacy-Focused**: All processing happens locally, ensuring your data never leaves your machine
-- **Flexible Deployment**: Supports CPU, NVIDIA GPU, and Intel GPU configurations
-- **Multiple Model Support**: Compatible with various models from Hugging Face and other sources
-- **Web Interface**: User-friendly chat interface for interacting with AI agents
-- **Advanced Capabilities**: Supports multimodal models, image generation, and more
-- **Docker Integration**: Easy deployment using Docker Compose
-
-### Environment Variables
-
-You can customize your LocalAGI setup using the following environment variables:
-
-- `MODEL_NAME`: Specify the model to use (e.g., `gemma-3-12b-it`)
-- `MULTIMODAL_MODEL`: Set a custom multimodal model
-- `IMAGE_MODEL`: Configure an image generation model
-
-For more advanced configuration and API documentation, visit the [LocalAGI GitHub repository](https://github.com/mudler/LocalAGI).
-
 ## What's Next?
 
 There is much more to explore with LocalAI! You can run any model from Hugging Face, perform video generation, and also voice cloning. For a comprehensive overview, check out the [features]({{% relref "docs/features" %}}) section.

docs/data/version.json

@@ -1,3 +1,3 @@
 {
-  "version": "v2.29.0"
+  "version": "v2.28.0"
 }

docs/static/install.sh

@@ -1,74 +1,31 @@
 #!/bin/sh
-# LocalAI Installer Script
-# This script installs LocalAI on Linux and macOS systems.
-# It automatically detects the system architecture and installs the appropriate version.
+# This script installs LocalAI on Linux.
+# It detects the current operating system architecture and installs the appropriate version of LocalAI.
 
 # Usage:
-#   Basic installation:
-#     curl https://localai.io/install.sh | sh
-#
-#   With environment variables:
-#     DOCKER_INSTALL=true USE_AIO=true API_KEY=your-key PORT=8080 THREADS=4 curl https://localai.io/install.sh | sh
-#
-#   To uninstall:
-#     curl https://localai.io/install.sh | sh -s -- --uninstall
-#
-# Environment Variables:
-#   DOCKER_INSTALL - Set to "true" to install Docker images (default: auto-detected)
-#   USE_AIO       - Set to "true" to use the all-in-one LocalAI image (default: false)
-#   API_KEY       - API key for securing LocalAI access (default: none)
-#   PORT          - Port to run LocalAI on (default: 8080)
-#   THREADS       - Number of CPU threads to use (default: auto-detected)
-#   MODELS_PATH   - Path to store models (default: /usr/share/local-ai/models)
-#   CORE_IMAGES   - Set to "true" to download core LocalAI images (default: false)
-#   P2P_TOKEN     - Token for P2P federation/worker mode (default: none)
-#   WORKER        - Set to "true" to run as a worker node (default: false)
-#   FEDERATED     - Set to "true" to enable federation mode (default: false)
-#   FEDERATED_SERVER - Set to "true" to run as a federation server (default: false)
+#   curl ... | ENV_VAR=... sh -
+#       or
+#   ENV_VAR=... ./install.sh
+#   To uninstall: ./install.sh --uninstall
 
 set -e
 set -o noglob
 #set -x
 
 # --- helper functions for logs ---
-# ANSI escape codes
-LIGHT_BLUE='\033[38;5;117m'
-ORANGE='\033[38;5;214m'
-RED='\033[38;5;196m'
-BOLD='\033[1m'
-RESET='\033[0m'
-
 info()
 {
-    echo -e "${BOLD}${LIGHT_BLUE}" '[INFO] ' "$@" "${RESET}"
+    echo ' ' "$@"
 }
 
 warn()
 {
-    echo -e "${BOLD}${ORANGE}" '[WARN] ' "$@" "${RESET}" >&2
+    echo '[WARN] ' "$@" >&2
 }
 
 fatal()
 {
-    echo -e "${BOLD}${RED}" '[ERROR] ' "$@" "${RESET}" >&2
-    exit 1
-}
-
-# --- custom choice functions ---
-# like the logging functions, but with the -n flag to prevent the new line and keep the cursor in line for choices inputs like y/n
-choice_info()
-{
-    echo -e -n "${BOLD}${LIGHT_BLUE}" '[INFO] ' "$@" "${RESET}"
-}
-
-choice_warn()
-{
-    echo -e -n "${BOLD}${ORANGE}" '[WARN] ' "$@" "${RESET}" >&2
-}
-
-choice_fatal()
-{
-    echo -e -n "${BOLD}${RED}" '[ERROR] ' "$@" "${RESET}" >&2
+    echo '[ERROR] ' "$@" >&2
     exit 1
 }
 
@@ -186,7 +143,7 @@ else
 fi
 THREADS=${THREADS:-$procs}
 LATEST_VERSION=$(curl -s "https://api.github.com/repos/mudler/LocalAI/releases/latest" | grep '"tag_name":' | sed -E 's/.*"([^"]+)".*/\1/')
-LOCALAI_VERSION="${LOCALAI_VERSION:-$LATEST_VERSION}" #changed due to VERSION beign already defined in Fedora 42 Cloud Edition
+VERSION="${VERSION:-$LATEST_VERSION}"
 MODELS_PATH=${MODELS_PATH:-/usr/share/local-ai/models}
 
 
@@ -253,7 +210,7 @@ WorkingDirectory=/usr/share/local-ai
 [Install]
 WantedBy=default.target
 EOF
-
+    
     $SUDO touch /etc/localai.env
     $SUDO echo "ADDRESS=0.0.0.0:$PORT" | $SUDO tee /etc/localai.env >/dev/null
     $SUDO echo "API_KEY=$API_KEY" | $SUDO tee -a /etc/localai.env >/dev/null
@@ -286,74 +243,23 @@ EOF
 
 # ref: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installing-with-yum-or-dnf
 install_container_toolkit_yum() {
-    info 'Installing NVIDIA container toolkit repository...'
+    info 'Installing NVIDIA repository...'
 
     curl -s -L https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo | \
     $SUDO  tee /etc/yum.repos.d/nvidia-container-toolkit.repo
 
     if [ "$PACKAGE_MANAGER" = "dnf" ]; then
-        DNF_VERSION=$($PACKAGE_MANAGER --version | grep -oE '[0-9]+\.[0-9]+\.[0-9]+' | head -n1 | cut -d. -f1)
-        if [ "$DNF_VERSION" -ge 5 ]; then
-            # DNF5: Use 'setopt' to enable the repository
-            $SUDO $PACKAGE_MANAGER config-manager setopt nvidia-container-toolkit-experimental.enabled=1
-        else
-            # DNF4: Use '--set-enabled' to enable the repository
-            $SUDO $PACKAGE_MANAGER config-manager --enable nvidia-container-toolkit-experimental
-        fi
-    else
+        $SUDO $PACKAGE_MANAGER config-manager --enable nvidia-container-toolkit-experimental
+    else 
         $SUDO $PACKAGE_MANAGER -y install yum-utils
         $SUDO $PACKAGE_MANAGER-config-manager --enable nvidia-container-toolkit-experimental
     fi
     $SUDO $PACKAGE_MANAGER install -y nvidia-container-toolkit
 }
 
-# Fedora, Rhel and other distro ships tunable SELinux booleans in the container-selinux policy to control device access.
-# In particular, enabling container_use_devices allows containers to use arbitrary host device labels (including GPU devices)
-# ref: https://github.com/containers/ramalama/blob/main/docs/ramalama-cuda.7.md#expected-output
-enable_selinux_container_booleans() {
-
-    # Check SELinux mode
-    SELINUX_MODE=$(getenforce)
-
-    if [ "$SELINUX_MODE" == "Enforcing" ]; then
-        # Check the status of container_use_devices
-        CONTAINER_USE_DEVICES=$(getsebool container_use_devices | awk '{print $3}')
-
-       if [ "$CONTAINER_USE_DEVICES" == "off" ]; then
-
-          #We want to give the user the choice to enable the SE booleans since it is a security config
-          warn "+-----------------------------------------------------------------------------------------------------------+"
-          warn "| WARNING:                                                                                                  |"
-          warn "| Your distribution ships tunable SELinux booleans in the container-selinux policy to control device access.|"
-          warn "| In particular, enabling \"container_use_devices\" allows containers to use arbitrary host device labels   |"
-          warn "| (including GPU devices).                                                                                  |"
-          warn "| This script can try to enable them enabling the \"container_use_devices\" flag.                           |"
-          warn "|                                                                                                           |"
-          warn "| Otherwise you can exit the install script and enable them yourself.                                       |"
-          warn "+-----------------------------------------------------------------------------------------------------------+"
-
-          while true; do
-              choice_warn "I understand that this script is going to change my SELinux configs, which is a security risk: (yes/exit) ";
-              read  Answer
-
-              if [ "$Answer" = "yes" ]; then
-                warn "Enabling \"container_use_devices\" persistently..."
-                $SUDO setsebool -P container_use_devices 1
-
-                break
-              elif [ "$Answer" = "exit" ]; then
-                  aborted
-              else
-                  warn "Invalid choice. Please enter 'yes' or 'exit'."
-              fi
-            done
-       fi
-    fi
-}
-
 # ref: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installing-with-apt
 install_container_toolkit_apt() {
-    info 'Installing NVIDIA container toolkit repository...'
+    info 'Installing NVIDIA repository...'
 
     curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | $SUDO gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
   && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
@@ -365,7 +271,7 @@ install_container_toolkit_apt() {
 
 # ref: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installing-with-zypper
 install_container_toolkit_zypper() {
-    info 'Installing NVIDIA zypper repository...'
+    info 'Installing NVIDIA repository...'
     $SUDO zypper ar https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo
     $SUDO zypper modifyrepo --enable nvidia-container-toolkit-experimental
     $SUDO zypper --gpg-auto-import-keys install -y nvidia-container-toolkit
@@ -394,29 +300,6 @@ install_container_toolkit() {
             opensuse*|suse*) install_container_toolkit_zypper ;;
             *) echo "Could not install nvidia container toolkit - unknown OS" ;;
     esac
-
-    # after installing the toolkit we need to add it to the docker runtimes, otherwise even with --gpu all
-    # the container would still run with runc and would not have access to nvidia-smi
-    # ref: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#configuring-docker
-    info "Adding NVIDIA Container Runtime to Docker runtimes..."
-    $SUDO nvidia-ctk runtime configure --runtime=docker
-
-    info "Restarting Docker Daemon"
-    $SUDO systemctl restart docker
-
-    # The NVML error arises because SELinux blocked the container’s attempts to open the GPU devices or related libraries.
-    # Without relaxing SELinux for the container, GPU commands like nvidia-smi report “Insufficient Permissions”
-    # This has been noted in NVIDIA’s documentation:
-    # ref: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/1.13.5/install-guide.html#id2
-    # ref: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/troubleshooting.html#nvml-insufficient-permissions-and-selinux
-    case $OS_NAME in
-            fedora|rhel|centos|rocky)
-                enable_selinux_container_booleans
-                ;;
-            opensuse-tumbleweed)
-                enable_selinux_container_booleans
-                ;;
-    esac
 }
 
 # ref: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#rhel-7-centos-7
@@ -424,21 +307,14 @@ install_container_toolkit() {
 # ref: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#rhel-9-rocky-9
 # ref: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#fedora
 install_cuda_driver_yum() {
-    info 'Installing NVIDIA CUDA repository...'
+    info 'Installing NVIDIA repository...'
     case $PACKAGE_MANAGER in
         yum)
             $SUDO $PACKAGE_MANAGER -y install yum-utils
             $SUDO $PACKAGE_MANAGER-config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/$1$2/$(uname -m)/cuda-$1$2.repo
             ;;
         dnf)
-            DNF_VERSION=$($PACKAGE_MANAGER --version | grep -oE '[0-9]+\.[0-9]+\.[0-9]+' | head -n1 | cut -d. -f1)
-            if [ "$DNF_VERSION" -ge 5 ]; then
-                # DNF5: Use 'addrepo' to add the repository
-                $SUDO $PACKAGE_MANAGER config-manager addrepo --id=nvidia-cuda --set=name="nvidia-cuda" --set=baseurl="https://developer.download.nvidia.com/compute/cuda/repos/$1$2/$(uname -m)/cuda-$1$2.repo"
-            else
-                # DNF4: Use '--add-repo' to add the repository
-                $SUDO $PACKAGE_MANAGER config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/$1$2/$(uname -m)/cuda-$1$2.repo
-            fi
+            $SUDO $PACKAGE_MANAGER config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/$1$2/$(uname -m)/cuda-$1$2.repo
             ;;
     esac
 
@@ -459,68 +335,10 @@ install_cuda_driver_yum() {
     $SUDO $PACKAGE_MANAGER -y install cuda-drivers
 }
 
