Update README.md

Signed-off-by: Ettore Di Giacinto <mudler@users.noreply.github.com>

.env

@@ -23,6 +23,12 @@ MODELS_PATH=/models
 ## Enable debug mode
 # DEBUG=true
 
+## Disables COMPEL (Diffusers)
+# COMPEL=0
+
+## Enable/Disable single backend (useful if only one GPU is available)
+# SINGLE_ACTIVE_BACKEND=true
+
 ## Specify a build type. Available: cublas, openblas, clblas.
 ## cuBLAS: This is a GPU-accelerated version of the complete standard BLAS (Basic Linear Algebra Subprograms) library. It's provided by Nvidia and is part of their CUDA toolkit.
 ## OpenBLAS: This is an open-source implementation of the BLAS library that aims to provide highly optimized code for various platforms. It includes support for multi-threading and can be compiled to use hardware-specific features for additional performance. OpenBLAS can run on many kinds of hardware, including CPUs from Intel, AMD, and ARM.
@@ -44,3 +50,40 @@ MODELS_PATH=/models
 
 ## Specify a default upload limit in MB (whisper)
 # UPLOAD_LIMIT
+
+## List of external GRPC backends (note on the container image this variable is already set to use extra backends available in extra/)
+# EXTERNAL_GRPC_BACKENDS=my-backend:127.0.0.1:9000,my-backend2:/usr/bin/backend.py
+
+### Advanced settings ###
+### Those are not really used by LocalAI, but from components in the stack ###
+##
+### Preload libraries
+# LD_PRELOAD=
+
+### Huggingface cache for models
+# HUGGINGFACE_HUB_CACHE=/usr/local/huggingface
+
+### Python backends GRPC max workers
+### Default number of workers for GRPC Python backends.
+### This actually controls wether a backend can process multiple requests or not.
+# PYTHON_GRPC_MAX_WORKERS=1
+
+### Define the number of parallel LLAMA.cpp workers (Defaults to 1)
+# LLAMACPP_PARALLEL=1
+
+### Enable to run parallel requests
+# PARALLEL_REQUESTS=true
+
+### Watchdog settings
+###
+# Enables watchdog to kill backends that are inactive for too much time
+# WATCHDOG_IDLE=true
+#
+# Enables watchdog to kill backends that are busy for too much time
+# WATCHDOG_BUSY=true
+#
+# Time in duration format (e.g. 1h30m) after which a backend is considered idle
+# WATCHDOG_IDLE_TIMEOUT=5m
+#
+# Time in duration format (e.g. 1h30m) after which a backend is considered busy
+# WATCHDOG_BUSY_TIMEOUT=5m

.github/PULL_REQUEST_TEMPLATE.md

@@ -8,16 +8,24 @@ This PR fixes #
 **[Signed commits](../CONTRIBUTING.md#signing-off-on-commits-developer-certificate-of-origin)**
 - [ ] Yes, I signed my commits.
  
-
 <!--
 Thank you for contributing to LocalAI! 
 
-Contributing Conventions:
+Contributing Conventions
+-------------------------
 
-1. Include descriptive PR titles with [<component-name>] prepended.
-2. Build and test your changes before submitting a PR. 
+The draft above helps to give a quick overview of your PR.
+
+Remember to remove this comment and to at least:
+
+1. Include descriptive PR titles with [<component-name>] prepended. We use [conventional commits](https://www.conventionalcommits.org/en/v1.0.0/).
+2. Build and test your changes before submitting a PR (`make build`). 
 3. Sign your commits
+4. **Tag maintainer:** for a quicker response, tag the relevant maintainer (see below).
+5. **X/Twitter handle:** we announce bigger features on X/Twitter. If your PR gets announced, and you'd like a mention, we'll gladly shout you out!
 
 By following the community's contribution conventions upfront, the review process will 
 be accelerated and your PR merged more quickly.
+
+If no one reviews your PR within a few days, please @-mention @mudler.
 -->

.github/workflows/bump_deps.yaml

@@ -12,6 +12,9 @@ jobs:
           - repository: "go-skynet/go-llama.cpp"
             variable: "GOLLAMA_VERSION"
             branch: "master"
+          - repository: "ggerganov/llama.cpp"
+            variable: "CPPLLAMA_VERSION"
+            branch: "master"
           - repository: "go-skynet/go-ggml-transformers.cpp"
             variable: "GOGGMLTRANSFORMERS_VERSION"
             branch: "master"
@@ -41,7 +44,7 @@ jobs:
             branch: "master"
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - name: Bump dependencies 🔧
         run: |
           bash .github/bump_deps.sh ${{ matrix.repository }} ${{ matrix.branch }} ${{ matrix.variable }}

.github/workflows/disabled/test-gpu.yml

@@ -0,0 +1,63 @@
+---
+name: 'GPU tests'
+
+on:
+  pull_request:
+  push:
+    branches:
+      - master
+    tags:
+      - '*'
+
+concurrency:
+  group: ci-gpu-tests-${{ github.head_ref || github.ref }}-${{ github.repository }}
+  cancel-in-progress: true
+
+jobs:
+  ubuntu-latest:
+    runs-on: gpu
+    strategy:
+      matrix:
+        go-version: ['1.21.x']
+    steps:
+      - name: Clone
+        uses: actions/checkout@v4
+        with: 
+          submodules: true
+      - name: Setup Go ${{ matrix.go-version }}
+        uses: actions/setup-go@v4
+        with:
+          go-version: ${{ matrix.go-version }}
+      # You can test your matrix by printing the current Go version
+      - name: Display Go version
+        run: go version
+      - name: Dependencies
+        run: |
+          sudo apt-get update
+          sudo DEBIAN_FRONTEND=noninteractive apt-get install -y make wget
+      - name: Build
+        run: |
+          if [ ! -e /run/systemd/system ]; then
+            sudo mkdir /run/systemd/system
+          fi
+          sudo mkdir -p /host/tests/${{ github.head_ref || github.ref }}
+          sudo chmod -R 777 /host/tests/${{ github.head_ref || github.ref }}
+          make \
+            TEST_DIR="/host/tests/${{ github.head_ref || github.ref }}" \
+            BUILD_TYPE=cublas \
+            prepare-e2e run-e2e-image test-e2e
+      - name: Release space from worker ♻
+        if: always()
+        run: |
+          sudo rm -rf build || true
+          sudo rm -rf bin || true
+          sudo rm -rf dist || true
+          sudo docker logs $(sudo docker ps -q --filter ancestor=localai-tests) > logs.txt
+          sudo cat logs.txt || true
+          sudo rm -rf logs.txt
+          make clean || true
+          make \
+            TEST_DIR="/host/tests/${{ github.head_ref || github.ref }}" \
+            teardown-e2e || true
+          sudo rm -rf /host/tests/${{ github.head_ref || github.ref }} || true
+          docker system prune -f -a --volumes || true

.github/workflows/image.yml

@@ -14,128 +14,144 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  docker:
+  extras-image-build:
+    uses: ./.github/workflows/image_build.yml
+    with:
+      tag-latest: ${{ matrix.tag-latest }}
+      tag-suffix: ${{ matrix.tag-suffix }}
+      ffmpeg: ${{ matrix.ffmpeg }}
+      image-type: ${{ matrix.image-type }}
+      build-type: ${{ matrix.build-type }}
+      cuda-major-version: ${{ matrix.cuda-major-version }}
+      cuda-minor-version: ${{ matrix.cuda-minor-version }}
+      platforms: ${{ matrix.platforms }}
+      runs-on: ${{ matrix.runs-on }}
+    secrets:
+      dockerUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+      dockerPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
     strategy:
+      # Pushing with all jobs in parallel
+      # eats the bandwidth of all the nodes
+      max-parallel: ${{ github.event_name != 'pull_request' && 2 || 4 }}
       matrix:
         include:
           - build-type: ''
-            platforms: 'linux/amd64,linux/arm64'
+            #platforms: 'linux/amd64,linux/arm64'
+            platforms: 'linux/amd64'
             tag-latest: 'auto'
             tag-suffix: ''
             ffmpeg: ''
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
+          - build-type: ''
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-ffmpeg'
+            ffmpeg: 'true'
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
           - build-type: 'cublas'
-            cuda-major-version: 11
-            cuda-minor-version: 7
+            cuda-major-version: "11"
+            cuda-minor-version: "7"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda11'
             ffmpeg: ''
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
           - build-type: 'cublas'
-            cuda-major-version: 12
-            cuda-minor-version: 1
+            cuda-major-version: "12"
+            cuda-minor-version: "1"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12'
             ffmpeg: ''
-          - build-type: ''
-            platforms: 'linux/amd64,linux/arm64'
-            tag-latest: 'false'
-            tag-suffix: '-ffmpeg'
-            ffmpeg: 'true'
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
           - build-type: 'cublas'
-            cuda-major-version: 11
-            cuda-minor-version: 7
+            cuda-major-version: "11"
+            cuda-minor-version: "7"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda11-ffmpeg'
             ffmpeg: 'true'
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
           - build-type: 'cublas'
-            cuda-major-version: 12
-            cuda-minor-version: 1
+            cuda-major-version: "12"
+            cuda-minor-version: "1"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12-ffmpeg'
             ffmpeg: 'true'
-
-    runs-on: ubuntu-latest
-    steps:
-      - name: Release space from worker
-        run: |
-          echo "Listing top largest packages"
-          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
-          head -n 30 <<< "${pkgs}"
-          echo
-          df -h
-          echo
-          sudo apt-get remove -y '^llvm-.*|^libllvm.*' || true
-          sudo apt-get remove --auto-remove android-sdk-platform-tools || true
-          sudo apt-get purge --auto-remove android-sdk-platform-tools || true
-          sudo rm -rf /usr/local/lib/android
-          sudo apt-get remove -y '^dotnet-.*|^aspnetcore-.*' || true
-          sudo rm -rf /usr/share/dotnet
-          sudo apt-get remove -y '^mono-.*' || true
-          sudo apt-get remove -y '^ghc-.*' || true
-          sudo apt-get remove -y '.*jdk.*|.*jre.*' || true
-          sudo apt-get remove -y 'php.*' || true
-          sudo apt-get remove -y hhvm powershell firefox monodoc-manual msbuild || true
-          sudo apt-get remove -y '^google-.*' || true
-          sudo apt-get remove -y azure-cli || true
-          sudo apt-get remove -y '^mongo.*-.*|^postgresql-.*|^mysql-.*|^mssql-.*' || true
-          sudo apt-get remove -y '^gfortran-.*' || true
-          sudo apt-get autoremove -y
-          sudo apt-get clean
-          echo
-          echo "Listing top largest packages"
-          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
-          head -n 30 <<< "${pkgs}"
-          echo
-          sudo rm -rfv build || true
-          df -h
-      - name: Checkout
-        uses: actions/checkout@v3
-
-      - name: Docker meta
-        id: meta
-        uses: docker/metadata-action@v4
-        with:
-          images: quay.io/go-skynet/local-ai
-          tags: |
-            type=ref,event=branch
-            type=semver,pattern={{raw}}
-            type=sha
-          flavor: |
-            latest=${{ matrix.tag-latest }}
-            suffix=${{ matrix.tag-suffix }}
-
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@master
-        with:
-          platforms: all
-
-      - name: Set up Docker Buildx
-        id: buildx
-        uses: docker/setup-buildx-action@master
-
-      - name: Login to DockerHub
-        if: github.event_name != 'pull_request'
-        uses: docker/login-action@v2
-        with:
-          registry: quay.io
-          username: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
-          password: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
-
-      - name: Build and push
-        uses: docker/build-push-action@v4
-        with:
-          builder: ${{ steps.buildx.outputs.name }}
-          build-args: |
-            BUILD_TYPE=${{ matrix.build-type }}
-            CUDA_MAJOR_VERSION=${{ matrix.cuda-major-version }}
-            CUDA_MINOR_VERSION=${{ matrix.cuda-minor-version }}
-            FFMPEG=${{ matrix.ffmpeg }}
-          context: .
-          file: ./Dockerfile
-          platforms: ${{ matrix.platforms }}
-          push: ${{ github.event_name != 'pull_request' }}
-          tags: ${{ steps.meta.outputs.tags }}
-          labels: ${{ steps.meta.outputs.labels }}
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
+          - build-type: ''
+            #platforms: 'linux/amd64,linux/arm64'
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: ''
+            ffmpeg: ''
+            image-type: 'extras'
+            runs-on: 'arc-runner-set'
+  core-image-build:
+    uses: ./.github/workflows/image_build.yml
+    with:
+      tag-latest: ${{ matrix.tag-latest }}
+      tag-suffix: ${{ matrix.tag-suffix }}
+      ffmpeg: ${{ matrix.ffmpeg }}
+      image-type: ${{ matrix.image-type }}
+      build-type: ${{ matrix.build-type }}
+      cuda-major-version: ${{ matrix.cuda-major-version }}
+      cuda-minor-version: ${{ matrix.cuda-minor-version }}
+      platforms: ${{ matrix.platforms }}
+      runs-on: ${{ matrix.runs-on }}
+    secrets:
+      dockerUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+      dockerPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+    strategy:
+      matrix:
+        include:
+          - build-type: ''
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-ffmpeg-core'
+            ffmpeg: 'true'
+            image-type: 'core'
+            runs-on: 'ubuntu-latest'
+          - build-type: 'cublas'
+            cuda-major-version: "11"
+            cuda-minor-version: "7"
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-cublas-cuda11-core'
+            ffmpeg: ''
+            image-type: 'core'
+            runs-on: 'ubuntu-latest'
+          - build-type: 'cublas'
+            cuda-major-version: "12"
+            cuda-minor-version: "1"
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-cublas-cuda12-core'
+            ffmpeg: ''
+            image-type: 'core'
+            runs-on: 'ubuntu-latest'
+          - 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: 'ubuntu-latest'
+          - build-type: 'cublas'
+            cuda-major-version: "12"
+            cuda-minor-version: "1"
+            platforms: 'linux/amd64'
+            tag-latest: 'false'
+            tag-suffix: '-cublas-cuda12-ffmpeg-core'
+            ffmpeg: 'true'
+            image-type: 'core'
+            runs-on: 'ubuntu-latest'

.github/workflows/image_build.yml

@@ -0,0 +1,147 @@
+---
+name: 'build container images (reusable)'
+
+on:
+  workflow_call:
+    inputs:
+      build-type:
+        description: 'Build type'
+        default: ''
+        type: string
+      cuda-major-version:
+        description: 'CUDA major version'
+        default: "11"
+        type: string
+      cuda-minor-version:
+        description: 'CUDA minor version'
+        default: "7"
+        type: string
+      platforms:
+        description: 'Platforms'
+        default: ''
+        type: string
+      tag-latest:
+        description: 'Tag latest'
+        default: ''
+        type: string
+      tag-suffix:
+        description: 'Tag suffix'
+        default: ''
+        type: string
+      ffmpeg:
+        description: 'FFMPEG'
+        default: ''
+        type: string
+      image-type:
+        description: 'Image type'
+        default: ''
+        type: string
+      runs-on:
+        description: 'Runs on'
+        required: true
+        default: ''
+        type: string
+    secrets:
+      dockerUsername:
+        required: true
+      dockerPassword:
+        required: true
+jobs:
+  reusable_image-build:
+    runs-on: ${{ inputs.runs-on }}
+    steps:
+      - name: Force Install GIT latest
+        run: |
+          sudo apt-get update \
+          && sudo apt-get install -y software-properties-common \
+          && sudo apt-get update \
+          && sudo add-apt-repository -y ppa:git-core/ppa \
+          && sudo apt-get update \
+          && sudo apt-get install -y git
+      - name: Checkout
+        uses: actions/checkout@v4
+      # - name: Release space from worker
+      #   run: |
+      #     echo "Listing top largest packages"
+      #     pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+      #     head -n 30 <<< "${pkgs}"
+      #     echo
+      #     df -h
+      #     echo
+      #     sudo apt-get remove -y '^llvm-.*|^libllvm.*' || true
+      #     sudo apt-get remove --auto-remove android-sdk-platform-tools || true
+      #     sudo apt-get purge --auto-remove android-sdk-platform-tools || true
+      #     sudo rm -rf /usr/local/lib/android
+      #     sudo apt-get remove -y '^dotnet-.*|^aspnetcore-.*' || true
+      #     sudo rm -rf /usr/share/dotnet
+      #     sudo apt-get remove -y '^mono-.*' || true
+      #     sudo apt-get remove -y '^ghc-.*' || true
+      #     sudo apt-get remove -y '.*jdk.*|.*jre.*' || true
+      #     sudo apt-get remove -y 'php.*' || true
+      #     sudo apt-get remove -y hhvm powershell firefox monodoc-manual msbuild || true
+      #     sudo apt-get remove -y '^google-.*' || true
+      #     sudo apt-get remove -y azure-cli || true
+      #     sudo apt-get remove -y '^mongo.*-.*|^postgresql-.*|^mysql-.*|^mssql-.*' || true
+      #     sudo apt-get remove -y '^gfortran-.*' || true
+      #     sudo apt-get remove -y microsoft-edge-stable || true
+      #     sudo apt-get remove -y firefox || true
+      #     sudo apt-get remove -y powershell || true
+      #     sudo apt-get remove -y r-base-core || true
+      #     sudo apt-get autoremove -y
+      #     sudo apt-get clean
+      #     echo
+      #     echo "Listing top largest packages"
+      #     pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+      #     head -n 30 <<< "${pkgs}"
+      #     echo
+      #     sudo rm -rfv build || true
+      #     df -h
+      - name: Docker meta
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: quay.io/go-skynet/local-ai
+          tags: |
+            type=ref,event=branch
+            type=semver,pattern={{raw}}
+            type=sha
+          flavor: |
+            latest=${{ inputs.tag-latest }}
+            suffix=${{ inputs.tag-suffix }}
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@master
+        with:
+          platforms: all
+
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@master
+
+      - name: Login to DockerHub
+        if: github.event_name != 'pull_request'
+        uses: docker/login-action@v3
+        with:
+          registry: quay.io
+          username: ${{ secrets.dockerUsername }}
+          password: ${{ secrets.dockerPassword }}
+
+      - name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          builder: ${{ steps.buildx.outputs.name }}
+          build-args: |
+            BUILD_TYPE=${{ inputs.build-type }}
+            CUDA_MAJOR_VERSION=${{ inputs.cuda-major-version }}
+            CUDA_MINOR_VERSION=${{ inputs.cuda-minor-version }}
+            FFMPEG=${{ inputs.ffmpeg }}
+            IMAGE_TYPE=${{ inputs.image-type }}
+          context: .
+          file: ./Dockerfile
+          platforms: ${{ inputs.platforms }}
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+      - name: job summary
+        run: |
+          echo "Built image: ${{ steps.meta.outputs.labels }}" >> $GITHUB_STEP_SUMMARY