-install_fedora_nvidia_kernel_drivers(){
-
-  #We want to give the user the choice to install the akmod kernel drivers or not, since it could break some setups
-  warn "+------------------------------------------------------------------------------------------------+"
-  warn "| WARNING:                                                                                       |"
-  warn "| Looks like the NVIDIA Kernel modules are not installed.                                        |"
-  warn "|                                                                                                |"
-  warn "| This script can try to install them using akmod-nvidia.                                        |"
-  warn "| - The script need the rpmfusion free and nonfree repos and will install them if not available. |"
-  warn "| - The akmod installation can sometimes inhibit the reboot command.                             |"
-  warn "|                                                                                                |"
-  warn "| Otherwise you can exit the install script and install them yourself.                           |"
-  warn "| NOTE: you will need to reboot after the installation.                                          |"
-  warn "+------------------------------------------------------------------------------------------------+"
-
-  while true; do
-    choice_warn "Do you wish for the script to try and install them? (akmod/exit) ";
-    read  Answer
-
-    if [ "$Answer" = "akmod" ]; then
-
-      DNF_VERSION=$($PACKAGE_MANAGER --version | grep -oE '[0-9]+\.[0-9]+\.[0-9]+' | head -n1 | cut -d. -f1)
-
-      OS_NAME=$ID
-      OS_VERSION=$VERSION_ID
-      FREE_URL="https://mirrors.rpmfusion.org/free/fedora/rpmfusion-free-release-${OS_VERSION}.noarch.rpm"
-      NONFREE_URL="https://mirrors.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-${OS_VERSION}.noarch.rpm"
-
-      curl -LO "$FREE_URL"
-      curl -LO "$NONFREE_URL"
-
-      if [ "$DNF_VERSION" -ge 5 ]; then
-          # DNF5:
-          $SUDO $PACKAGE_MANAGER install -y "rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm" "rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm"
-          $SUDO $PACKAGE_MANAGER install -y akmod-nvidia
-      else
-          # DNF4:
-          $SUDO $PACKAGE_MANAGER install -y "rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm" "rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm"
-          $SUDO $PACKAGE_MANAGER install -y akmod-nvidia
-      fi
-
-      $SUDO rm "rpmfusion-free-release-$(rpm -E %fedora).noarch.rpm"
-      $SUDO rm "rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm"
-
-      install_cuda_driver_yum $OS_NAME '41'
-
-      info "Nvidia driver installation complete, please reboot now and run the Install script again to complete the setup."
-      exit
-
-    elif [ "$Answer" = "exit" ]; then
-
-        aborted
-    else
-        warn "Invalid choice. Please enter 'akmod' or 'exit'."
-    fi
-  done
-}
-
 # ref: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#ubuntu
 # ref: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#debian
 install_cuda_driver_apt() {
-    info 'Installing NVIDIA CUDA repository...'
+    info 'Installing NVIDIA repository...'
     curl -fsSL -o $TEMP_DIR/cuda-keyring.deb https://developer.download.nvidia.com/compute/cuda/repos/$1$2/$(uname -m)/cuda-keyring_1.1-1_all.deb
 
     case $1 in
@@ -559,7 +377,7 @@ install_cuda() {
         case $OS_NAME in
             centos|rhel) install_cuda_driver_yum 'rhel' $(echo $OS_VERSION | cut -d '.' -f 1) ;;
             rocky) install_cuda_driver_yum 'rhel' $(echo $OS_VERSION | cut -c1) ;;
-            fedora) [ $OS_VERSION -lt '41' ] && install_cuda_driver_yum $OS_NAME $OS_VERSION || install_cuda_driver_yum $OS_NAME '41';;
+            fedora) [ $OS_VERSION -lt '37' ] && install_cuda_driver_yum $OS_NAME $OS_VERSION || install_cuda_driver_yum $OS_NAME '37';;
             amzn) install_cuda_driver_yum 'fedora' '37' ;;
             debian) install_cuda_driver_apt $OS_NAME $OS_VERSION ;;
             ubuntu) install_cuda_driver_apt $OS_NAME $(echo $OS_VERSION | sed 's/\.//') ;;
@@ -635,7 +453,7 @@ install_docker() {
         $SUDO systemctl start docker
     fi
 
-    info "Creating LocalAI Docker volume..."
+    info "Starting LocalAI Docker container..."
     # Create volume if doesn't exist already
     if ! $SUDO docker volume inspect local-ai-data > /dev/null 2>&1; then
         $SUDO docker volume create local-ai-data
@@ -649,7 +467,7 @@ install_docker() {
         # if $SUDO docker ps --format '{{.Names}}' | grep -q local-ai; then
         #     info "LocalAI Docker container is already running."
         #     exit 0
-        # fi
+        # fi 
 
         # info "Starting LocalAI Docker container..."
         # $SUDO docker start local-ai
@@ -666,28 +484,22 @@ install_docker() {
 
     IMAGE_TAG=
     if [ "$HAS_CUDA" ]; then
-        IMAGE_TAG=${LOCALAI_VERSION}-cublas-cuda12-ffmpeg
+        IMAGE_TAG=${VERSION}-cublas-cuda12-ffmpeg
         # CORE
         if [ "$CORE_IMAGES" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-cublas-cuda12-ffmpeg-core
+            IMAGE_TAG=${VERSION}-cublas-cuda12-ffmpeg-core
         fi
         # AIO
         if [ "$USE_AIO" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-aio-gpu-nvidia-cuda-12
+            IMAGE_TAG=${VERSION}-aio-gpu-nvidia-cuda-12
         fi
 
-        info "Checking Nvidia Kernel Drivers presence..."
         if ! available nvidia-smi; then
-          OS_NAME=$ID
-          OS_VERSION=$VERSION_ID
-
-            case $OS_NAME in
-                debian|ubuntu) $SUDO apt-get -y install nvidia-cuda-toolkit;;
-                fedora) install_fedora_nvidia_kernel_drivers;;
-            esac
+            info "Installing nvidia-cuda-toolkit..."
+            # TODO:
+            $SUDO apt-get -y install nvidia-cuda-toolkit
         fi
 
-        info "Starting LocalAI Docker container..."
         $SUDO docker run -v local-ai-data:/build/models \
             --gpus all \
             --restart=always \
@@ -696,17 +508,16 @@ install_docker() {
             $envs \
             -d -p $PORT:8080 --name local-ai localai/localai:$IMAGE_TAG $STARTCOMMAND
     elif [ "$HAS_AMD" ]; then
-        IMAGE_TAG=${LOCALAI_VERSION}-hipblas-ffmpeg
+        IMAGE_TAG=${VERSION}-hipblas-ffmpeg
         # CORE
         if [ "$CORE_IMAGES" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-hipblas-ffmpeg-core
+            IMAGE_TAG=${VERSION}-hipblas-ffmpeg-core
         fi
         # AIO
         if [ "$USE_AIO" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-aio-gpu-hipblas
+            IMAGE_TAG=${VERSION}-aio-gpu-hipblas
         fi
 
-        info "Starting LocalAI Docker container..."
         $SUDO docker run -v local-ai-data:/build/models \
             --device /dev/dri \
             --device /dev/kfd \
@@ -716,17 +527,16 @@ install_docker() {
             $envs \
             -d -p $PORT:8080 --name local-ai localai/localai:$IMAGE_TAG $STARTCOMMAND
     elif [ "$HAS_INTEL" ]; then
-        IMAGE_TAG=${LOCALAI_VERSION}-sycl-f32-ffmpeg
+        IMAGE_TAG=${VERSION}-sycl-f32-ffmpeg
         # CORE
         if [ "$CORE_IMAGES" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-sycl-f32-ffmpeg-core
+            IMAGE_TAG=${VERSION}-sycl-f32-ffmpeg-core
         fi
         # AIO
         if [ "$USE_AIO" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-aio-gpu-intel-f32
+            IMAGE_TAG=${VERSION}-aio-gpu-intel-f32
         fi
 
-        info "Starting LocalAI Docker container..."
         $SUDO docker run -v local-ai-data:/build/models \
             --device /dev/dri \
             --restart=always \
@@ -735,17 +545,15 @@ install_docker() {
             $envs \
             -d -p $PORT:8080 --name local-ai localai/localai:$IMAGE_TAG $STARTCOMMAND
     else
-        IMAGE_TAG=${LOCALAI_VERSION}-ffmpeg
+        IMAGE_TAG=${VERSION}-ffmpeg
         # CORE
         if [ "$CORE_IMAGES" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-ffmpeg-core
+            IMAGE_TAG=${VERSION}-ffmpeg-core
         fi
         # AIO
         if [ "$USE_AIO" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-aio-cpu
-        fi
-
-        info "Starting LocalAI Docker container..."
+            IMAGE_TAG=${VERSION}-aio-cpu
+        fi        
         $SUDO docker run -v local-ai-data:/models \
                 --restart=always \
                 -e MODELS_PATH=/models \
@@ -762,8 +570,8 @@ install_docker() {
 install_binary_darwin() {
     [ "$(uname -s)" = "Darwin" ] || fatal 'This script is intended to run on macOS only.'
 
-    info "Downloading LocalAI ${LOCALAI_VERSION}..."
-    curl --fail --show-error --location --progress-bar -o $TEMP_DIR/local-ai "https://github.com/mudler/LocalAI/releases/download/${LOCALAI_VERSION}/local-ai-Darwin-${ARCH}"
+    info "Downloading LocalAI ${VERSION}..."
+    curl --fail --show-error --location --progress-bar -o $TEMP_DIR/local-ai "https://github.com/mudler/LocalAI/releases/download/${VERSION}/local-ai-Darwin-${ARCH}"
 
     info "Installing to /usr/local/bin/local-ai"
     install -o0 -g0 -m755 $TEMP_DIR/local-ai /usr/local/bin/local-ai
@@ -794,8 +602,8 @@ install_binary() {
         exit 1
     fi
 
-    info "Downloading LocalAI ${LOCALAI_VERSION}..."
-    curl --fail --location --progress-bar -o $TEMP_DIR/local-ai "https://github.com/mudler/LocalAI/releases/download/${LOCALAI_VERSION}/local-ai-Linux-${ARCH}"
+    info "Downloading LocalAI ${VERSION}..."
+    curl --fail --location --progress-bar -o $TEMP_DIR/local-ai "https://github.com/mudler/LocalAI/releases/download/${VERSION}/local-ai-Linux-${ARCH}"
 
     for BINDIR in /usr/local/bin /usr/bin /bin; do
         echo $PATH | grep -q $BINDIR && break || continue
@@ -849,7 +657,7 @@ detect_start_command() {
     if [ "$WORKER" = true ]; then
         if [ -n "$P2P_TOKEN" ]; then
             STARTCOMMAND="worker p2p-llama-cpp-rpc"
-        else
+        else 
             STARTCOMMAND="worker llama-cpp-rpc"
         fi
     elif [ "$FEDERATED" = true ]; then

gallery/qwen3.yaml

@@ -1,39 +0,0 @@
----
-name: "qwen3"
-
-config_file: |
-  mmap: true
-  template:
-    chat_message: |
-      <|im_start|>{{ .RoleName }}
-      {{ if .FunctionCall -}}
-      {{ else if eq .RoleName "tool" -}}
-      {{ end -}}
-      {{ if .Content -}}
-      {{.Content }}
-      {{ end -}}
-      {{ if .FunctionCall -}}
-      {{toJson .FunctionCall}}
-      {{ end -}}<|im_end|>
-    function: |
-      <|im_start|>system
-      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. Here are the available tools:
-      {{range .Functions}}
-      {'type': 'function', 'function': {'name': '{{.Name}}', 'description': '{{.Description}}', 'parameters': {{toJson .Parameters}} }}
-      {{end}}
-      For each function call return a json object with function name and arguments
-      <|im_end|>
-      {{.Input -}}
-      <|im_start|>assistant
-    chat: |
-      {{.Input -}}
-      <|im_start|>assistant
-    completion: |
-      {{.Input}}
-  context_size: 8192
-  f16: true
-  stopwords:
-  - '<|im_end|>'
-  - '<dummy32000>'
-  - '</s>'
-  - '<|endoftext|>'

go.mod

@@ -6,7 +6,6 @@ toolchain go1.23.1
 
 require (
 	dario.cat/mergo v1.0.1
-	github.com/GeertJohan/go.rice v1.0.3
 	github.com/Masterminds/sprig/v3 v3.3.0
 	github.com/alecthomas/kong v0.9.0
 	github.com/census-instrumentation/opencensus-proto v0.4.1
@@ -28,7 +27,6 @@ require (
 	github.com/golang/protobuf v1.5.4
 	github.com/google/go-containerregistry v0.19.2
 	github.com/google/uuid v1.6.0
-	github.com/gpustack/gguf-parser-go v0.17.0
 	github.com/grpc-ecosystem/grpc-gateway v1.5.0
 	github.com/hpcloud/tail v1.0.0
 	github.com/ipfs/go-log v1.0.5
@@ -44,6 +42,7 @@ require (
 	github.com/onsi/ginkgo/v2 v2.22.2
 	github.com/onsi/gomega v1.36.2
 	github.com/orcaman/writerseeker v0.0.0-20200621085525-1d3f536ff85e
+	github.com/ory/dockertest/v3 v3.10.0
 	github.com/otiai10/openaigo v1.7.0
 	github.com/phayes/freeport v0.0.0-20220201140144-74d24b5ae9f5
 	github.com/prometheus/client_golang v1.20.5
@@ -55,7 +54,7 @@ require (
 	github.com/streamer45/silero-vad-go v0.2.1
 	github.com/stretchr/testify v1.10.0
 	github.com/swaggo/swag v1.16.3
-	github.com/testcontainers/testcontainers-go v0.35.0
+	github.com/thxcode/gguf-parser-go v0.1.0
 	github.com/tmc/langchaingo v0.1.12
 	github.com/valyala/fasthttp v1.55.0
 	go.opentelemetry.io/otel v1.34.0
@@ -75,26 +74,19 @@ require (
 	cloud.google.com/go/auth v0.4.1 // indirect
 	cloud.google.com/go/auth/oauth2adapt v0.2.2 // indirect
 	cloud.google.com/go/compute/metadata v0.5.0 // indirect
-	github.com/containerd/platforms v0.2.1 // indirect
-	github.com/cpuguy83/dockercfg v0.3.2 // indirect
-	github.com/daaku/go.zipexe v1.0.2 // indirect
-	github.com/distribution/reference v0.6.0 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/envoyproxy/protoc-gen-validate v1.1.0 // indirect
 	github.com/fasthttp/websocket v1.5.3 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
 	github.com/go-task/slim-sprig/v3 v3.0.0 // indirect
+	github.com/go-viper/mapstructure/v2 v2.0.0 // indirect
 	github.com/google/s2a-go v0.1.7 // indirect
 	github.com/googleapis/enterprise-certificate-proxy v0.3.2 // indirect
 	github.com/googleapis/gax-go/v2 v2.12.4 // indirect
 	github.com/json-iterator/go v1.1.12 // indirect
-	github.com/magiconair/properties v1.8.7 // indirect
 	github.com/moby/docker-image-spec v1.3.1 // indirect
-	github.com/moby/patternmatcher v0.6.0 // indirect
-	github.com/moby/sys/user v0.1.0 // indirect
 	github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd // indirect
 	github.com/modern-go/reflect2 v1.0.2 // indirect
-	github.com/morikuni/aec v1.0.0 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/pion/datachannel v1.5.10 // indirect
 	github.com/pion/dtls/v2 v2.2.12 // indirect
@@ -118,7 +110,6 @@ require (
 	github.com/pion/turn/v2 v2.1.6 // indirect
 	github.com/pion/turn/v4 v4.0.0 // indirect
 	github.com/pion/webrtc/v4 v4.0.9 // indirect
-	github.com/rs/dnscache v0.0.0-20230804202142-fc85eb664529 // indirect
 	github.com/savsgio/gotils v0.0.0-20230208104028-c358bd845dee // indirect
 	github.com/shirou/gopsutil/v4 v4.24.7 // indirect
 	github.com/wlynxg/anet v0.0.5 // indirect
@@ -137,6 +128,7 @@ require (
 	github.com/Masterminds/semver/v3 v3.3.0 // indirect
 	github.com/Microsoft/go-winio v0.6.2 // indirect
 	github.com/Microsoft/hcsshim v0.11.7 // indirect
+	github.com/Nvveen/Gotty v0.0.0-20120604004816-cd527374f1e5 // indirect
 	github.com/StackExchange/wmi v1.2.1 // indirect
 	github.com/alecthomas/chroma/v2 v2.8.0 // indirect
 	github.com/andybalholm/brotli v1.1.0 // indirect
@@ -160,7 +152,7 @@ require (
 	github.com/dlclark/regexp2 v1.10.0 // indirect
 	github.com/docker/cli v27.0.3+incompatible // indirect
 	github.com/docker/distribution v2.8.2+incompatible // indirect
-	github.com/docker/docker v27.1.1+incompatible
+	github.com/docker/docker v27.0.3+incompatible
 	github.com/docker/docker-credential-helpers v0.7.0 // indirect
 	github.com/docker/go-connections v0.5.0 // indirect
 	github.com/docker/go-units v0.5.0 // indirect
@@ -189,13 +181,14 @@ require (
 	github.com/google/go-cmp v0.6.0 // indirect
 	github.com/google/gopacket v1.1.19 // indirect
 	github.com/google/pprof v0.0.0-20250208200701-d0013a598941 // indirect
+	github.com/google/shlex v0.0.0-20191202100458-e7afc7fbc510 // indirect
 	github.com/gorilla/css v1.0.1 // indirect
 	github.com/gorilla/websocket v1.5.3 // indirect
 	github.com/hashicorp/errwrap v1.1.0 // indirect
 	github.com/hashicorp/go-multierror v1.1.1 // indirect
 	github.com/hashicorp/golang-lru v1.0.2 // indirect
 	github.com/hashicorp/golang-lru/v2 v2.0.7 // indirect
-	github.com/henvic/httpretty v0.1.4 // indirect
+	github.com/henvic/httpretty v0.1.3 // indirect
 	github.com/huandu/xstrings v1.5.0 // indirect
 	github.com/huin/goupnp v1.3.0 // indirect
 	github.com/ipfs/boxo v0.27.4 // indirect
@@ -262,6 +255,7 @@ require (
 	github.com/olekukonko/tablewriter v0.0.5 // indirect
 	github.com/opencontainers/go-digest v1.0.0 // indirect
 	github.com/opencontainers/image-spec v1.1.0
+	github.com/opencontainers/runc v1.1.12 // indirect
 	github.com/opencontainers/runtime-spec v1.2.0 // indirect
 	github.com/opentracing/opentracing-go v1.2.0 // indirect
 	github.com/pbnjay/memory v0.0.0-20210728143218-7b4eea64cf58 // indirect
@@ -284,7 +278,7 @@ require (
 	github.com/shoenig/go-m1cpu v0.1.6 // indirect
 	github.com/shopspring/decimal v1.4.0 // indirect
 	github.com/sirupsen/logrus v1.9.3 // indirect
-	github.com/smallnest/ringbuffer v0.0.0-20241116012123-461381446e3d // indirect
+	github.com/smallnest/ringbuffer v0.0.0-20240423223918-bab516b2000b // indirect
 	github.com/songgao/packets v0.0.0-20160404182456-549a10cd4091 // indirect
 	github.com/spaolacci/murmur3 v1.1.0 // indirect
 	github.com/spf13/cast v1.7.0 // indirect
@@ -299,6 +293,9 @@ require (
 	github.com/vishvananda/netlink v1.3.0 // indirect
 	github.com/vishvananda/netns v0.0.5 // indirect
 	github.com/whyrusleeping/go-keyspace v0.0.0-20160322163242-5b898ac5add1 // indirect
+	github.com/xeipuuv/gojsonpointer v0.0.0-20190905194746-02993c407bfb // indirect
+	github.com/xeipuuv/gojsonreference v0.0.0-20180127040603-bd5ef7bd5415 // indirect
+	github.com/xeipuuv/gojsonschema v1.2.0 // indirect
 	github.com/xi2/xz v0.0.0-20171230120015-48954b6210f8 // indirect
 	github.com/yuin/goldmark v1.5.4 // indirect
 	github.com/yuin/goldmark-emoji v1.0.2 // indirect