.github/workflows/release.yaml

@@ -19,7 +19,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Clone
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           submodules: true
       - uses: actions/setup-go@v4
@@ -29,6 +29,12 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install build-essential ffmpeg
+
+          git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+              cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
+                -DgRPC_BUILD_TESTS=OFF \
+                ../.. && sudo make -j12 install
+
       - name: Build
         id: build
         env:
@@ -60,18 +66,26 @@ jobs:
     runs-on: macOS-latest
     steps:
       - name: Clone
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           submodules: true
       - uses: actions/setup-go@v4
         with:
           go-version: '>=1.21.0'
+      - name: Dependencies
+        run: |
+          git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+              cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
+                -DgRPC_BUILD_TESTS=OFF \
+                ../.. && make -j12 install && rm -rf grpc
       - name: Build
         id: build
         env:
           CMAKE_ARGS: "${{ matrix.defines }}"
           BUILD_ID: "${{ matrix.build }}"
         run: |
+          export C_INCLUDE_PATH=/usr/local/include
+          export CPLUS_INCLUDE_PATH=/usr/local/include
           make dist
       - uses: actions/upload-artifact@v3
         with:

.github/workflows/test.yml

@@ -14,14 +14,46 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  ubuntu-latest:
+  tests-linux:
     runs-on: ubuntu-latest
     strategy:
       matrix:
         go-version: ['1.21.x']
     steps:
+      - name: Release space from worker
+        run: |
+          echo "Listing top largest packages"
+          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+          head -n 30 <<< "${pkgs}"
+          echo
+          df -h
+          echo
+          sudo apt-get remove -y '^llvm-.*|^libllvm.*' || true
+          sudo apt-get remove --auto-remove android-sdk-platform-tools || true
+          sudo apt-get purge --auto-remove android-sdk-platform-tools || true
+          sudo rm -rf /usr/local/lib/android
+          sudo apt-get remove -y '^dotnet-.*|^aspnetcore-.*' || true
+          sudo rm -rf /usr/share/dotnet
+          sudo apt-get remove -y '^mono-.*' || true
+          sudo apt-get remove -y '^ghc-.*' || true
+          sudo apt-get remove -y '.*jdk.*|.*jre.*' || true
+          sudo apt-get remove -y 'php.*' || true
+          sudo apt-get remove -y hhvm powershell firefox monodoc-manual msbuild || true
+          sudo apt-get remove -y '^google-.*' || true
+          sudo apt-get remove -y azure-cli || true
+          sudo apt-get remove -y '^mongo.*-.*|^postgresql-.*|^mysql-.*|^mssql-.*' || true
+          sudo apt-get remove -y '^gfortran-.*' || true
+          sudo apt-get autoremove -y
+          sudo apt-get clean
+          echo
+          echo "Listing top largest packages"
+          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+          head -n 30 <<< "${pkgs}"
+          echo
+          sudo rm -rfv build || true
+          df -h
       - name: Clone
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with: 
           submodules: true
       - name: Setup Go ${{ matrix.go-version }}
@@ -35,38 +67,42 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install build-essential ffmpeg
-          
+          curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
+             sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
+              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
+             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
+             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
+             sudo apt-get update && \
+             sudo apt-get install -y conda
           sudo apt-get install -y ca-certificates cmake curl patch
           sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
-          sudo pip install -r extra/requirements.txt
+          
+          sudo rm -rfv /usr/bin/conda || true
+          PATH=$PATH:/opt/conda/bin make -C backend/python/sentencetransformers
 
-          sudo mkdir /build && sudo chmod -R 777 /build && cd /build && \
-          curl -L "https://github.com/gabime/spdlog/archive/refs/tags/v1.11.0.tar.gz" | \
-          tar -xzvf - && \
-          mkdir -p "spdlog-1.11.0/build" && \
-          cd "spdlog-1.11.0/build" && \
-          cmake ..  && \
-          make -j8 && \
-          sudo cmake --install . --prefix /usr && mkdir -p "lib/Linux-$(uname -m)" && \
-          cd /build && \
-          mkdir -p "lib/Linux-$(uname -m)/piper_phonemize" && \
-          curl -L "https://github.com/rhasspy/piper-phonemize/releases/download/v1.0.0/libpiper_phonemize-amd64.tar.gz" | \
-          tar -C "lib/Linux-$(uname -m)/piper_phonemize" -xzvf - && ls -liah /build/lib/Linux-$(uname -m)/piper_phonemize/ && \
-          sudo cp -rfv /build/lib/Linux-$(uname -m)/piper_phonemize/lib/. /usr/lib/ && \
-          sudo ln -s /usr/lib/libpiper_phonemize.so /usr/lib/libpiper_phonemize.so.1 && \
-          sudo cp -rfv /build/lib/Linux-$(uname -m)/piper_phonemize/include/. /usr/include/
+          # Pre-build piper before we start tests in order to have shared libraries in place
+          make sources/go-piper && \
+          GO_TAGS="tts" make -C sources/go-piper piper.o && \
+          sudo cp -rfv sources/go-piper/piper/build/pi/lib/. /usr/lib/ && \
+          # Pre-build stable diffusion before we install a newer version of abseil (not compatible with stablediffusion-ncn)
+          GO_TAGS="stablediffusion tts" GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build
+
+          git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+              cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
+                -DgRPC_BUILD_TESTS=OFF \
+                ../.. && sudo make -j12 install
       - name: Test
         run: |
-          ESPEAK_DATA="/build/lib/Linux-$(uname -m)/piper_phonemize/lib/espeak-ng-data" GO_TAGS="tts stablediffusion" make test
+          GO_TAGS="stablediffusion tts" make test
 
-  macOS-latest:
+  tests-apple:
     runs-on: macOS-latest
     strategy:
       matrix:
         go-version: ['1.21.x']
     steps:
       - name: Clone
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with: 
           submodules: true
       - name: Setup Go ${{ matrix.go-version }}
@@ -76,6 +112,14 @@ jobs:
       # You can test your matrix by printing the current Go version
       - name: Display Go version
         run: go version
+      - name: Dependencies
+        run: |
+          git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+              cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
+                -DgRPC_BUILD_TESTS=OFF \
+                ../.. && make -j12 install && rm -rf grpc
       - name: Test
         run: |
+          export C_INCLUDE_PATH=/usr/local/include
+          export CPLUS_INCLUDE_PATH=/usr/local/include
           CMAKE_ARGS="-DLLAMA_F16C=OFF -DLLAMA_AVX512=OFF -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF" make test