go.sum

@@ -17,15 +17,10 @@ dmitri.shuralyov.com/html/belt v0.0.0-20180602232347-f7d459c86be0/go.mod h1:JLBr
 dmitri.shuralyov.com/service/change v0.0.0-20181023043359-a85b471d5412/go.mod h1:a1inKt/atXimZ4Mv927x+r7UpyzRUf4emIoiiSC2TN4=
 dmitri.shuralyov.com/state v0.0.0-20180228185332-28bcc343414c/go.mod h1:0PRwlb0D6DFvNNtx+9ybjezNCa8XF0xaYcETyp6rHWU=
 git.apache.org/thrift.git v0.0.0-20180902110319-2566ecd5d999/go.mod h1:fPE2ZNJGynbRyZ4dJvy6G277gSllfV2HJqblrnkyeyg=
-github.com/AdaLogics/go-fuzz-headers v0.0.0-20230811130428-ced1acdcaa24 h1:bvDV9vkmnHYOMsOr4WLk+Vo07yKIzd94sVoIqshQ4bU=
-github.com/AdaLogics/go-fuzz-headers v0.0.0-20230811130428-ced1acdcaa24/go.mod h1:8o94RPi1/7XTJvwPpRSzSUedZrtlirdB3r9Z20bi2f8=
 github.com/Azure/go-ansiterm v0.0.0-20230124172434-306776ec8161 h1:L/gRVlceqvL25UVaW/CKtUDjefjrs0SPonmDGUVOYP0=
 github.com/Azure/go-ansiterm v0.0.0-20230124172434-306776ec8161/go.mod h1:xomTg63KZ2rFqZQzSB4Vz2SUXa1BpHTVz9L5PTmPC4E=
 github.com/BurntSushi/toml v0.3.1/go.mod h1:xHWCNGjB5oqiDr8zfno3MHue2Ht5sIBksp03qcyfWMU=
 github.com/BurntSushi/toml v1.2.1/go.mod h1:CxXYINrC8qIiEnFrOxCa7Jy5BFHlXnUU2pbicEuybxQ=
-github.com/GeertJohan/go.incremental v1.0.0/go.mod h1:6fAjUhbVuX1KcMD3c8TEgVUqmo4seqhv0i0kdATSkM0=
-github.com/GeertJohan/go.rice v1.0.3 h1:k5viR+xGtIhF61125vCE1cmJ5957RQGXG6dmbaWZSmI=
-github.com/GeertJohan/go.rice v1.0.3/go.mod h1:XVdrU4pW00M4ikZed5q56tPf1v2KwnIKeIdc9CBYNt4=
 github.com/KyleBanks/depth v1.2.1 h1:5h8fQADFrWtarTdtDudMmGsC7GPbOAu6RVB3ffsVFHc=
 github.com/KyleBanks/depth v1.2.1/go.mod h1:jzSb9d0L43HxTQfT+oSA1EEp2q+ne2uh6XgeJcm8brE=
 github.com/MakeNowJust/heredoc v1.0.0 h1:cXCdzVdstXyiTqTvfqk9SDHpKNjxuom+DOlyEeQ4pzQ=
@@ -40,9 +35,10 @@ github.com/Microsoft/go-winio v0.6.2 h1:F2VQgta7ecxGYO8k3ZZz3RS8fVIXVxONVUPlNERo
 github.com/Microsoft/go-winio v0.6.2/go.mod h1:yd8OoFMLzJbo9gZq8j5qaps8bJ9aShtEA8Ipt1oGCvU=
 github.com/Microsoft/hcsshim v0.11.7 h1:vl/nj3Bar/CvJSYo7gIQPyRWc9f3c6IeSNavBTSZNZQ=
 github.com/Microsoft/hcsshim v0.11.7/go.mod h1:MV8xMfmECjl5HdO7U/3/hFVnkmSBjAjmA09d4bExKcU=
+github.com/Nvveen/Gotty v0.0.0-20120604004816-cd527374f1e5 h1:TngWCqHvy9oXAN6lEVMRuU21PR1EtLVZJmdB18Gu3Rw=
+github.com/Nvveen/Gotty v0.0.0-20120604004816-cd527374f1e5/go.mod h1:lmUJ/7eu/Q8D7ML55dXQrVaamCz2vxCfdQBasLZfHKk=
 github.com/StackExchange/wmi v1.2.1 h1:VIkavFPXSjcnS+O8yTq7NI32k0R5Aj+v39y29VYDOSA=
 github.com/StackExchange/wmi v1.2.1/go.mod h1:rcmrprowKIVzvc+NUiLncP2uuArMWLCbu9SBzvHz7e8=
-github.com/akavel/rsrc v0.8.0/go.mod h1:uLoCtb9J+EyAqh+26kdrTgmzRBFPGOolLWKpdxkKq+c=
 github.com/alecthomas/assert/v2 v2.6.0 h1:o3WJwILtexrEUk3cUVal3oiQY2tfgr/FHWiz/v2n4FU=
 github.com/alecthomas/assert/v2 v2.6.0/go.mod h1:Bze95FyfUr7x34QZrjL+XP+0qgp/zg8yS+TtBj1WA3k=
 github.com/alecthomas/chroma/v2 v2.8.0 h1:w9WJUjFFmHHB2e8mRpL9jjy3alYDlU0QLDezj1xE264=
@@ -97,16 +93,12 @@ github.com/containerd/errdefs v0.1.0 h1:m0wCRBiu1WJT/Fr+iOoQHMQS/eP5myQ8lCv4Dz5Z
 github.com/containerd/errdefs v0.1.0/go.mod h1:YgWiiHtLmSeBrvpw+UfPijzbLaB77mEG1WwJTDETIV0=
 github.com/containerd/log v0.1.0 h1:TCJt7ioM2cr/tfR8GPbGf9/VRAX8D2B4PjzCpfX540I=
 github.com/containerd/log v0.1.0/go.mod h1:VRRf09a7mHDIRezVKTRCrOq78v577GXq3bSa3EhrzVo=
-github.com/containerd/platforms v0.2.1 h1:zvwtM3rz2YHPQsF2CHYM8+KtB5dvhISiXh5ZpSBQv6A=
-github.com/containerd/platforms v0.2.1/go.mod h1:XHCb+2/hzowdiut9rkudds9bE5yJ7npe7dG/wG+uFPw=
 github.com/containerd/stargz-snapshotter/estargz v0.14.3 h1:OqlDCK3ZVUO6C3B/5FSkDwbkEETK84kQgEeFwDC+62k=
 github.com/containerd/stargz-snapshotter/estargz v0.14.3/go.mod h1:KY//uOCIkSuNAHhJogcZtrNHdKrA99/FCCRjE3HD36o=
 github.com/coreos/go-systemd v0.0.0-20181012123002-c6f51f82210d/go.mod h1:F5haX7vjVVG0kc13fIWeqUViNPyEJxv/OmvnBo0Yme4=
 github.com/coreos/go-systemd/v22 v22.1.0/go.mod h1:xO0FLkIi5MaZafQlIrOotqXZ90ih+1atmu1JpKERPPk=
 github.com/coreos/go-systemd/v22 v22.5.0 h1:RrqgGjYQKalulkV8NGVIfkXQf6YYmOyiJKk8iXXhfZs=
 github.com/coreos/go-systemd/v22 v22.5.0/go.mod h1:Y58oyj3AT4RCenI/lSvhwexgC+NSVTIJ3seZv2GcEnc=
-github.com/cpuguy83/dockercfg v0.3.2 h1:DlJTyZGBDlXqUZ2Dk2Q3xHs/FtnooJJVaad2S9GKorA=
-github.com/cpuguy83/dockercfg v0.3.2/go.mod h1:sugsbF4//dDlL/i+S+rtpIWp+5h0BHJHfjj5/jFyUJc=
 github.com/cpuguy83/go-md2man/v2 v2.0.0-20190314233015-f79a8a8ca69d/go.mod h1:maD7wRr/U5Z6m/iR4s+kqSMx2CaBsrgA7czyZG/E6dU=
 github.com/cpuguy83/go-md2man/v2 v2.0.0/go.mod h1:maD7wRr/U5Z6m/iR4s+kqSMx2CaBsrgA7czyZG/E6dU=
 github.com/cpuguy83/go-md2man/v2 v2.0.2/go.mod h1:tgQtvFlXSQOSOSIRvRPT7W67SCa46tRHOmNcaadrF8o=
@@ -116,8 +108,6 @@ github.com/creachadair/otp v0.5.0 h1:q3Th7CXm2zlmCdBjw5tEPFOj4oWJMnVL5HXlq0sNKS0
 github.com/creachadair/otp v0.5.0/go.mod h1:0kceI87EnYFNYSTL121goJVAnk3eJhaed9H0nMuJUkA=
 github.com/creack/pty v1.1.18 h1:n56/Zwd5o6whRC5PMGretI4IdRLlmBXYNjScPaBgsbY=
 github.com/creack/pty v1.1.18/go.mod h1:MOBLtS5ELjhRRrroQr9kyvTxUAFNvYEK993ew/Vr4O4=
-github.com/daaku/go.zipexe v1.0.2 h1:Zg55YLYTr7M9wjKn8SY/WcpuuEi+kR2u4E8RhvpyXmk=
-github.com/daaku/go.zipexe v1.0.2/go.mod h1:5xWogtqlYnfBXkSB1o9xysukNP9GTvaNkqzUZbt3Bw8=
 github.com/dave-gray101/v2keyauth v0.0.0-20240624150259-c45d584d25e2 h1:flLYmnQFZNo04x2NPehMbf30m7Pli57xwZ0NFqR/hb0=
 github.com/dave-gray101/v2keyauth v0.0.0-20240624150259-c45d584d25e2/go.mod h1:NtWqRzAp/1tw+twkW8uuBenEVVYndEAZACWU3F3xdoQ=
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
@@ -129,16 +119,14 @@ github.com/decred/dcrd/crypto/blake256 v1.0.1 h1:7PltbUIQB7u/FfZ39+DGa/ShuMyJ5il
 github.com/decred/dcrd/crypto/blake256 v1.0.1/go.mod h1:2OfgNZ5wDpcsFmHmCK5gZTPcCXqlm2ArzUIkw9czNJo=
 github.com/decred/dcrd/dcrec/secp256k1/v4 v4.3.0 h1:rpfIENRNNilwHwZeG5+P150SMrnNEcHYvcCuK6dPZSg=
 github.com/decred/dcrd/dcrec/secp256k1/v4 v4.3.0/go.mod h1:v57UDF4pDQJcEfFUCRop3lJL149eHGSe9Jvczhzjo/0=
-github.com/distribution/reference v0.6.0 h1:0IXCQ5g4/QMHHkarYzh5l+u8T3t73zM5QvfrDyIgxBk=
-github.com/distribution/reference v0.6.0/go.mod h1:BbU0aIcezP1/5jX/8MP0YiH4SdvB5Y4f/wlDRiLyi3E=
 github.com/dlclark/regexp2 v1.10.0 h1:+/GIL799phkJqYW+3YbOd8LCcbHzT0Pbo8zl70MHsq0=
 github.com/dlclark/regexp2 v1.10.0/go.mod h1:DHkYz0B9wPfa6wondMfaivmHpzrQ3v9q8cnmRbL6yW8=
 github.com/docker/cli v27.0.3+incompatible h1:usGs0/BoBW8MWxGeEtqPMkzOY56jZ6kYlSN5BLDioCQ=
 github.com/docker/cli v27.0.3+incompatible/go.mod h1:JLrzqnKDaYBop7H2jaqPtU4hHvMKP+vjCwu2uszcLI8=
 github.com/docker/distribution v2.8.2+incompatible h1:T3de5rq0dB1j30rp0sA2rER+m322EBzniBPB6ZIzuh8=
 github.com/docker/distribution v2.8.2+incompatible/go.mod h1:J2gT2udsDAN96Uj4KfcMRqY0/ypR+oyYUYmja8H+y+w=
-github.com/docker/docker v27.1.1+incompatible h1:hO/M4MtV36kzKldqnA37IWhebRA+LnqqcqDja6kVaKY=
-github.com/docker/docker v27.1.1+incompatible/go.mod h1:eEKB0N0r5NX/I1kEveEz05bcu8tLC/8azJZsviup8Sk=
+github.com/docker/docker v27.0.3+incompatible h1:aBGI9TeQ4MPlhquTQKq9XbK79rKFVwXNUAYz9aXyEBE=
+github.com/docker/docker v27.0.3+incompatible/go.mod h1:eEKB0N0r5NX/I1kEveEz05bcu8tLC/8azJZsviup8Sk=
 github.com/docker/docker-credential-helpers v0.7.0 h1:xtCHsjxogADNZcdv1pKUHXryefjlVRqWqIhk/uXJp0A=
 github.com/docker/docker-credential-helpers v0.7.0/go.mod h1:rETQfLdHNT3foU5kuNkFR1R1V12OJRRO5lzt2D1b5X0=
 github.com/docker/go-connections v0.5.0 h1:USnMq7hx7gwdVZq1L49hLXaFtUdTADjXGp+uj1Br63c=
@@ -177,6 +165,8 @@ github.com/frankban/quicktest v1.14.6/go.mod h1:4ptaffx2x8+WTWXmUCuVU6aPUX1/Mz7z
 github.com/fsnotify/fsnotify v1.4.7/go.mod h1:jwhsz4b93w/PPRr/qN1Yymfu8t87LnFCMoQvtojpjFo=
 github.com/fsnotify/fsnotify v1.7.0 h1:8JEhPFa5W2WU7YfeZzPNqzMP6Lwt7L2715Ggo0nosvA=
 github.com/fsnotify/fsnotify v1.7.0/go.mod h1:40Bi/Hjc2AVfZrqy+aj+yEI+/bRxZnMJyTJwOpGvigM=
+github.com/ggerganov/whisper.cpp/bindings/go v0.0.0-20240626202019-c118733a29ad h1:dQ93Vd6i25o+zH9vvnZ8mu7jtJQ6jT3D+zE3V8Q49n0=
+github.com/ggerganov/whisper.cpp/bindings/go v0.0.0-20240626202019-c118733a29ad/go.mod h1:QIjZ9OktHFG7p+/m3sMvrAJKKdWrr1fZIK0rM6HZlyo=
 github.com/ghodss/yaml v1.0.0 h1:wQHKEahhL6wmXdzwWG11gIVCkOv05bNOh+Rxn0yngAk=
 github.com/ghodss/yaml v1.0.0/go.mod h1:4dBDuWmgqj2HViK6kFavaiC9ZROes6MMH2rRYeMEF04=
 github.com/gliderlabs/ssh v0.1.1/go.mod h1:U7qILu1NlMHj9FlMhZLlkCdDnU1DBEAqr0aevW3Awn0=
@@ -206,8 +196,12 @@ github.com/go-openapi/swag v0.23.0 h1:vsEVJDUo2hPJ2tu0/Xc+4noaxyEffXNIs3cOULZ+Gr
 github.com/go-openapi/swag v0.23.0/go.mod h1:esZ8ITTYEsH1V2trKHjAN8Ai7xHb8RV+YSZ577vPjgQ=
 github.com/go-skynet/go-llama.cpp v0.0.0-20240314183750-6a8041ef6b46 h1:lALhXzDkqtp12udlDLLg+ybXVMmL7Ox9tybqVLWxjPE=
 github.com/go-skynet/go-llama.cpp v0.0.0-20240314183750-6a8041ef6b46/go.mod h1:iub0ugfTnflE3rcIuqV2pQSo15nEw3GLW/utm5gyERo=
+github.com/go-sql-driver/mysql v1.7.1 h1:lUIinVbN1DY0xBg0eMOzmmtGoHwWBbvnWubQUrtU8EI=
+github.com/go-sql-driver/mysql v1.7.1/go.mod h1:OXbVy3sEdcQ2Doequ6Z5BW6fXNQTmx+9S1MCJN5yJMI=
 github.com/go-task/slim-sprig/v3 v3.0.0 h1:sUs3vkvUymDpBKi3qH1YSqBQk9+9D/8M2mN1vB6EwHI=
 github.com/go-task/slim-sprig/v3 v3.0.0/go.mod h1:W848ghGpv3Qj3dhTPRyJypKRiqCdHZiAzKg9hl15HA8=
+github.com/go-viper/mapstructure/v2 v2.0.0 h1:dhn8MZ1gZ0mzeodTG3jt5Vj/o87xZKuNAprG2mQfMfc=
+github.com/go-viper/mapstructure/v2 v2.0.0/go.mod h1:oJDH3BJKyqBA2TXFhDsKDGDTlndYOZ6rGS0BRZIxGhM=
 github.com/go-yaml/yaml v2.1.0+incompatible/go.mod h1:w2MrLa16VYP0jy6N7M5kHaCkaLENm+P+Tv+MfurjSw0=
 github.com/godbus/dbus/v5 v5.0.3/go.mod h1:xhWf0FNVPg57R7Z0UbKHbJfkEywrmjJnf7w5xrFpKfA=
 github.com/godbus/dbus/v5 v5.0.4/go.mod h1:xhWf0FNVPg57R7Z0UbKHbJfkEywrmjJnf7w5xrFpKfA=
@@ -282,6 +276,8 @@ github.com/google/pprof v0.0.0-20250208200701-d0013a598941/go.mod h1:vavhavw2zAx
 github.com/google/renameio v0.1.0/go.mod h1:KWCgfxg9yswjAJkECMjeO8J8rahYeXnNhOm40UhjYkI=
 github.com/google/s2a-go v0.1.7 h1:60BLSyTrOV4/haCDW4zb1guZItoSq8foHCXrAnjBo/o=
 github.com/google/s2a-go v0.1.7/go.mod h1:50CgR4k1jNlWBu4UfS4AcfhVe1r6pdZPygJ3R8F0Qdw=
+github.com/google/shlex v0.0.0-20191202100458-e7afc7fbc510 h1:El6M4kTTCOh6aBiKaUGG7oYTSPP8MxqL4YI3kZKwcP4=
+github.com/google/shlex v0.0.0-20191202100458-e7afc7fbc510/go.mod h1:pupxD2MaaD3pAXIBCelhxNneeOaAeabZDe5s4K6zSpQ=
 github.com/google/uuid v1.1.2/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/google/uuid v1.3.1/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
@@ -299,13 +295,9 @@ github.com/gorilla/css v1.0.1 h1:ntNaBIghp6JmvWnxbZKANoLyuXTPZ4cAMlo6RyhlbO8=
 github.com/gorilla/css v1.0.1/go.mod h1:BvnYkspnSzMmwRK+b8/xgNPLiIuNZr6vbZBTPQ2A3b0=
 github.com/gorilla/websocket v1.5.3 h1:saDtZ6Pbx/0u+bgYQ3q96pZgCzfhKXGPqt7kZ72aNNg=
 github.com/gorilla/websocket v1.5.3/go.mod h1:YR8l580nyteQvAITg2hZ9XVh4b55+EU/adAjf1fMHhE=
-github.com/gpustack/gguf-parser-go v0.17.0 h1:DkSziWLsiQM0pqqkr/zMcaBn94KY7iQTi4zmaHixDus=
-github.com/gpustack/gguf-parser-go v0.17.0/go.mod h1:GvHh1Kvvq5ojCOsJ5UpwiJJmIjFw3Qk5cW7R+CZ3IJo=
 github.com/gregjones/httpcache v0.0.0-20180305231024-9cad4c3443a7/go.mod h1:FecbI9+v66THATjSRHfNgh1IVFe/9kFxbXtjV0ctIMA=
 github.com/grpc-ecosystem/grpc-gateway v1.5.0 h1:WcmKMm43DR7RdtlkEXQJyo5ws8iTp98CyhCCbOHMvNI=
 github.com/grpc-ecosystem/grpc-gateway v1.5.0/go.mod h1:RSKVYQBd5MCa4OVpNdGskqpgL2+G+NZTnrVHpWWfpdw=
-github.com/grpc-ecosystem/grpc-gateway/v2 v2.22.0 h1:asbCHRVmodnJTuQ3qamDwqVOIjwqUPTYmYuemVOx+Ys=
-github.com/grpc-ecosystem/grpc-gateway/v2 v2.22.0/go.mod h1:ggCgvZ2r7uOoQjOyu2Y1NhHmEPPzzuhWgcza5M1Ji1I=
 github.com/hashicorp/errwrap v1.0.0/go.mod h1:YH+1FKiLXxHSkmPseP+kNlulaMuP3n2brvKWEqk/Jc4=
 github.com/hashicorp/errwrap v1.1.0 h1:OxrOeh75EUXMY8TBjag2fzXGZ40LB6IKw45YeGUDY2I=
 github.com/hashicorp/errwrap v1.1.0/go.mod h1:YH+1FKiLXxHSkmPseP+kNlulaMuP3n2brvKWEqk/Jc4=
@@ -315,8 +307,8 @@ github.com/hashicorp/golang-lru v1.0.2 h1:dV3g9Z/unq5DpblPpw+Oqcv4dU/1omnb4Ok8iP
 github.com/hashicorp/golang-lru v1.0.2/go.mod h1:iADmTwqILo4mZ8BN3D2Q6+9jd8WM5uGBxy+E8yxSoD4=
 github.com/hashicorp/golang-lru/v2 v2.0.7 h1:a+bsQ5rvGLjzHuww6tVxozPZFVghXaHOwFs4luLUK2k=
 github.com/hashicorp/golang-lru/v2 v2.0.7/go.mod h1:QeFd9opnmA6QUJc5vARoKUSoFhyfM2/ZepoAG6RGpeM=
-github.com/henvic/httpretty v0.1.4 h1:Jo7uwIRWVFxkqOnErcoYfH90o3ddQyVrSANeS4cxYmU=
-github.com/henvic/httpretty v0.1.4/go.mod h1:Dn60sQTZfbt2dYsdUSNsCljyF4AfdqnuJFDLJA1I4AM=
+github.com/henvic/httpretty v0.1.3 h1:4A6vigjz6Q/+yAfTD4wqipCv+Px69C7Th/NhT0ApuU8=
+github.com/henvic/httpretty v0.1.3/go.mod h1:UUEv7c2kHZ5SPQ51uS3wBpzPDibg2U3Y+IaXyHy5GBg=
 github.com/hexops/gotextdiff v1.0.3 h1:gitA9+qJrrTCsiCl7+kh75nPqQt1cx4ZkudSTLoUqJM=
 github.com/hexops/gotextdiff v1.0.3/go.mod h1:pSWU5MAI3yDq+fZBTazCSJysOMbxWL1BSow5/V2vxeg=
 github.com/hpcloud/tail v1.0.0 h1:nfCOvKYfkgYP8hkirhJocXT2+zOD8yUNjXaWfTlyFKI=
@@ -395,6 +387,8 @@ github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
 github.com/kr/text v0.2.0/go.mod h1:eLer722TekiGuMkidMxC/pM04lWEeraHUUmBw8l2grE=
 github.com/kylelemons/godebug v1.1.0 h1:RPNrshWIDI6G2gRW9EHilWtl7Z6Sb1BR0xunSBf0SNc=
 github.com/kylelemons/godebug v1.1.0/go.mod h1:9/0rRGxNHcop5bhtWyNeEfOS8JIWk580+fNqagV/RAw=
+github.com/lib/pq v0.0.0-20180327071824-d34b9ff171c2 h1:hRGSmZu7j271trc9sneMrpOW7GN5ngLm8YUZIPzf394=
+github.com/lib/pq v0.0.0-20180327071824-d34b9ff171c2/go.mod h1:5WUZQaWbwv1U+lTReE5YruASi9Al49XbQIvNi/34Woo=
 github.com/libp2p/go-buffer-pool v0.1.0 h1:oK4mSFcQz7cTQIfqbe4MIj9gLW+mnanjyFtc6cdF0Y8=
 github.com/libp2p/go-buffer-pool v0.1.0/go.mod h1:N+vh8gMqimBzdKkSMVuydVDq+UV5QTWy5HSiZacSbPg=
 github.com/libp2p/go-cidranger v1.1.0 h1:ewPN8EZ0dd1LSnrtuwd4709PXVcITVeuwbag38yPW7c=
@@ -434,8 +428,6 @@ github.com/lucasb-eyer/go-colorful v1.2.0/go.mod h1:R4dSotOR9KMtayYi1e77YzuveK+i
 github.com/lufia/plan9stats v0.0.0-20240819163618-b1d8f4d146e7 h1:5RK988zAqB3/AN3opGfRpoQgAVqr6/A5+qRTi67VUZY=
 github.com/lufia/plan9stats v0.0.0-20240819163618-b1d8f4d146e7/go.mod h1:ilwx/Dta8jXAgpFYFvSWEMwxmbWXyiUHkd5FwyKhb5k=
 github.com/lunixbochs/vtclean v1.0.0/go.mod h1:pHhQNgMf3btfWnGBVipUOjRYhoOsdGqdm/+2c2E2WMI=
-github.com/magiconair/properties v1.8.7 h1:IeQXZAiQcpL9mgcAe1Nu6cX9LLw6ExEHKjN0VQdvPDY=
-github.com/magiconair/properties v1.8.7/go.mod h1:Dhd985XPs7jluiymwWYZ0G4Z61jb3vdS329zhj2hYo0=
 github.com/mailru/easyjson v0.0.0-20190312143242-1de009706dbe/go.mod h1:C1wdFJiN94OJF2b5HbByQZoLdCWB1Yqtg26g4irojpc=
 github.com/mailru/easyjson v0.7.7 h1:UGYAvKxe3sBsEDzO8ZeWOSlIQfWFlxbzLZe7hwFURr0=
 github.com/mailru/easyjson v0.7.7/go.mod h1:xzfreul335JAWq5oZzymOObrkdz5UnU4kGfJJLY9Nlc=
@@ -482,14 +474,10 @@ github.com/mitchellh/reflectwalk v1.0.2 h1:G2LzWKi524PWgd3mLHV8Y5k7s6XUvT0Gef6zx
 github.com/mitchellh/reflectwalk v1.0.2/go.mod h1:mSTlrgnPZtwu0c4WaC2kGObEpuNDbx0jmZXqmk4esnw=
 github.com/moby/docker-image-spec v1.3.1 h1:jMKff3w6PgbfSa69GfNg+zN/XLhfXJGnEx3Nl2EsFP0=
 github.com/moby/docker-image-spec v1.3.1/go.mod h1:eKmb5VW8vQEh/BAr2yvVNvuiJuY6UIocYsFu/DxxRpo=
-github.com/moby/patternmatcher v0.6.0 h1:GmP9lR19aU5GqSSFko+5pRqHi+Ohk1O69aFiKkVGiPk=
-github.com/moby/patternmatcher v0.6.0/go.mod h1:hDPoyOpDY7OrrMDLaYoY3hf52gNCR/YOUYxkhApJIxc=
 github.com/moby/sys/mountinfo v0.6.2 h1:BzJjoreD5BMFNmD9Rus6gdd1pLuecOFPt8wC+Vygl78=
 github.com/moby/sys/mountinfo v0.6.2/go.mod h1:IJb6JQeOklcdMU9F5xQ8ZALD+CUr5VlGpwtX+VE0rpI=
 github.com/moby/sys/sequential v0.5.0 h1:OPvI35Lzn9K04PBbCLW0g4LcFAJgHsvXsRyewg5lXtc=
 github.com/moby/sys/sequential v0.5.0/go.mod h1:tH2cOOs5V9MlPiXcQzRC+eEyab644PWKGRYaaV5ZZlo=
-github.com/moby/sys/user v0.1.0 h1:WmZ93f5Ux6het5iituh9x2zAG7NFY9Aqi49jjE1PaQg=
-github.com/moby/sys/user v0.1.0/go.mod h1:fKJhFOnsCN6xZ5gSfbM6zaHGgDJMrqt9/reuj4T7MmU=
 github.com/moby/term v0.5.0 h1:xt8Q1nalod/v7BqbG21f8mQPqH+xAaC9C3N3wfWbVP0=
 github.com/moby/term v0.5.0/go.mod h1:8FzsFHVUBGZdbDsJw/ot+X+d5HLUbvklYLJ9uGfcI3Y=
 github.com/modern-go/concurrent v0.0.0-20180228061459-e0a39a4cb421/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=
@@ -498,13 +486,13 @@ github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd/go.mod h1:6dJ
 github.com/modern-go/reflect2 v1.0.1/go.mod h1:bx2lNnkwVCuqBIxFjflWJWanXIb3RllmbCylyMrvgv0=
 github.com/modern-go/reflect2 v1.0.2 h1:xBagoLtFs94CBntxluKeaWgTMpvLxC4ur3nMaC9Gz0M=
 github.com/modern-go/reflect2 v1.0.2/go.mod h1:yWuevngMOJpCy52FWWMvUC8ws7m/LJsjYzDa0/r8luk=
-github.com/morikuni/aec v1.0.0 h1:nP9CBfwrvYnBRgY6qfDQkygYDmYwOilePFkwzv4dU8A=
-github.com/morikuni/aec v1.0.0/go.mod h1:BbKIizmSmc5MMPqRYbxO4ZU0S0+P200+tUnFx7PXmsc=
 github.com/mr-tron/base58 v1.1.2/go.mod h1:BinMc/sQntlIE1frQmRFPUoPA1Zkr8VRgBdjWI2mNwc=
 github.com/mr-tron/base58 v1.2.0 h1:T/HDJBh4ZCPbU39/+c3rRvE0uKBQlU27+QI8LJ4t64o=
 github.com/mr-tron/base58 v1.2.0/go.mod h1:BinMc/sQntlIE1frQmRFPUoPA1Zkr8VRgBdjWI2mNwc=
 github.com/mudler/edgevpn v0.30.1 h1:4yyhNFJX62NpRp50sxiyZE5E/sdAqEZX+aE5Mv7QS60=
 github.com/mudler/edgevpn v0.30.1/go.mod h1:IAJkkJ0oH3rwsSGOGTFT4UBYFqYuD/QyaKzTLB3P/eU=
+github.com/mudler/go-piper v0.0.0-20241023091659-2494246fd9fc h1:RxwneJl1VgvikiX28EkpdAyL4yQVnJMrbquKospjHyA=
+github.com/mudler/go-piper v0.0.0-20241023091659-2494246fd9fc/go.mod h1:O7SwdSWMilAWhBZMK9N9Y/oBDyMMzshE3ju8Xkexwig=
 github.com/mudler/go-processmanager v0.0.0-20240820160718-8b802d3ecf82 h1:FVT07EI8njvsD4tC2Hw8Xhactp5AWhsQWD4oTeQuSAU=
 github.com/mudler/go-processmanager v0.0.0-20240820160718-8b802d3ecf82/go.mod h1:Urp7LG5jylKoDq0663qeBh0pINGcRl35nXdKx82PSoU=
 github.com/mudler/water v0.0.0-20221010214108-8c7313014ce0 h1:Qh6ghkMgTu6siFbTf7L3IszJmshMhXxNL4V+t7IIA6w=
@@ -541,7 +529,6 @@ github.com/neelance/astrewrite v0.0.0-20160511093645-99348263ae86/go.mod h1:kHJE
 github.com/neelance/sourcemap v0.0.0-20151028013722-8c68805598ab/go.mod h1:Qr6/a/Q4r9LP1IltGz7tA7iOK1WonHEYhu1HRBA7ZiM=
 github.com/nikolalohinski/gonja/v2 v2.3.2 h1:UgLFfqi7L9XfX0PEcE4eUpvGojVQL5KhBfJJaBp7ZxY=
 github.com/nikolalohinski/gonja/v2 v2.3.2/go.mod h1:1Wcc/5huTu6y36e0sOFR1XQoFlylw3c3H3L5WOz0RDg=
-github.com/nkovacs/streamquote v1.0.0/go.mod h1:BN+NaZ2CmdKqUuTUXUEm9j95B2TRbpOWpxbJYzzgUsc=
 github.com/nwaples/rardecode v1.1.0 h1:vSxaY8vQhOcVr4mm5e8XllHWTiM4JF507A0Katqw7MQ=
 github.com/nwaples/rardecode v1.1.0/go.mod h1:5DzqNKiOdpKKBH87u8VlvAnPZMXcGRhxWkRpHbbfGS0=
 github.com/nxadm/tail v1.4.11 h1:8feyoE3OzPrcshW5/MJ4sGESc5cqmGkGCWlco4l0bqY=
@@ -558,6 +545,8 @@ github.com/opencontainers/go-digest v1.0.0 h1:apOUWs51W5PlhuyGyz9FCeeBIOUDA/6nW8
 github.com/opencontainers/go-digest v1.0.0/go.mod h1:0JzlMkj0TRzQZfJkVvzbP0HBR3IKzErnv2BNG4W4MAM=
 github.com/opencontainers/image-spec v1.1.0 h1:8SG7/vwALn54lVB/0yZ/MMwhFrPYtpEHQb2IpWsCzug=
 github.com/opencontainers/image-spec v1.1.0/go.mod h1:W4s4sFTMaBeK1BQLXbG4AdM2szdn85PY75RI83NrTrM=
+github.com/opencontainers/runc v1.1.12 h1:BOIssBaW1La0/qbNZHXOOa71dZfZEQOzW7dqQf3phss=
+github.com/opencontainers/runc v1.1.12/go.mod h1:S+lQwSfncpBha7XTy/5lBwWgm5+y5Ma/O44Ekby9FK8=
 github.com/opencontainers/runtime-spec v1.0.2/go.mod h1:jwyrGlmzljRJv/Fgzds9SsS/C5hL+LL3ko9hs6T5lQ0=
 github.com/opencontainers/runtime-spec v1.2.0 h1:z97+pHb3uELt/yiAWD691HNHQIF07bE7dzrbT927iTk=
 github.com/opencontainers/runtime-spec v1.2.0/go.mod h1:jwyrGlmzljRJv/Fgzds9SsS/C5hL+LL3ko9hs6T5lQ0=
@@ -566,6 +555,8 @@ github.com/opentracing/opentracing-go v1.2.0/go.mod h1:GxEUsuufX4nBwe+T+Wl9TAgYr
 github.com/openzipkin/zipkin-go v0.1.1/go.mod h1:NtoC/o8u3JlF1lSlyPNswIbeQH9bJTmOf0Erfk+hxe8=