.gitignore

@@ -1,15 +1,10 @@
 # go-llama build artifacts
-go-llama
-go-llama-stable
-/gpt4all
-go-stable-diffusion
-go-piper
-/go-bert
-go-ggllm
-/piper
+/sources/
 __pycache__/
 *.a
 get-sources
+/backend/cpp/llama/grpc-server
+/backend/cpp/llama/llama.cpp
 
 go-ggml-transformers
 go-gpt2

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "docs/themes/hugo-theme-relearn"]
+	path = docs/themes/hugo-theme-relearn
+	url = https://github.com/McShelby/hugo-theme-relearn.git

CONTRIBUTING.md

@@ -0,0 +1,72 @@
+# Contributing to localAI
+
+Thank you for your interest in contributing to LocalAI! We appreciate your time and effort in helping to improve our project. Before you get started, please take a moment to review these guidelines.
+
+## Table of Contents
+
+- [Getting Started](#getting-started)
+  - [Prerequisites](#prerequisites)
+  - [Setting up the Development Environment](#setting-up-the-development-environment)
+- [Contributing](#contributing)
+  - [Submitting an Issue](#submitting-an-issue)
+  - [Creating a Pull Request (PR)](#creating-a-pull-request-pr)
+- [Coding Guidelines](#coding-guidelines)
+- [Testing](#testing)
+- [Documentation](#documentation)
+- [Community and Communication](#community-and-communication)
+
+
+
+## Getting Started
+
+### Prerequisites
+
+- Golang [1.21]
+- Git
+- macOS/Linux
+
+### Setting up the Development Environment and running localAI in the local environment
+
+1. Clone the repository: `git clone https://github.com/go-skynet/LocalAI.git`
+2. Navigate to the project directory: `cd LocalAI`
+3. Install the required dependencies: `make prepare`
+4. Run LocalAI: `make run`
+
+## Contributing
+
+We welcome contributions from everyone! To get started, follow these steps:
+
+### Submitting an Issue
+
+If you find a bug, have a feature request, or encounter any issues, please check the [issue tracker](https://github.com/go-skynet/LocalAI/issues) to see if a similar issue has already been reported. If not, feel free to [create a new issue](https://github.com/go-skynet/LocalAI/issues/new) and provide as much detail as possible.
+
+### Creating a Pull Request (PR)
+
+1. Fork the repository.
+2. Create a new branch with a descriptive name: `git checkout -b [branch name]`
+3. Make your changes and commit them.
+4. Push the changes to your fork: `git push origin [branch name]`
+5. Create a new pull request from your branch to the main project's `main` or `master` branch.
+6. Provide a clear description of your changes in the pull request.
+7. Make any requested changes during the review process.
+8. Once your PR is approved, it will be merged into the main project.
+
+## Coding Guidelines
+
+- No specific coding guidelines at the moment. Please make sure the code can be tested. The most popular lint tools like []`golangci-lint`](https://golangci-lint.run) can help you here.
+
+## Testing
+
+`make test` cannot handle all the model now. Please be sure to add a test case for the new features or the part was changed.
+
+## Documentation
+
+- We are welcome the contribution of the documents, please open new PR in the official document repo [localai-website](https://github.com/go-skynet/localai-website)
+
+## Community and Communication
+
+- You can reach out via the Github issue tracker.
+- Open a new discussion at [Discussion](https://github.com/go-skynet/LocalAI/discussions)
+- Join the Discord channel [Discord](https://discord.gg/uJAeKSAGDy)
+
+---

Dockerfile

@@ -1,22 +1,27 @@
 ARG GO_VERSION=1.21-bullseye
+ARG IMAGE_TYPE=extras
+# extras or core
 
-FROM golang:$GO_VERSION as requirements
+
+FROM golang:$GO_VERSION as requirements-core
 
 ARG BUILD_TYPE
 ARG CUDA_MAJOR_VERSION=11
 ARG CUDA_MINOR_VERSION=7
-ARG SPDLOG_VERSION="1.11.0"
-ARG PIPER_PHONEMIZE_VERSION='1.0.0'
 ARG TARGETARCH
 ARG TARGETVARIANT
 
 ENV BUILD_TYPE=${BUILD_TYPE}
-ENV EXTERNAL_GRPC_BACKENDS="huggingface-embeddings:/build/extra/grpc/huggingface/huggingface.py,autogptq:/build/extra/grpc/autogptq/autogptq.py,bark:/build/extra/grpc/bark/ttsbark.py,diffusers:/build/extra/grpc/diffusers/backend_diffusers.py,exllama:/build/extra/grpc/exllama/exllama.py"
+ENV EXTERNAL_GRPC_BACKENDS="huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,petals:/build/backend/python/petals/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,exllama:/build/backend/python/exllama/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh"
 ENV GALLERIES='[{"name":"model-gallery", "url":"github:go-skynet/model-gallery/index.yaml"}, {"url": "github:go-skynet/model-gallery/huggingface.yaml","name":"huggingface"}]'
 ARG GO_TAGS="stablediffusion tts"
 
 RUN apt-get update && \
-    apt-get install -y ca-certificates cmake curl patch pip
+    apt-get install -y ca-certificates curl patch pip cmake && apt-get clean
+
+
+COPY --chmod=644 custom-ca-certs/* /usr/local/share/ca-certificates/
+RUN update-ca-certificates
 
 # Use the variables in subsequent instructions
 RUN echo "Target Architecture: $TARGETARCH"
@@ -30,63 +35,52 @@ RUN if [ "${BUILD_TYPE}" = "cublas" ]; then \
     dpkg -i cuda-keyring_1.0-1_all.deb && \
     rm -f cuda-keyring_1.0-1_all.deb && \
     apt-get update && \
-    apt-get install -y cuda-nvcc-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcublas-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcusparse-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcusolver-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+    apt-get install -y cuda-nvcc-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcublas-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcusparse-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcusolver-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION}  && apt-get clean \
     ; fi
 ENV PATH /usr/local/cuda/bin:${PATH}
 
-# Extras requirements
-COPY extra/requirements.txt /build/extra/requirements.txt
-ENV PATH="/root/.cargo/bin:${PATH}"
-RUN pip install --upgrade pip
-RUN curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
-RUN if [ "${TARGETARCH}" = "amd64" ]; then \
-        pip install git+https://github.com/suno-ai/bark.git diffusers invisible_watermark transformers accelerate safetensors;\
-    fi
-RUN if [ "${BUILD_TYPE}" = "cublas" ] && [ "${TARGETARCH}" = "amd64" ]; then \
-        pip install torch && pip install auto-gptq https://github.com/jllllll/exllama/releases/download/0.0.10/exllama-0.0.10+cu${CUDA_MAJOR_VERSION}${CUDA_MINOR_VERSION}-cp39-cp39-linux_x86_64.whl;\
-    fi
-RUN pip install -r /build/extra/requirements.txt && rm -rf /build/extra/requirements.txt
+# OpenBLAS requirements and stable diffusion
+RUN apt-get install -y \
+    libopenblas-dev \
+    libopencv-dev \ 
+    && apt-get clean
+
+# Set up OpenCV
+RUN ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
 
 WORKDIR /build
 
-# OpenBLAS requirements
-RUN apt-get install -y libopenblas-dev
-
-# Stable Diffusion requirements
-RUN apt-get install -y libopencv-dev && \
-    ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
-
-
-# piper requirements
-# Use pre-compiled Piper phonemization library (includes onnxruntime)
-#RUN if echo "${GO_TAGS}" | grep -q "tts"; then \
 RUN test -n "$TARGETARCH" \
     || (echo 'warn: missing $TARGETARCH, either set this `ARG` manually, or run using `docker buildkit`')
 
-RUN curl -L "https://github.com/gabime/spdlog/archive/refs/tags/v${SPDLOG_VERSION}.tar.gz" | \
-    tar -xzvf - && \
-    mkdir -p "spdlog-${SPDLOG_VERSION}/build" && \
-    cd "spdlog-${SPDLOG_VERSION}/build" && \
-    cmake ..  && \
-    make -j8 && \
-    cmake --install . --prefix /usr && mkdir -p "lib/Linux-$(uname -m)" && \
-    cd /build && \
-    mkdir -p "lib/Linux-$(uname -m)/piper_phonemize" && \
-    curl -L "https://github.com/rhasspy/piper-phonemize/releases/download/v${PIPER_PHONEMIZE_VERSION}/libpiper_phonemize-${TARGETARCH:-$(go env GOARCH)}${TARGETVARIANT}.tar.gz" | \
-    tar -C "lib/Linux-$(uname -m)/piper_phonemize" -xzvf - && ls -liah /build/lib/Linux-$(uname -m)/piper_phonemize/ && \
-    cp -rfv /build/lib/Linux-$(uname -m)/piper_phonemize/lib/. /usr/lib/ && \
-    ln -s /usr/lib/libpiper_phonemize.so /usr/lib/libpiper_phonemize.so.1 && \
-    cp -rfv /build/lib/Linux-$(uname -m)/piper_phonemize/include/. /usr/include/
+# Extras requirements
+FROM requirements-core as requirements-extras
+
+RUN curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
+    install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
+    gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
+    echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list && \
+    echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list && \
+    apt-get update && \
+    apt-get install -y conda
+
+ENV PATH="/root/.cargo/bin:${PATH}"
+RUN pip install --upgrade pip
+RUN curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
+
+
 # \
 #    ; fi
 
 ###################################
 ###################################
 
-FROM requirements as builder
+FROM requirements-${IMAGE_TYPE} as builder
 
 ARG GO_TAGS="stablediffusion tts"
-
+ARG GRPC_BACKENDS
+ARG BUILD_GRPC=true
+ENV GRPC_BACKENDS=${GRPC_BACKENDS}
 ENV GO_TAGS=${GO_TAGS}
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
 ENV NVIDIA_REQUIRE_CUDA="cuda>=${CUDA_MAJOR_VERSION}.0"
@@ -94,28 +88,47 @@ ENV NVIDIA_VISIBLE_DEVICES=all
 
 WORKDIR /build
 
-COPY Makefile .
-RUN make get-sources
-COPY go.mod .
-RUN make prepare
 COPY . .
 COPY .git .
+RUN make prepare
 
-RUN ESPEAK_DATA=/build/lib/Linux-$(uname -m)/piper_phonemize/lib/espeak-ng-data make build
+# stablediffusion does not tolerate a newer version of abseil, build it first
+RUN GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build
+
+RUN if [ "${BUILD_GRPC}" = "true" ]; then \
+    git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+    cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
+      -DgRPC_BUILD_TESTS=OFF \
+       ../.. && make -j12 install \
+    ; fi
+
+# Rebuild with defaults backends
+RUN make build
+
+RUN if [ ! -d "/build/sources/go-piper/piper/build/pi/lib/" ]; then \
+    mkdir -p /build/sources/go-piper/piper/build/pi/lib/ \
+    touch /build/sources/go-piper/piper/build/pi/lib/keep \
+    ; fi
 
 ###################################
 ###################################
 
-FROM requirements
+FROM requirements-${IMAGE_TYPE}
 
 ARG FFMPEG
 ARG BUILD_TYPE
 ARG TARGETARCH
+ARG IMAGE_TYPE=extras
 
 ENV BUILD_TYPE=${BUILD_TYPE}
 ENV REBUILD=false
 ENV HEALTHCHECK_ENDPOINT=http://localhost:8080/readyz
 
+ARG CUDA_MAJOR_VERSION=11
+ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
+ENV NVIDIA_REQUIRE_CUDA="cuda>=${CUDA_MAJOR_VERSION}.0"
+ENV NVIDIA_VISIBLE_DEVICES=all
+
 # Add FFmpeg
 RUN if [ "${FFMPEG}" = "true" ]; then \
     apt-get install -y ffmpeg \
@@ -128,12 +141,62 @@ WORKDIR /build
 # see https://github.com/go-skynet/LocalAI/pull/658#discussion_r1241971626 and
 # https://github.com/go-skynet/LocalAI/pull/434
 COPY . .
-RUN make prepare-sources
+
+COPY --from=builder /build/sources ./sources/
+COPY --from=builder /build/grpc ./grpc/
+
+RUN make prepare-sources && cd /build/grpc/cmake/build && make install && rm -rf grpc
+
+# Copy the binary
 COPY --from=builder /build/local-ai ./
-# To resolve exllama import error
-RUN if [ "${BUILD_TYPE}" = "cublas" ] && [ "${TARGETARCH:-$(go env GOARCH)}" = "amd64" ]; then \
-        cp -rfv /usr/local/lib/python3.9/dist-packages/exllama extra/grpc/exllama/;\
+
+# Copy shared libraries for piper
+COPY --from=builder /build/sources/go-piper/piper/build/pi/lib/* /usr/lib/
+
+# do not let stablediffusion rebuild (requires an older version of absl)
+COPY --from=builder /build/backend-assets/grpc/stablediffusion ./backend-assets/grpc/stablediffusion
+
+## Duplicated from Makefile to avoid having a big layer that's hard to push
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/autogptq \
+    ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/bark \
+    ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/diffusers \
+    ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/vllm \
+    ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/sentencetransformers \
+    ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/transformers \
+    ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/vall-e-x \
+    ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/exllama \
+    ; fi
+RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
+	PATH=$PATH:/opt/conda/bin make -C backend/python/petals \
+    ; fi
+
+# Copy VALLE-X as it's not a real "lib"
+# TODO: this is wrong - we should copy the lib into the conda env path
+RUN if [ -d /usr/lib/vall-e-x ]; then \
+    cp -rfv /usr/lib/vall-e-x/* ./ ; \ 
     fi
+
+# we also copy exllama libs over to resolve exllama import error
+# TODO: check if this is still needed
+RUN if [ -d /usr/local/lib/python3.9/dist-packages/exllama ]; then \
+        cp -rfv /usr/local/lib/python3.9/dist-packages/exllama backend/python/exllama/;\
+    fi
+
 # Define the health check command
 HEALTHCHECK --interval=1m --timeout=10m --retries=10 \
   CMD curl -f $HEALTHCHECK_ENDPOINT || exit 1

Entitlements.plist

@@ -0,0 +1,10 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+    <key>com.apple.security.network.client</key>
+    <true/>
+    <key>com.apple.security.network.server</key>
+    <true/>
+</dict>
+</plist>

Makefile

@@ -4,10 +4,12 @@ GOVET=$(GOCMD) vet
 BINARY_NAME=local-ai
 
 # llama.cpp versions
-GOLLAMA_VERSION?=bf63302a2be787674e6ca4227a8aaeb95a8eb6b1
+GOLLAMA_VERSION?=aeba71ee842819da681ea537e78846dc75949ac0
 
 GOLLAMA_STABLE_VERSION?=50cee7712066d9e38306eccadcfbb44ea87df4b7
 
+CPPLLAMA_VERSION?=1f5cd83275fabb43f2ae92c30033b384a3eb37b4
+
 # gpt4all version
 GPT4ALL_REPO?=https://github.com/nomic-ai/gpt4all
 GPT4ALL_VERSION?=27a8b020c36b0df8f8b82a252d261cda47cf44b8
@@ -26,23 +28,23 @@ WHISPER_CPP_VERSION?=85ed71aaec8e0612a84c0b67804bde75aa75a273
 BERT_VERSION?=6abe312cded14042f6b7c3cd8edf082713334a4d
 
 # go-piper version
-PIPER_VERSION?=56b8a81b4760a6fbee1a82e62f007ae7e8f010a7
-
-# go-bloomz version
-BLOOMZ_VERSION?=1834e77b83faafe912ad4092ccf7f77937349e2f
+PIPER_VERSION?=5a4c9e28c84bac09ab6baa9f88457d852cb46bb2
 
 # stablediffusion version
-STABLEDIFFUSION_VERSION?=d89260f598afb809279bc72aa0107b4292587632
-
-# Go-ggllm
-GOGGLLM_VERSION?=862477d16eefb0805261c19c9b0d053e3b2b684b
+STABLEDIFFUSION_VERSION?=902db5f066fd137697e3b69d0fa10d4782bd2c2f
 
 export BUILD_TYPE?=
+export STABLE_BUILD_TYPE?=$(BUILD_TYPE)
+export CMAKE_ARGS?=
 CGO_LDFLAGS?=
 CUDA_LIBPATH?=/usr/local/cuda/lib64/
 GO_TAGS?=
 BUILD_ID?=git
 
+TEST_DIR=/tmp/test
+
+RANDOM := $(shell bash -c 'echo $$RANDOM')
+
 VERSION?=$(shell git describe --always --tags || echo "dev" )
 # go tool nm ./local-ai | grep Commit
 LD_FLAGS?=
@@ -50,7 +52,6 @@ override LD_FLAGS += -X "github.com/go-skynet/LocalAI/internal.Version=$(VERSION
 override LD_FLAGS += -X "github.com/go-skynet/LocalAI/internal.Commit=$(shell git rev-parse HEAD)"
 