 github.com/orcaman/writerseeker v0.0.0-20200621085525-1d3f536ff85e h1:s2RNOM/IGdY0Y6qfTeUKhDawdHDpK9RGBdx80qN4Ttw=
 github.com/orcaman/writerseeker v0.0.0-20200621085525-1d3f536ff85e/go.mod h1:nBdnFKj15wFbf94Rwfq4m30eAcyY9V/IyKAGQFtqkW0=
+github.com/ory/dockertest/v3 v3.10.0 h1:4K3z2VMe8Woe++invjaTB7VRyQXQy5UY+loujO4aNE4=
+github.com/ory/dockertest/v3 v3.10.0/go.mod h1:nr57ZbRWMqfsdGdFNLHz5jjNdDb7VVFnzAeW1n5N1Lg=
 github.com/otiai10/mint v1.6.1 h1:kgbTJmOpp/0ce7hk3H8jiSuR0MXmpwWRfqUdKww17qg=
 github.com/otiai10/mint v1.6.1/go.mod h1:MJm72SBthJjz8qhefc4z1PYEieWmy8Bku7CjcAqyUSM=
 github.com/otiai10/openaigo v1.7.0 h1:AOQcOjRRM57ABvz+aI2oJA/Qsz1AydKbdZAlGiKyCqg=
@@ -669,8 +660,6 @@ github.com/rivo/uniseg v0.4.7/go.mod h1:FN3SvrM+Zdj16jyLfmOkMNblXMcoc8DfTHruCPUc
 github.com/rogpeppe/go-internal v1.3.0/go.mod h1:M8bDsm7K2OlrFYOpmOWEs/qY81heoFRclV5y23lUDJ4=
 github.com/rogpeppe/go-internal v1.13.1 h1:KvO1DLK/DRN07sQ1LQKScxyZJuNnedQ5/wKSR38lUII=
 github.com/rogpeppe/go-internal v1.13.1/go.mod h1:uMEvuHeurkdAXX61udpOXGD/AzZDWNMNyH2VO9fmH0o=
-github.com/rs/dnscache v0.0.0-20230804202142-fc85eb664529 h1:18kd+8ZUlt/ARXhljq+14TwAoKa61q6dX8jtwOf6DH8=
-github.com/rs/dnscache v0.0.0-20230804202142-fc85eb664529/go.mod h1:qe5TWALJ8/a1Lqznoc5BDHpYX/8HU60Hm2AwRmqzxqA=
 github.com/rs/xid v1.5.0/go.mod h1:trrq9SKmegXys3aeAKXMUTdJsYXVwGY3RLcfgqegfbg=
 github.com/rs/zerolog v1.33.0 h1:1cU2KZkvPxNyfgEmhHAz/1A9Bz+llsdYzklWFzgp0r8=
 github.com/rs/zerolog v1.33.0/go.mod h1:/7mN4D5sKwJLZQ2b/znpjC3/GQWY/xaDXUM0kKWRHss=
@@ -723,8 +712,8 @@ github.com/sirupsen/logrus v1.7.0/go.mod h1:yWOB1SBYBC5VeMP7gHvWumXLIWorT60ONWic
 github.com/sirupsen/logrus v1.9.0/go.mod h1:naHLuLoDiP4jHNo9R0sCBMtWGeIprob74mVsIT4qYEQ=
 github.com/sirupsen/logrus v1.9.3 h1:dueUQJ1C2q9oE3F7wvmSGAaVtTmUizReu6fjN8uqzbQ=
 github.com/sirupsen/logrus v1.9.3/go.mod h1:naHLuLoDiP4jHNo9R0sCBMtWGeIprob74mVsIT4qYEQ=
-github.com/smallnest/ringbuffer v0.0.0-20241116012123-461381446e3d h1:3VwvTjiRPA7cqtgOWddEL+JrcijMlXUmj99c/6YyZoY=
-github.com/smallnest/ringbuffer v0.0.0-20241116012123-461381446e3d/go.mod h1:tAG61zBM1DYRaGIPloumExGvScf08oHuo0kFoOqdbT0=
+github.com/smallnest/ringbuffer v0.0.0-20240423223918-bab516b2000b h1:e9eeuSYSLmUKxy7ALzKcxo7ggTceQaVcBhjDIcewa9c=
+github.com/smallnest/ringbuffer v0.0.0-20240423223918-bab516b2000b/go.mod h1:tAG61zBM1DYRaGIPloumExGvScf08oHuo0kFoOqdbT0=
 github.com/smartystreets/assertions v1.2.0/go.mod h1:tcbTF8ujkAEcZ8TElKY+i30BzYlVhC/LOxJk7iOWnoo=
 github.com/smartystreets/assertions v1.13.0 h1:Dx1kYM01xsSqKPno3aqLnrwac2LetPvN23diwyr69Qs=
 github.com/smartystreets/assertions v1.13.0/go.mod h1:wDmR7qL282YbGsPy6H/yAsesrxfxaaSlJazyFLYVFx8=
@@ -743,8 +732,6 @@ github.com/streamer45/silero-vad-go v0.2.1/go.mod h1:B+2FXs/5fZ6pzl6unUZYhZqkYdO
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
 github.com/stretchr/objx v0.4.0/go.mod h1:YvHI0jy2hoMjB+UWwv71VJQ9isScKT/TqJzVSSt89Yw=
 github.com/stretchr/objx v0.5.0/go.mod h1:Yh+to48EsGEfYuaHDzXPcE3xhTkx73EhmCGUpEOglKo=
-github.com/stretchr/objx v0.5.2 h1:xuMeJ0Sdp5ZMRXx/aWO6RZxdr3beISkG5/G/aIRr3pY=
-github.com/stretchr/objx v0.5.2/go.mod h1:FRsXN1f5AsAjCGJKqEizvkpNtU+EGNCLh3NxZ/8L+MA=
 github.com/stretchr/testify v1.2.2/go.mod h1:a8OnRcib4nhh0OaRAV+Yts87kKdq0PP7pXfy6kDkUVs=
 github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=
 github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81PSLYec5m4=
@@ -761,8 +748,8 @@ github.com/swaggo/files/v2 v2.0.0/go.mod h1:24kk2Y9NYEJ5lHuCra6iVwkMjIekMCaFq/0J
 github.com/swaggo/swag v1.16.3 h1:PnCYjPCah8FK4I26l2F/KQ4yz3sILcVUN3cTlBFA9Pg=
 github.com/swaggo/swag v1.16.3/go.mod h1:DImHIuOFXKpMFAQjcC7FG4m3Dg4+QuUgUzJmKjI/gRk=
 github.com/tarm/serial v0.0.0-20180830185346-98f6abe2eb07/go.mod h1:kDXzergiv9cbyO7IOYJZWg1U88JhDg3PB6klq9Hg2pA=
-github.com/testcontainers/testcontainers-go v0.35.0 h1:uADsZpTKFAtp8SLK+hMwSaa+X+JiERHtd4sQAFmXeMo=
-github.com/testcontainers/testcontainers-go v0.35.0/go.mod h1:oEVBj5zrfJTrgjwONs1SsRbnBtH9OKl+IGl3UMcr2B4=
+github.com/thxcode/gguf-parser-go v0.1.0 h1:J4QruXyEQGjrAKeKZFlsD2na9l4XF5+bjR194d+wJS4=
+github.com/thxcode/gguf-parser-go v0.1.0/go.mod h1:Tn1PsO/YDEtLIxm1+QDCjIIH9L/9Sr7+KpxZKm0sEuE=
 github.com/tinylib/msgp v1.1.8 h1:FCXC1xanKO4I8plpHGH2P7koL/RzZs12l/+r7vakfm0=
 github.com/tinylib/msgp v1.1.8/go.mod h1:qkpG+2ldGg4xRFmx+jfTvZPxfGFhi64BcnL9vkCm/Tw=
 github.com/tklauser/go-sysconf v0.3.14 h1:g5vzr9iPFFz24v2KZXs/pvpvh8/V9Fw6vQK5ZZb78yU=
@@ -781,7 +768,6 @@ github.com/valyala/bytebufferpool v1.0.0 h1:GqA5TC/0021Y/b9FG4Oi9Mr3q7XYx6Kllzaw
 github.com/valyala/bytebufferpool v1.0.0/go.mod h1:6bBcMArwyJ5K/AmCkWv1jt77kVWyCJ6HpOuEn7z0Csc=
 github.com/valyala/fasthttp v1.55.0 h1:Zkefzgt6a7+bVKHnu/YaYSOPfNYNisSVBo/unVCf8k8=
 github.com/valyala/fasthttp v1.55.0/go.mod h1:NkY9JtkrpPKmgwV3HTaS2HWaJss9RSIsRVfcxxoHiOM=
-github.com/valyala/fasttemplate v1.0.1/go.mod h1:UQGH1tvbgY+Nz5t2n7tXsz52dQxojPUpymEIMZ47gx8=
 github.com/valyala/tcplisten v1.0.0 h1:rBHj/Xf+E1tRGZyWIWwJDiRY0zc1Js+CV5DqwacVSA8=
 github.com/valyala/tcplisten v1.0.0/go.mod h1:T0xQ8SeCZGxckz9qRXTfG43PvQ/mcWh7FwZEA7Ioqkc=
 github.com/vbatts/tar-split v0.11.3 h1:hLFqsOLQ1SsppQNTMpkpPXClLDfC2A3Zgy9OUU+RVck=
@@ -800,6 +786,13 @@ github.com/whyrusleeping/go-keyspace v0.0.0-20160322163242-5b898ac5add1/go.mod h
 github.com/wlynxg/anet v0.0.3/go.mod h1:eay5PRQr7fIVAMbTbchTnO9gG65Hg/uYGdc7mguHxoA=
 github.com/wlynxg/anet v0.0.5 h1:J3VJGi1gvo0JwZ/P1/Yc/8p63SoW98B5dHkYDmpgvvU=
 github.com/wlynxg/anet v0.0.5/go.mod h1:eay5PRQr7fIVAMbTbchTnO9gG65Hg/uYGdc7mguHxoA=
+github.com/xeipuuv/gojsonpointer v0.0.0-20180127040702-4e3ac2762d5f/go.mod h1:N2zxlSyiKSe5eX1tZViRH5QA0qijqEDrYZiPEAiq3wU=
+github.com/xeipuuv/gojsonpointer v0.0.0-20190905194746-02993c407bfb h1:zGWFAtiMcyryUHoUjUJX0/lt1H2+i2Ka2n+D3DImSNo=
+github.com/xeipuuv/gojsonpointer v0.0.0-20190905194746-02993c407bfb/go.mod h1:N2zxlSyiKSe5eX1tZViRH5QA0qijqEDrYZiPEAiq3wU=
+github.com/xeipuuv/gojsonreference v0.0.0-20180127040603-bd5ef7bd5415 h1:EzJWgHovont7NscjpAxXsDA8S8BMYve8Y5+7cuRE7R0=
+github.com/xeipuuv/gojsonreference v0.0.0-20180127040603-bd5ef7bd5415/go.mod h1:GwrjFmJcFw6At/Gs6z4yjiIwzuJ1/+UwLxMQDVQXShQ=
+github.com/xeipuuv/gojsonschema v1.2.0 h1:LhYJRs+L4fBtjZUfuSZIKGeVu0QRy8e5Xi7D17UxZ74=
+github.com/xeipuuv/gojsonschema v1.2.0/go.mod h1:anYRn/JVcOK2ZgGU+IjEV4nwlhoK5sQluxsYJ78Id3Y=
 github.com/xi2/xz v0.0.0-20171230120015-48954b6210f8 h1:nIPpBwaJSVYIxUFsDv3M8ofmx9yWTog9BfvIu0q41lo=
 github.com/xi2/xz v0.0.0-20171230120015-48954b6210f8/go.mod h1:HUYIGzjTL3rfEspMxjDjgmT5uz5wzYJKVo23qUhYTos=
 github.com/yuin/goldmark v1.1.27/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
@@ -822,10 +815,6 @@ go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.56.0 h1:UP6IpuH
 go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp v0.56.0/go.mod h1:qxuZLtbq5QDtdeSHsS7bcf6EH6uO6jUAgk764zd3rhM=
 go.opentelemetry.io/otel v1.34.0 h1:zRLXxLCgL1WyKsPVrgbSdMN4c0FMkDAskSTQP+0hdUY=
 go.opentelemetry.io/otel v1.34.0/go.mod h1:OWFPOQ+h4G8xpyjgqo4SxJYdDQ/qmRH+wivy7zzx9oI=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.31.0 h1:K0XaT3DwHAcV4nKLzcQvwAgSyisUghWoY20I7huthMk=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.31.0/go.mod h1:B5Ki776z/MBnVha1Nzwp5arlzBbE3+1jk+pGmaP5HME=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.31.0 h1:lUsI2TYsQw2r1IASwoROaCnjdj2cvC2+Jbxvk6nHnWU=
-go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracehttp v1.31.0/go.mod h1:2HpZxxQurfGxJlJDblybejHB6RX6pmExPNe517hREw4=
 go.opentelemetry.io/otel/exporters/prometheus v0.50.0 h1:2Ewsda6hejmbhGFyUvWZjUThC98Cf8Zy6g0zkIimOng=
 go.opentelemetry.io/otel/exporters/prometheus v0.50.0/go.mod h1:pMm5PkUo5YwbLiuEf7t2xg4wbP0/eSJrMxIMxKosynY=
 go.opentelemetry.io/otel/metric v1.34.0 h1:+eTR3U0MyfWjRDhmFMxe2SsW64QrZ84AOhvqS7Y+PoQ=
@@ -836,8 +825,6 @@ go.opentelemetry.io/otel/sdk/metric v1.28.0 h1:OkuaKgKrgAbYrrY0t92c+cC+2F6hsFNnC
 go.opentelemetry.io/otel/sdk/metric v1.28.0/go.mod h1:cWPjykihLAPvXKi4iZc1dpER3Jdq2Z0YLse3moQUCpg=
 go.opentelemetry.io/otel/trace v1.34.0 h1:+ouXS2V8Rd4hp4580a8q23bg0azF2nI8cqLYnC8mh/k=
 go.opentelemetry.io/otel/trace v1.34.0/go.mod h1:Svm7lSjQD7kG7KJ/MUHPVXSDGz2OX4h0M2jHBhmSfRE=
-go.opentelemetry.io/proto/otlp v1.3.1 h1:TrMUixzpM0yuc/znrFTP9MMRh8trP93mkCiDVeXrui0=
-go.opentelemetry.io/proto/otlp v1.3.1/go.mod h1:0X1WI4de4ZsLrrJNLAQbFeLCm3T7yBkR0XqQ7niQU+8=
 go.uber.org/atomic v1.6.0/go.mod h1:sABNBOSYdrvTF6hTgEIbc7YasKWGhgEQZyfxyTvoXHQ=
 go.uber.org/atomic v1.7.0/go.mod h1:fEN4uk6kAWBTFdckzkM89CLk9XfWZrxpCo0nPH17wJc=
 go.uber.org/dig v1.18.0 h1:imUL1UiY0Mg4bqbFfsRQO5G4CGRBec/ZujWTvSVp3pw=
@@ -1107,8 +1094,8 @@ gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C
 gopkg.in/yaml.v3 v3.0.0-20210107192922-496545a6307b/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
 gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
-gotest.tools/v3 v3.5.1 h1:EENdUnS3pdur5nybKYIh2Vfgc8IUNBjxDPSjtiJcOzU=
-gotest.tools/v3 v3.5.1/go.mod h1:isy3WKz7GK6uNw/sbHzfKBLvlvXwUyV06n6brMxxopU=
+gotest.tools/v3 v3.3.0 h1:MfDY1b1/0xN1CyMlQDac0ziEy9zJQd9CXBRRDHw2jJo=
+gotest.tools/v3 v3.3.0/go.mod h1:Mcr9QNxkg0uMvy/YElmo4SpXgJKWgQvYrT7Kw5RzJ1A=
 grpc.go4.org v0.0.0-20170609214715-11d0a25b4919/go.mod h1:77eQGdRu53HpSqPFJFmuJdjuHRquDANNeA4x7B8WQ9o=
 gvisor.dev/gvisor v0.0.0-20230927004350-cbd86285d259 h1:TbRPT0HtzFP3Cno1zZo7yPzEEnfu8EjLfl6IU9VfqkQ=
 gvisor.dev/gvisor v0.0.0-20230927004350-cbd86285d259/go.mod h1:AVgIgHMwK63XvmAzWG9vLQ41YnVHN0du0tEC46fI7yY=