 OPTIONAL_TARGETS?=
-ESPEAK_DATA?=
 
 OS := $(shell uname -s)
 ARCH := $(shell uname -m)
@@ -60,13 +61,26 @@ WHITE  := $(shell tput -Txterm setaf 7)
 CYAN   := $(shell tput -Txterm setaf 6)
 RESET  := $(shell tput -Txterm sgr0)
 
+# Default Docker bridge IP
+E2E_BRIDGE_IP?=172.17.0.1
+
 ifndef UNAME_S
 UNAME_S := $(shell uname -s)
 endif
 
-# workaround for rwkv.cpp
-ifeq ($(UNAME_S),Darwin)
-        CGO_LDFLAGS += -lcblas -framework Accelerate 
+ifeq ($(OS),Darwin)
+	CGO_LDFLAGS += -lcblas -framework Accelerate
+	ifeq ($(OSX_SIGNING_IDENTITY),)
+		OSX_SIGNING_IDENTITY := $(shell security find-identity -v -p codesigning | grep '"' | head -n 1 | sed -E 's/.*"(.*)"/\1/')
+	endif
+
+	# on OSX, if BUILD_TYPE is blank, we should default to use Metal
+	ifeq ($(BUILD_TYPE),)
+		BUILD_TYPE=metal
+	# disable metal if on Darwin and any other value is explicitly passed.
+	else ifneq ($(BUILD_TYPE),metal)
+		CMAKE_ARGS+=-DLLAMA_METAL=OFF
+	endif
 endif
 
 ifeq ($(BUILD_TYPE),openblas)
@@ -78,6 +92,18 @@ ifeq ($(BUILD_TYPE),cublas)
 	export LLAMA_CUBLAS=1
 endif
 
+ifeq ($(BUILD_TYPE),hipblas)
+	ROCM_HOME ?= /opt/rocm
+	export CXX=$(ROCM_HOME)/llvm/bin/clang++
+	export CC=$(ROCM_HOME)/llvm/bin/clang
+	# llama-ggml has no hipblas support, so override it here.
+	export STABLE_BUILD_TYPE=
+	GPU_TARGETS ?= gfx900,gfx90a,gfx1030,gfx1031,gfx1100
+	AMDGPU_TARGETS ?= "$(GPU_TARGETS)"
+	CMAKE_ARGS+=-DLLAMA_HIPBLAS=ON -DAMDGPU_TARGETS="$(AMDGPU_TARGETS)" -DGPU_TARGETS="$(GPU_TARGETS)"
+	CGO_LDFLAGS += -O3 --rtlib=compiler-rt -unwindlib=libgcc -lhipblas -lrocblas --hip-link
+endif
+
 ifeq ($(BUILD_TYPE),metal)
 	CGO_LDFLAGS+=-framework Foundation -framework Metal -framework MetalKit -framework MetalPerformanceShaders
 	export LLAMA_METAL=1
@@ -100,169 +126,149 @@ endif
 ifeq ($(findstring tts,$(GO_TAGS)),tts)
 #	OPTIONAL_TARGETS+=go-piper/libpiper_binding.a
 #	OPTIONAL_TARGETS+=backend-assets/espeak-ng-data
+	PIPER_CGO_CXXFLAGS+=-I$(shell pwd)/sources/go-piper/piper/src/cpp -I$(shell pwd)/sources/go-piper/piper/build/fi/include -I$(shell pwd)/sources/go-piper/piper/build/pi/include -I$(shell pwd)/sources/go-piper/piper/build/si/include
+ 	PIPER_CGO_LDFLAGS+=-L$(shell pwd)/sources/go-piper/piper/build/fi/lib -L$(shell pwd)/sources/go-piper/piper/build/pi/lib -L$(shell pwd)/sources/go-piper/piper/build/si/lib -lfmt -lspdlog -lucd
 	OPTIONAL_GRPC+=backend-assets/grpc/piper
 endif
 
-GRPC_BACKENDS?=backend-assets/grpc/langchain-huggingface backend-assets/grpc/falcon-ggml backend-assets/grpc/bert-embeddings backend-assets/grpc/falcon backend-assets/grpc/bloomz backend-assets/grpc/llama backend-assets/grpc/llama-stable backend-assets/grpc/gpt4all backend-assets/grpc/dolly backend-assets/grpc/gpt2 backend-assets/grpc/gptj backend-assets/grpc/gptneox backend-assets/grpc/mpt backend-assets/grpc/replit backend-assets/grpc/starcoder backend-assets/grpc/rwkv backend-assets/grpc/whisper $(OPTIONAL_GRPC)
+ALL_GRPC_BACKENDS=backend-assets/grpc/langchain-huggingface backend-assets/grpc/falcon-ggml backend-assets/grpc/bert-embeddings backend-assets/grpc/llama backend-assets/grpc/llama-cpp backend-assets/grpc/llama-ggml backend-assets/grpc/gpt4all backend-assets/grpc/dolly backend-assets/grpc/gpt2 backend-assets/grpc/gptj backend-assets/grpc/gptneox backend-assets/grpc/mpt backend-assets/grpc/replit backend-assets/grpc/starcoder backend-assets/grpc/rwkv backend-assets/grpc/whisper $(OPTIONAL_GRPC)
+GRPC_BACKENDS?=$(ALL_GRPC_BACKENDS) $(OPTIONAL_GRPC)
+
+# If empty, then we build all
+ifeq ($(GRPC_BACKENDS),)
+	GRPC_BACKENDS=$(ALL_GRPC_BACKENDS)
+endif
 
 .PHONY: all test build vendor
 
 all: help
 
 ## GPT4ALL
-gpt4all:
-	git clone --recurse-submodules $(GPT4ALL_REPO) gpt4all
-	cd gpt4all && git checkout -b build $(GPT4ALL_VERSION) && git submodule update --init --recursive --depth 1
-
-## go-ggllm
-go-ggllm:
-	git clone --recurse-submodules https://github.com/mudler/go-ggllm.cpp go-ggllm
-	cd go-ggllm && git checkout -b build $(GOGGLLM_VERSION) && git submodule update --init --recursive --depth 1
-
-go-ggllm/libggllm.a: go-ggllm
-	$(MAKE) -C go-ggllm BUILD_TYPE=$(BUILD_TYPE) libggllm.a
+sources/gpt4all:
+	git clone --recurse-submodules $(GPT4ALL_REPO) sources/gpt4all
+	cd sources/gpt4all && git checkout -b build $(GPT4ALL_VERSION) && git submodule update --init --recursive --depth 1
 
 ## go-piper
-go-piper:
-	git clone --recurse-submodules https://github.com/mudler/go-piper go-piper
-	cd go-piper && git checkout -b build $(PIPER_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-piper:
+	git clone --recurse-submodules https://github.com/mudler/go-piper sources/go-piper
+	cd sources/go-piper && git checkout -b build $(PIPER_VERSION) && git submodule update --init --recursive --depth 1
 
 ## BERT embeddings
-go-bert:
-	git clone --recurse-submodules https://github.com/go-skynet/go-bert.cpp go-bert
-	cd go-bert && git checkout -b build $(BERT_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-bert:
+	git clone --recurse-submodules https://github.com/go-skynet/go-bert.cpp sources/go-bert
+	cd sources/go-bert && git checkout -b build $(BERT_VERSION) && git submodule update --init --recursive --depth 1
 
 ## stable diffusion
-go-stable-diffusion:
-	git clone --recurse-submodules https://github.com/mudler/go-stable-diffusion go-stable-diffusion
-	cd go-stable-diffusion && git checkout -b build $(STABLEDIFFUSION_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-stable-diffusion:
+	git clone --recurse-submodules https://github.com/mudler/go-stable-diffusion sources/go-stable-diffusion
+	cd sources/go-stable-diffusion && git checkout -b build $(STABLEDIFFUSION_VERSION) && git submodule update --init --recursive --depth 1
 
-go-stable-diffusion/libstablediffusion.a:
-	$(MAKE) -C go-stable-diffusion libstablediffusion.a
+sources/go-stable-diffusion/libstablediffusion.a:
+	$(MAKE) -C sources/go-stable-diffusion libstablediffusion.a
 
 ## RWKV
-go-rwkv:
-	git clone --recurse-submodules $(RWKV_REPO) go-rwkv
-	cd go-rwkv && git checkout -b build $(RWKV_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-rwkv:
+	git clone --recurse-submodules $(RWKV_REPO) sources/go-rwkv
+	cd sources/go-rwkv && git checkout -b build $(RWKV_VERSION) && git submodule update --init --recursive --depth 1
 
-go-rwkv/librwkv.a: go-rwkv
-	cd go-rwkv && cd rwkv.cpp &&	cmake . -DRWKV_BUILD_SHARED_LIBRARY=OFF &&	cmake --build . && 	cp librwkv.a ..
+sources/go-rwkv/librwkv.a: sources/go-rwkv
+	cd sources/go-rwkv && cd rwkv.cpp &&	cmake . -DRWKV_BUILD_SHARED_LIBRARY=OFF &&	cmake --build . && 	cp librwkv.a ..
 
-## bloomz
-bloomz:
-	git clone --recurse-submodules https://github.com/go-skynet/bloomz.cpp bloomz
-	cd bloomz && git checkout -b build $(BLOOMZ_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-bert/libgobert.a: sources/go-bert
+	$(MAKE) -C sources/go-bert libgobert.a
 
-bloomz/libbloomz.a: bloomz
-	cd bloomz && make libbloomz.a
-
-go-bert/libgobert.a: go-bert
-	$(MAKE) -C go-bert libgobert.a
-
-backend-assets/gpt4all: gpt4all/gpt4all-bindings/golang/libgpt4all.a
+backend-assets/gpt4all: sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a
 	mkdir -p backend-assets/gpt4all
-	@cp gpt4all/gpt4all-bindings/golang/buildllm/*.so backend-assets/gpt4all/ || true
-	@cp gpt4all/gpt4all-bindings/golang/buildllm/*.dylib backend-assets/gpt4all/ || true
-	@cp gpt4all/gpt4all-bindings/golang/buildllm/*.dll backend-assets/gpt4all/ || true
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.so backend-assets/gpt4all/ || true
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dylib backend-assets/gpt4all/ || true
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dll backend-assets/gpt4all/ || true
 
-backend-assets/espeak-ng-data:
+backend-assets/espeak-ng-data: sources/go-piper
 	mkdir -p backend-assets/espeak-ng-data
-ifdef ESPEAK_DATA
-	@cp -rf $(ESPEAK_DATA)/. backend-assets/espeak-ng-data
-else
-	@echo "ESPEAK_DATA not set, skipping tts. Note that this will break the tts functionality."
-	@touch backend-assets/espeak-ng-data/keep
-endif
+	$(MAKE) -C sources/go-piper piper.o
+	@cp -rf sources/go-piper/piper/build/pi/share/espeak-ng-data/. backend-assets/espeak-ng-data
 
-gpt4all/gpt4all-bindings/golang/libgpt4all.a: gpt4all
-	$(MAKE) -C gpt4all/gpt4all-bindings/golang/ libgpt4all.a
+sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a: sources/gpt4all
+	$(MAKE) -C sources/gpt4all/gpt4all-bindings/golang/ libgpt4all.a
 
 ## CEREBRAS GPT
-go-ggml-transformers:
-	git clone --recurse-submodules https://github.com/go-skynet/go-ggml-transformers.cpp go-ggml-transformers
-	cd go-ggml-transformers && git checkout -b build $(GOGPT2_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-ggml-transformers:
+	git clone --recurse-submodules https://github.com/go-skynet/go-ggml-transformers.cpp sources/go-ggml-transformers
+	cd sources/go-ggml-transformers && git checkout -b build $(GOGPT2_VERSION) && git submodule update --init --recursive --depth 1
 
-go-ggml-transformers/libtransformers.a: go-ggml-transformers
-	$(MAKE) -C go-ggml-transformers BUILD_TYPE=$(BUILD_TYPE) libtransformers.a
+sources/go-ggml-transformers/libtransformers.a: sources/go-ggml-transformers
+	$(MAKE) -C sources/go-ggml-transformers BUILD_TYPE=$(BUILD_TYPE) libtransformers.a
 
-whisper.cpp:
-	git clone https://github.com/ggerganov/whisper.cpp.git
-	cd whisper.cpp && git checkout -b build $(WHISPER_CPP_VERSION) && git submodule update --init --recursive --depth 1
+sources/whisper.cpp:
+	git clone https://github.com/ggerganov/whisper.cpp.git sources/whisper.cpp
+	cd sources/whisper.cpp && git checkout -b build $(WHISPER_CPP_VERSION) && git submodule update --init --recursive --depth 1
 
-whisper.cpp/libwhisper.a: whisper.cpp
-	cd whisper.cpp && make libwhisper.a
+sources/whisper.cpp/libwhisper.a: sources/whisper.cpp
+	cd sources/whisper.cpp && make libwhisper.a
 
-go-llama:
-	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp go-llama
-	cd go-llama && git checkout -b build $(GOLLAMA_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-llama:
+	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp sources/go-llama
+	cd sources/go-llama && git checkout -b build $(GOLLAMA_VERSION) && git submodule update --init --recursive --depth 1
 
-go-llama-stable:
-	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp go-llama-stable
-	cd go-llama-stable && git checkout -b build $(GOLLAMA_STABLE_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-llama-ggml:
+	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp sources/go-llama-ggml
+	cd sources/go-llama-ggml && git checkout -b build $(GOLLAMA_STABLE_VERSION) && git submodule update --init --recursive --depth 1
 
-go-llama/libbinding.a: go-llama
-	$(MAKE) -C go-llama BUILD_TYPE=$(BUILD_TYPE) libbinding.a
+sources/go-llama/libbinding.a: sources/go-llama
+	$(MAKE) -C sources/go-llama BUILD_TYPE=$(BUILD_TYPE) libbinding.a
 
-go-llama-stable/libbinding.a: go-llama-stable
-	$(MAKE) -C go-llama-stable BUILD_TYPE=$(BUILD_TYPE) libbinding.a
+sources/go-llama-ggml/libbinding.a: sources/go-llama-ggml
+	$(MAKE) -C sources/go-llama-ggml BUILD_TYPE=$(STABLE_BUILD_TYPE) libbinding.a
 
-go-piper/libpiper_binding.a:
-	$(MAKE) -C go-piper libpiper_binding.a example/main
+sources/go-piper/libpiper_binding.a: sources/go-piper
+	$(MAKE) -C sources/go-piper libpiper_binding.a example/main
 
-get-sources: go-llama go-llama-stable go-ggllm go-ggml-transformers gpt4all go-piper go-rwkv whisper.cpp go-bert bloomz go-stable-diffusion
+backend/cpp/llama/llama.cpp:
+	$(MAKE) -C backend/cpp/llama llama.cpp	
+
+get-sources: backend/cpp/llama/llama.cpp sources/go-llama sources/go-llama-ggml sources/go-ggml-transformers sources/gpt4all sources/go-piper sources/go-rwkv sources/whisper.cpp sources/go-bert sources/go-stable-diffusion
 	touch $@
 
 replace:
-	$(GOCMD) mod edit -replace github.com/nomic-ai/gpt4all/gpt4all-bindings/golang=$(shell pwd)/gpt4all/gpt4all-bindings/golang
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-ggml-transformers.cpp=$(shell pwd)/go-ggml-transformers
-	$(GOCMD) mod edit -replace github.com/donomii/go-rwkv.cpp=$(shell pwd)/go-rwkv
-	$(GOCMD) mod edit -replace github.com/ggerganov/whisper.cpp=$(shell pwd)/whisper.cpp
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-bert.cpp=$(shell pwd)/go-bert
-	$(GOCMD) mod edit -replace github.com/go-skynet/bloomz.cpp=$(shell pwd)/bloomz
-	$(GOCMD) mod edit -replace github.com/mudler/go-stable-diffusion=$(shell pwd)/go-stable-diffusion
-	$(GOCMD) mod edit -replace github.com/mudler/go-piper=$(shell pwd)/go-piper
-	$(GOCMD) mod edit -replace github.com/mudler/go-ggllm.cpp=$(shell pwd)/go-ggllm
+	$(GOCMD) mod edit -replace github.com/nomic-ai/gpt4all/gpt4all-bindings/golang=$(shell pwd)/sources/gpt4all/gpt4all-bindings/golang
+	$(GOCMD) mod edit -replace github.com/go-skynet/go-ggml-transformers.cpp=$(shell pwd)/sources/go-ggml-transformers
+	$(GOCMD) mod edit -replace github.com/donomii/go-rwkv.cpp=$(shell pwd)/sources/go-rwkv
+	$(GOCMD) mod edit -replace github.com/ggerganov/whisper.cpp=$(shell pwd)/sources/whisper.cpp
+	$(GOCMD) mod edit -replace github.com/go-skynet/go-bert.cpp=$(shell pwd)/sources/go-bert
+	$(GOCMD) mod edit -replace github.com/mudler/go-stable-diffusion=$(shell pwd)/sources/go-stable-diffusion
+	$(GOCMD) mod edit -replace github.com/mudler/go-piper=$(shell pwd)/sources/go-piper
 