pkg/assets/extract.go

@@ -1,37 +1,35 @@
 package assets
 
 import (
+	"embed"
 	"fmt"
+	"io/fs"
 	"os"
 	"path/filepath"
 
-	rice "github.com/GeertJohan/go.rice"
 	"github.com/mudler/LocalAI/pkg/library"
 )
 
-const backendAssetsDir = "backend-assets"
-
 func ResolvePath(dir string, paths ...string) string {
-	return filepath.Join(append([]string{dir, backendAssetsDir}, paths...)...)
+	return filepath.Join(append([]string{dir, "backend-assets"}, paths...)...)
 }
 
-func ExtractFiles(content *rice.Box, extractDir string) error {
-	// Create the target directory with backend-assets subdirectory
-	backendAssetsDir := filepath.Join(extractDir, backendAssetsDir)
-	err := os.MkdirAll(backendAssetsDir, 0750)
+func ExtractFiles(content embed.FS, extractDir string) error {
+	// Create the target directory if it doesn't exist
+	err := os.MkdirAll(extractDir, 0750)
 	if err != nil {
 		return fmt.Errorf("failed to create directory: %v", err)
 	}
 
-	// Walk through the rice box and extract files
-	err = content.Walk("", func(path string, info os.FileInfo, err error) error {
+	// Walk through the embedded FS and extract files
+	err = fs.WalkDir(content, ".", func(path string, d fs.DirEntry, err error) error {
 		if err != nil {
 			return err
 		}
 
 		// Reconstruct the directory structure in the target directory
-		targetFile := filepath.Join(backendAssetsDir, path)
-		if info.IsDir() {
+		targetFile := filepath.Join(extractDir, path)
+		if d.IsDir() {
 			// Create the directory in the target directory
 			err := os.MkdirAll(targetFile, 0750)
 			if err != nil {
@@ -40,8 +38,8 @@ func ExtractFiles(content *rice.Box, extractDir string) error {
 			return nil
 		}
 
-		// Read the file from the rice box
-		fileData, err := content.Bytes(path)
+		// Read the file from the embedded FS
+		fileData, err := content.ReadFile(path)
 		if err != nil {
 			return fmt.Errorf("failed to read file: %v", err)
 		}
@@ -58,7 +56,7 @@ func ExtractFiles(content *rice.Box, extractDir string) error {
 	// If there is a lib directory, set LD_LIBRARY_PATH to include it
 	// we might use this mechanism to carry over e.g. Nvidia CUDA libraries
 	// from the embedded FS to the target directory
-	library.LoadExtractedLibs(backendAssetsDir)
+	library.LoadExtractedLibs(extractDir)
 
 	return err
 }

pkg/assets/list.go

@@ -1,19 +1,19 @@
 package assets
 
 import (
-	"os"
+	"embed"
+	"io/fs"
 
-	rice "github.com/GeertJohan/go.rice"
 	"github.com/rs/zerolog/log"
 )
 
-func ListFiles(content *rice.Box) (files []string) {
-	err := content.Walk("", func(path string, info os.FileInfo, err error) error {
+func ListFiles(content embed.FS) (files []string) {
+	err := fs.WalkDir(content, ".", func(path string, d fs.DirEntry, err error) error {
 		if err != nil {
 			return err
 		}
 
-		if info.IsDir() {
+		if d.IsDir() {
 			return nil
 		}
 
@@ -21,7 +21,7 @@ func ListFiles(content *rice.Box) (files []string) {
 		return nil
 	})
 	if err != nil {
-		log.Error().Err(err).Msg("error walking the rice box")
+		log.Error().Err(err).Msg("error walking the embedded filesystem")
 	}
 	return
 }

pkg/grpc/backend.go

@@ -39,7 +39,6 @@ type Backend interface {
 	LoadModel(ctx context.Context, in *pb.ModelOptions, opts ...grpc.CallOption) (*pb.Result, error)
 	PredictStream(ctx context.Context, in *pb.PredictOptions, f func(reply *pb.Reply), opts ...grpc.CallOption) error
 	GenerateImage(ctx context.Context, in *pb.GenerateImageRequest, opts ...grpc.CallOption) (*pb.Result, error)
-	GenerateVideo(ctx context.Context, in *pb.GenerateVideoRequest, opts ...grpc.CallOption) (*pb.Result, error)
 	TTS(ctx context.Context, in *pb.TTSRequest, opts ...grpc.CallOption) (*pb.Result, error)
 	SoundGeneration(ctx context.Context, in *pb.SoundGenerationRequest, opts ...grpc.CallOption) (*pb.Result, error)
 	AudioTranscription(ctx context.Context, in *pb.TranscriptRequest, opts ...grpc.CallOption) (*pb.TranscriptResult, error)

pkg/grpc/base/base.go

@@ -53,10 +53,6 @@ func (llm *Base) GenerateImage(*pb.GenerateImageRequest) error {
 	return fmt.Errorf("unimplemented")
 }
 
-func (llm *Base) GenerateVideo(*pb.GenerateVideoRequest) error {
-	return fmt.Errorf("unimplemented")
-}
-
 func (llm *Base) AudioTranscription(*pb.TranscriptRequest) (pb.TranscriptResult, error) {
 	return pb.TranscriptResult{}, fmt.Errorf("unimplemented")
 }

pkg/grpc/client.go

@@ -215,28 +215,6 @@ func (c *Client) GenerateImage(ctx context.Context, in *pb.GenerateImageRequest,
 	return client.GenerateImage(ctx, in, opts...)
 }
 
-func (c *Client) GenerateVideo(ctx context.Context, in *pb.GenerateVideoRequest, opts ...grpc.CallOption) (*pb.Result, error) {
-	if !c.parallel {
-		c.opMutex.Lock()
-		defer c.opMutex.Unlock()
-	}
-	c.setBusy(true)
-	defer c.setBusy(false)
-	c.wdMark()
-	defer c.wdUnMark()
-	conn, err := grpc.Dial(c.address, grpc.WithTransportCredentials(insecure.NewCredentials()),
-		grpc.WithDefaultCallOptions(
-			grpc.MaxCallRecvMsgSize(50*1024*1024), // 50MB
-			grpc.MaxCallSendMsgSize(50*1024*1024), // 50MB
-		))
-	if err != nil {
-		return nil, err
-	}
-	defer conn.Close()
-	client := pb.NewBackendClient(conn)
-	return client.GenerateVideo(ctx, in, opts...)
-}
-
 func (c *Client) TTS(ctx context.Context, in *pb.TTSRequest, opts ...grpc.CallOption) (*pb.Result, error) {
 	if !c.parallel {
 		c.opMutex.Lock()

pkg/grpc/embed.go

@@ -47,10 +47,6 @@ func (e *embedBackend) GenerateImage(ctx context.Context, in *pb.GenerateImageRe
 	return e.s.GenerateImage(ctx, in)
 }
 
-func (e *embedBackend) GenerateVideo(ctx context.Context, in *pb.GenerateVideoRequest, opts ...grpc.CallOption) (*pb.Result, error) {
-	return e.s.GenerateVideo(ctx, in)
-}
-
 func (e *embedBackend) TTS(ctx context.Context, in *pb.TTSRequest, opts ...grpc.CallOption) (*pb.Result, error) {
 	return e.s.TTS(ctx, in)
 }

pkg/grpc/interface.go

@@ -14,7 +14,6 @@ type LLM interface {
 	Load(*pb.ModelOptions) error
 	Embeddings(*pb.PredictOptions) ([]float32, error)
 	GenerateImage(*pb.GenerateImageRequest) error
-	GenerateVideo(*pb.GenerateVideoRequest) error
 	AudioTranscription(*pb.TranscriptRequest) (pb.TranscriptResult, error)
 	TTS(*pb.TTSRequest) error
 	SoundGeneration(*pb.SoundGenerationRequest) error

pkg/grpc/server.go

@@ -75,18 +75,6 @@ func (s *server) GenerateImage(ctx context.Context, in *pb.GenerateImageRequest)
 	return &pb.Result{Message: "Image generated", Success: true}, nil
 }
 
-func (s *server) GenerateVideo(ctx context.Context, in *pb.GenerateVideoRequest) (*pb.Result, error) {
-	if s.llm.Locking() {
-		s.llm.Lock()
-		defer s.llm.Unlock()
-	}
-	err := s.llm.GenerateVideo(in)
-	if err != nil {
-		return &pb.Result{Message: fmt.Sprintf("Error generating video: %s", err.Error()), Success: false}, err
-	}
-	return &pb.Result{Message: "Video generated", Success: true}, nil
-}
-
 func (s *server) TTS(ctx context.Context, in *pb.TTSRequest) (*pb.Result, error) {
 	if s.llm.Locking() {
 		s.llm.Lock()

pkg/library/dynaload.go

@@ -25,7 +25,7 @@ func LoadExtractedLibs(dir string) error {
 	}
 
 	var err error = nil
-	for _, libDir := range []string{filepath.Join(dir, "lib"), filepath.Join(dir, "lib")} {
+	for _, libDir := range []string{filepath.Join(dir, "backend-assets", "lib"), filepath.Join(dir, "lib")} {
 		err = errors.Join(err, LoadExternal(libDir))
 	}
 	return err

pkg/model/initializers.go

@@ -181,6 +181,10 @@ func orderBackends(backends map[string][]string) ([]string, error) {
 // selectGRPCProcessByHostCapabilities selects the GRPC process to start based on system capabilities
 // Note: this is now relevant only for llama.cpp
 func selectGRPCProcessByHostCapabilities(backend, assetDir string, f16 bool) string {
+	foundCUDA := false
+	foundAMDGPU := false
+	foundIntelGPU := false
+	var grpcProcess string
 
 	// Select backend now just for llama.cpp
 	if backend != LLamaCPP {
@@ -194,26 +198,50 @@ func selectGRPCProcessByHostCapabilities(backend, assetDir string, f16 bool) str
 	}
 
 	// Check for GPU-binaries that are shipped with single binary releases
-	gpuBinaries := map[string]string{
-		"nvidia": LLamaCPPCUDA,
-		"amd":    LLamaCPPHipblas,
-		"intel":  LLamaCPPSycl16,
-	}
-
-	if !f16 {
-		gpuBinaries["intel"] = LLamaCPPSycl32
-	}
-
-	for vendor, binary := range gpuBinaries {
-		if xsysinfo.HasGPU(vendor) {
-			p := backendPath(assetDir, binary)
-			if _, err := os.Stat(p); err == nil {
-				log.Info().Msgf("[%s] attempting to load with %s variant (vendor: %s)", backend, binary, vendor)
-				return p
+	gpus, err := xsysinfo.GPUs()
+	if err == nil {
+		for _, gpu := range gpus {
+			if strings.Contains(gpu.String(), "nvidia") {
+				p := backendPath(assetDir, LLamaCPPCUDA)
+				if _, err := os.Stat(p); err == nil {
+					log.Info().Msgf("[%s] attempting to load with CUDA variant", backend)
+					grpcProcess = p
+					foundCUDA = true
+				} else {
+					log.Debug().Msgf("Nvidia GPU device found, no embedded CUDA variant found. You can ignore this message if you are using container with CUDA support")
+				}
+			}
+			if strings.Contains(gpu.String(), "amd") {
+				p := backendPath(assetDir, LLamaCPPHipblas)
+				if _, err := os.Stat(p); err == nil {
+					log.Info().Msgf("[%s] attempting to load with HIPBLAS variant", backend)
+					grpcProcess = p
+					foundAMDGPU = true
+				} else {
+					log.Debug().Msgf("AMD GPU device found, no embedded HIPBLAS variant found. You can ignore this message if you are using container with HIPBLAS support")
+				}
+			}
+			if strings.Contains(gpu.String(), "intel") {
+				backend := LLamaCPPSycl16
+				if !f16 {
+					backend = LLamaCPPSycl32
+				}
+				p := backendPath(assetDir, backend)
+				if _, err := os.Stat(p); err == nil {
+					log.Info().Msgf("[%s] attempting to load with Intel variant", backend)
+					grpcProcess = p
+					foundIntelGPU = true
+				} else {
+					log.Debug().Msgf("Intel GPU device found, no embedded SYCL variant found. You can ignore this message if you are using container with SYCL support")
+				}
 			}
 		}
 	}
 
+	if foundCUDA || foundAMDGPU || foundIntelGPU {
+		return grpcProcess
+	}
+
 	// No GPU found or no specific binaries found, try to load the CPU variant(s)
 
 	// Select a binary based on availability/capability

pkg/templates/multimodal.go

@@ -21,8 +21,7 @@ type MultimodalContent struct {
 	ID int
 }
 
-// https://github.com/ggml-org/llama.cpp/blob/be1d4a13db26750fac702ceb3af88ae4f39dc9f4/tools/mtmd/mtmd.h#L42
-const DefaultMultiModalTemplate = "{{ range .Audio }}[audio-{{.ID}}]{{end}}{{ range .Images }}<__image__>{{end}}{{ range .Video }}[vid-{{.ID}}]{{end}}{{.Text}}"
+const DefaultMultiModalTemplate = "{{ range .Audio }}[audio-{{.ID}}]{{end}}{{ range .Images }}[img-{{.ID}}]{{end}}{{ range .Video }}[vid-{{.ID}}]{{end}}{{.Text}}"
 
 func TemplateMultiModal(templateString string, opts MultiModalOptions, text string) (string, error) {
 	if templateString == "" {

pkg/xsysinfo/gguf.go

@@ -1,52 +0,0 @@
-package xsysinfo
-
-import (
-	"errors"
-
-	gguf "github.com/gpustack/gguf-parser-go"
-)
-
-type VRAMEstimate struct {
-	TotalVRAM       uint64
-	AvailableVRAM   uint64
-	ModelSize       uint64
-	EstimatedLayers int
-	EstimatedVRAM   uint64
-	IsFullOffload   bool
-}
-
-func EstimateGGUFVRAMUsage(f *gguf.GGUFFile, availableVRAM uint64) (*VRAMEstimate, error) {
-	// Get model metadata
-	m := f.Metadata()
-	a := f.Architecture()
-
-	// Calculate base model size
-	modelSize := uint64(m.Size)
-
-	if a.BlockCount == 0 {
-		return nil, errors.New("block count is 0")
-	}
-
-	// Estimate number of layers that can fit in VRAM
-	// Each layer typically requires about 1/32 of the model size
-	layerSize := modelSize / uint64(a.BlockCount)
-	estimatedLayers := int(availableVRAM / layerSize)
-
-	// If we can't fit even one layer, we need to do full offload
-	isFullOffload := estimatedLayers <= 0
-	if isFullOffload {
-		estimatedLayers = 0
-	}
-
-	// Calculate estimated VRAM usage
-	estimatedVRAM := uint64(estimatedLayers) * layerSize
-
-	return &VRAMEstimate{
-		TotalVRAM:       availableVRAM,
-		AvailableVRAM:   availableVRAM,
-		ModelSize:       modelSize,
-		EstimatedLayers: estimatedLayers,
-		EstimatedVRAM:   estimatedVRAM,
-		IsFullOffload:   isFullOffload,
-	}, nil
-}

pkg/xsysinfo/gpu.go

@@ -1,8 +1,6 @@
 package xsysinfo
 
 import (
-	"strings"
-
 	"github.com/jaypipes/ghw"
 	"github.com/jaypipes/ghw/pkg/gpu"
 )
@@ -15,37 +13,3 @@ func GPUs() ([]*gpu.GraphicsCard, error) {
 
 	return gpu.GraphicsCards, nil
 }
-
-func TotalAvailableVRAM() (uint64, error) {
-	gpus, err := GPUs()
-	if err != nil {
-		return 0, err
-	}
-
-	var totalVRAM uint64
-	for _, gpu := range gpus {
-		if gpu != nil && gpu.Node != nil && gpu.Node.Memory != nil {
-			if gpu.Node.Memory.TotalUsableBytes > 0 {
-				totalVRAM += uint64(gpu.Node.Memory.TotalUsableBytes)
-			}
-		}
-	}
-
-	return totalVRAM, nil
-}
-
-func HasGPU(vendor string) bool {
-	gpus, err := GPUs()
-	if err != nil {
-		return false
-	}
-	if vendor == "" {
-		return len(gpus) > 0
-	}
-	for _, gpu := range gpus {
-		if strings.Contains(gpu.String(), vendor) {
-			return true
-		}
-	}
-	return false
-}

swagger/docs.go

@@ -812,30 +812,6 @@ const docTemplate = `{
                     }
                 }
             }
-        },
-        "/video": {
-            "post": {
-                "summary": "Creates a video given a prompt.",
-                "parameters": [
-                    {
-                        "description": "query params",
-                        "name": "request",
-                        "in": "body",
-                        "required": true,
-                        "schema": {
-                            "$ref": "#/definitions/schema.OpenAIRequest"
-                        }
-                    }
-                ],
-                "responses": {
-                    "200": {
-                        "description": "Response",
-                        "schema": {
-                            "$ref": "#/definitions/schema.OpenAIResponse"
-                        }
-                    }
-                }
-            }
         }
     },
     "definitions": {
@@ -1660,6 +1636,7 @@ const docTemplate = `{
                     "type": "string"
                 },
                 "model_base_name": {
+                    "description": "AutoGPTQ",
                     "type": "string"
                 },
                 "n": {
@@ -1743,6 +1720,10 @@ const docTemplate = `{
                 },
                 "typical_p": {
                     "type": "number"
+                },
+                "use_fast_tokenizer": {
+                    "description": "AutoGPTQ",
+                    "type": "boolean"
                 }
             }
         },

swagger/swagger.json

@@ -805,30 +805,6 @@
                     }
                 }
             }
-        },
-        "/video": {
-            "post": {
-                "summary": "Creates a video given a prompt.",
-                "parameters": [
-                    {
-                        "description": "query params",
-                        "name": "request",
-                        "in": "body",
-                        "required": true,
-                        "schema": {
-                            "$ref": "#/definitions/schema.OpenAIRequest"
-                        }
-                    }
-                ],
-                "responses": {
-                    "200": {
-                        "description": "Response",
-                        "schema": {
-                            "$ref": "#/definitions/schema.OpenAIResponse"
-                        }
-                    }
-                }
-            }
         }
     },
     "definitions": {
@@ -1653,6 +1629,7 @@
                     "type": "string"
                 },
                 "model_base_name": {
+                    "description": "AutoGPTQ",
                     "type": "string"
                 },
                 "n": {
@@ -1736,6 +1713,10 @@
                 },
                 "typical_p": {
                     "type": "number"
+                },
+                "use_fast_tokenizer": {
+                    "description": "AutoGPTQ",
+                    "type": "boolean"
                 }
             }
         },

swagger/swagger.yaml

@@ -555,6 +555,7 @@ definitions:
       model:
         type: string
       model_base_name:
+        description: AutoGPTQ
         type: string
       "n":
         description: Also part of the OpenAI official spec. use it for returning multiple
@@ -615,6 +616,9 @@ definitions:
         type: boolean
       typical_p:
         type: number
+      use_fast_tokenizer:
+        description: AutoGPTQ
+        type: boolean
     required:
     - file
     type: object
@@ -1248,21 +1252,6 @@ paths:
           schema:
             $ref: '#/definitions/proto.VADResponse'
       summary: Detect voice fragments in an audio stream
-  /video:
-    post:
-      parameters:
-      - description: query params
-        in: body
-        name: request
-        required: true
-        schema:
-          $ref: '#/definitions/schema.OpenAIRequest'
-      responses:
-        "200":
-          description: Response
-          schema:
-            $ref: '#/definitions/schema.OpenAIResponse'
-      summary: Creates a video given a prompt.
 securityDefinitions:
   BearerAuth:
     in: header

tests/e2e-aio/e2e_suite_test.go

@@ -7,27 +7,24 @@ import (
 	"runtime"
 	"testing"
 
-	"github.com/docker/go-connections/nat"
 	. "github.com/onsi/ginkgo/v2"
 	. "github.com/onsi/gomega"
+	"github.com/ory/dockertest/v3"
+	"github.com/ory/dockertest/v3/docker"
 	"github.com/sashabaranov/go-openai"
-	"github.com/testcontainers/testcontainers-go"
-	"github.com/testcontainers/testcontainers-go/wait"
 )
 
-var container testcontainers.Container
+var pool *dockertest.Pool
+var resource *dockertest.Resource
 var client *openai.Client
 
 var containerImage = os.Getenv("LOCALAI_IMAGE")
 var containerImageTag = os.Getenv("LOCALAI_IMAGE_TAG")
 var modelsDir = os.Getenv("LOCALAI_MODELS_DIR")
+var apiPort = os.Getenv("LOCALAI_API_PORT")
 var apiEndpoint = os.Getenv("LOCALAI_API_ENDPOINT")
 var apiKey = os.Getenv("LOCALAI_API_KEY")
 
-const (
-	defaultApiPort = "8080"
-)
-
 func TestLocalAI(t *testing.T) {
 	RegisterFailHandler(Fail)
 	RunSpecs(t, "LocalAI E2E test suite")
@@ -35,14 +32,15 @@ func TestLocalAI(t *testing.T) {
 
 var _ = BeforeSuite(func() {
 
+	if apiPort == "" {
+		apiPort = "8080"
+	}
+
 	var defaultConfig openai.ClientConfig
 	if apiEndpoint == "" {
 		startDockerImage()
-		apiPort, err := container.MappedPort(context.Background(), nat.Port(defaultApiPort))
-		Expect(err).To(Not(HaveOccurred()))
-
 		defaultConfig = openai.DefaultConfig(apiKey)
-		apiEndpoint = "http://localhost:" + apiPort.Port() + "/v1" // So that other tests can reference this value safely.
+		apiEndpoint = "http://localhost:" + apiPort + "/v1" // So that other tests can reference this value safely.
 		defaultConfig.BaseURL = apiEndpoint
 	} else {
 		GinkgoWriter.Printf("docker apiEndpoint set from env: %q\n", apiEndpoint)
@@ -60,23 +58,26 @@ var _ = BeforeSuite(func() {
 })
 
 var _ = AfterSuite(func() {
-	if container != nil {
-		Expect(container.Terminate(context.Background())).To(Succeed())
+	if resource != nil {
+		Expect(pool.Purge(resource)).To(Succeed())
 	}
+	//dat, err := os.ReadFile(resource.Container.LogPath)
+	//Expect(err).To(Not(HaveOccurred()))
+	//Expect(string(dat)).To(ContainSubstring("GRPC Service Ready"))
+	//fmt.Println(string(dat))
 })
 
 var _ = AfterEach(func() {
-	// Add any cleanup needed after each test
+	//Expect(dbClient.Clear()).To(Succeed())
 })
 
-type logConsumer struct {
-}
-
-func (l *logConsumer) Accept(log testcontainers.Log) {
-	GinkgoWriter.Write([]byte(log.Content))
-}
-
 func startDockerImage() {
+	p, err := dockertest.NewPool("")
+	Expect(err).To(Not(HaveOccurred()))
+	Expect(p.Client.Ping()).To(Succeed())
+
+	pool = p
+
 	// get cwd
 	cwd, err := os.Getwd()
 	Expect(err).To(Not(HaveOccurred()))
@@ -87,43 +88,20 @@ func startDockerImage() {
 	}
 
 	proc := runtime.NumCPU()
-
-	req := testcontainers.ContainerRequest{
-
-		Image:        fmt.Sprintf("%s:%s", containerImage, containerImageTag),
-		ExposedPorts: []string{defaultApiPort},
-		LogConsumerCfg: &testcontainers.LogConsumerConfig{
-			Consumers: []testcontainers.LogConsumer{
-				&logConsumer{},
-			},
+	options := &dockertest.RunOptions{
+		Repository: containerImage,
+		Tag:        containerImageTag,
+		//	Cmd:        []string{"server", "/data"},
+		PortBindings: map[docker.Port][]docker.PortBinding{
+			"8080/tcp": []docker.PortBinding{{HostPort: apiPort}},
 		},
-		Env: map[string]string{
-			"MODELS_PATH":                   "/models",
-			"DEBUG":                         "true",
-			"THREADS":                       fmt.Sprint(proc),
-			"LOCALAI_SINGLE_ACTIVE_BACKEND": "true",
-		},
-		Files: []testcontainers.ContainerFile{
-			{
-				HostFilePath:      md,
-				ContainerFilePath: "/models",
-				FileMode:          0o755,
-			},
-		},
-		WaitingFor: wait.ForAll(
-			wait.ForListeningPort(nat.Port(defaultApiPort)),
-		//	wait.ForHTTP("/v1/models").WithPort(nat.Port(apiPort)).WithStartupTimeout(50*time.Minute),
-		),
+		Env:    []string{"MODELS_PATH=/models", "DEBUG=true", "THREADS=" + fmt.Sprint(proc), "LOCALAI_SINGLE_ACTIVE_BACKEND=true"},
+		Mounts: []string{md + ":/models"},
 	}
 
-	GinkgoWriter.Printf("Launching Docker Container %s:%s\n", containerImage, containerImageTag)
-
-	ctx := context.Background()
-	c, err := testcontainers.GenericContainer(ctx, testcontainers.GenericContainerRequest{
-		ContainerRequest: req,
-		Started:          true,
-	})
+	GinkgoWriter.Printf("Launching Docker Container %q\n%+v\n", containerImageTag, options)
+	r, err := pool.RunWithOptions(options)
 	Expect(err).To(Not(HaveOccurred()))
 
-	container = c
+	resource = r
 }