 prepare-sources: get-sources replace
 	$(GOCMD) mod download
+	touch $@
 
 ## GENERIC
 rebuild: ## Rebuilds the project
 	$(GOCMD) clean -cache
-	$(MAKE) -C go-llama clean
-	$(MAKE) -C go-llama-stable clean
-	$(MAKE) -C gpt4all/gpt4all-bindings/golang/ clean
-	$(MAKE) -C go-ggml-transformers clean
-	$(MAKE) -C go-rwkv clean
-	$(MAKE) -C whisper.cpp clean
-	$(MAKE) -C go-stable-diffusion clean
-	$(MAKE) -C go-bert clean
-	$(MAKE) -C bloomz clean
-	$(MAKE) -C go-piper clean
-	$(MAKE) -C go-ggllm clean
+	$(MAKE) -C sources/go-llama clean
+	$(MAKE) -C sources/go-llama-ggml clean
+	$(MAKE) -C sources/gpt4all/gpt4all-bindings/golang/ clean
+	$(MAKE) -C sources/go-ggml-transformers clean
+	$(MAKE) -C sources/go-rwkv clean
+	$(MAKE) -C sources/whisper.cpp clean
+	$(MAKE) -C sources/go-stable-diffusion clean
+	$(MAKE) -C sources/go-bert clean
+	$(MAKE) -C sources/go-piper clean
 	$(MAKE) build
 
-prepare: prepare-sources $(OPTIONAL_TARGETS) 
+prepare: prepare-sources $(OPTIONAL_TARGETS)
 	touch $@
 
 clean: ## Remove build related file
 	$(GOCMD) clean -cache
 	rm -f prepare
-	rm -rf ./go-llama
-	rm -rf ./gpt4all	
-	rm -rf ./go-llama-stable
-	rm -rf ./go-gpt2
-	rm -rf ./go-stable-diffusion
-	rm -rf ./go-ggml-transformers
-	rm -rf ./backend-assets
-	rm -rf ./go-rwkv
-	rm -rf ./go-bert
-	rm -rf ./bloomz
-	rm -rf ./whisper.cpp
-	rm -rf ./go-piper
-	rm -rf ./go-ggllm
+	rm -rf ./sources
 	rm -rf $(BINARY_NAME)
 	rm -rf release/
+	rm -rf ./backend/cpp/grpc/grpc_repo
+	rm -rf ./backend/cpp/grpc/build
+	rm -rf ./backend/cpp/grpc/installed_packages
+	$(MAKE) -C backend/cpp/llama clean
 
 ## Build:
 
@@ -278,6 +284,9 @@ dist: build
 	mkdir -p release
 	cp $(BINARY_NAME) release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-$(ARCH)
 
+osx-signed: build
+	codesign --deep --force --sign "$(OSX_SIGNING_IDENTITY)" --entitlements "./Entitlements.plist" "./$(BINARY_NAME)"
+
 ## Run
 run: prepare ## run local-ai
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) run ./
@@ -285,12 +294,12 @@ run: prepare ## run local-ai
 test-models/testmodel:
 	mkdir test-models
 	mkdir test-dir
-	wget https://huggingface.co/nnakasato/ggml-model-test/resolve/main/ggml-model-q4.bin -O test-models/testmodel
-	wget https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.en.bin -O test-models/whisper-en
-	wget https://huggingface.co/skeskinen/ggml/resolve/main/all-MiniLM-L6-v2/ggml-model-q4_0.bin -O test-models/bert
-	wget https://cdn.openai.com/whisper/draft-20220913a/micro-machines.wav -O test-dir/audio.wav
-	wget https://huggingface.co/mudler/rwkv-4-raven-1.5B-ggml/resolve/main/RWKV-4-Raven-1B5-v11-Eng99%2525-Other1%2525-20230425-ctx4096_Q4_0.bin -O test-models/rwkv
-	wget https://raw.githubusercontent.com/saharNooby/rwkv.cpp/5eb8f09c146ea8124633ab041d9ea0b1f1db4459/rwkv/20B_tokenizer.json -O test-models/rwkv.tokenizer.json
+	wget -q https://huggingface.co/nnakasato/ggml-model-test/resolve/main/ggml-model-q4.bin -O test-models/testmodel
+	wget -q https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.en.bin -O test-models/whisper-en
+	wget -q https://huggingface.co/mudler/all-MiniLM-L6-v2/resolve/main/ggml-model-q4_0.bin -O test-models/bert
+	wget -q https://cdn.openai.com/whisper/draft-20220913a/micro-machines.wav -O test-dir/audio.wav
+	wget -q https://huggingface.co/mudler/rwkv-4-raven-1.5B-ggml/resolve/main/RWKV-4-Raven-1B5-v11-Eng99%2525-Other1%2525-20230425-ctx4096_Q4_0.bin -O test-models/rwkv
+	wget -q https://raw.githubusercontent.com/saharNooby/rwkv.cpp/5eb8f09c146ea8124633ab041d9ea0b1f1db4459/rwkv/20B_tokenizer.json -O test-models/rwkv.tokenizer.json
 	cp tests/models_fixtures/* test-models
 
 prepare-test: grpcs
@@ -301,14 +310,34 @@ test: prepare test-models/testmodel grpcs
 	@echo 'Running tests'
 	export GO_TAGS="tts stablediffusion"
 	$(MAKE) prepare-test
-	HUGGINGFACE_GRPC=$(abspath ./)/extra/grpc/huggingface/huggingface.py TEST_DIR=$(abspath ./)/test-dir/ FIXTURES=$(abspath ./)/tests/fixtures CONFIG_FILE=$(abspath ./)/test-models/config.yaml MODELS_PATH=$(abspath ./)/test-models \
-	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --label-filter="!gpt4all && !llama && !llama-gguf" --flake-attempts 5 -v -r ./api ./pkg
+	HUGGINGFACE_GRPC=$(abspath ./)/backend/python/sentencetransformers/run.sh TEST_DIR=$(abspath ./)/test-dir/ FIXTURES=$(abspath ./)/tests/fixtures CONFIG_FILE=$(abspath ./)/test-models/config.yaml MODELS_PATH=$(abspath ./)/test-models \
+	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --label-filter="!gpt4all && !llama && !llama-gguf"  --flake-attempts 5 --fail-fast -v -r ./api ./pkg
 	$(MAKE) test-gpt4all
 	$(MAKE) test-llama
 	$(MAKE) test-llama-gguf
 	$(MAKE) test-tts
 	$(MAKE) test-stablediffusion
 
+prepare-e2e:
+	mkdir -p $(TEST_DIR)
+	cp -rfv $(abspath ./tests/e2e-fixtures)/gpu.yaml $(TEST_DIR)/gpu.yaml
+	test -e $(TEST_DIR)/ggllm-test-model.bin || wget -q https://huggingface.co/TheBloke/CodeLlama-7B-Instruct-GGUF/resolve/main/codellama-7b-instruct.Q2_K.gguf -O $(TEST_DIR)/ggllm-test-model.bin
+	docker build --build-arg BUILD_GRPC=true --build-arg GRPC_BACKENDS="$(GRPC_BACKENDS)" --build-arg IMAGE_TYPE=core --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg CUDA_MAJOR_VERSION=11 --build-arg CUDA_MINOR_VERSION=7 --build-arg FFMPEG=true -t localai-tests .
+
+run-e2e-image:
+	ls -liah $(abspath ./tests/e2e-fixtures)
+	docker run -p 5390:8080 -e MODELS_PATH=/models -e THREADS=1 -e DEBUG=true -d --rm -v $(TEST_DIR):/models --gpus all --name e2e-tests-$(RANDOM) localai-tests
+
+test-e2e:
+	@echo 'Running e2e tests'
+	BUILD_TYPE=$(BUILD_TYPE) \
+	LOCALAI_API=http://$(E2E_BRIDGE_IP):5390/v1 \
+	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --flake-attempts 5 -v -r ./tests/e2e
+
+teardown-e2e:
+	rm -rf $(TEST_DIR) || true
+	docker stop $$(docker ps -q --filter ancestor=localai-tests)
+
 test-gpt4all: prepare-test
 	TEST_DIR=$(abspath ./)/test-dir/ FIXTURES=$(abspath ./)/tests/fixtures CONFIG_FILE=$(abspath ./)/test-models/config.yaml MODELS_PATH=$(abspath ./)/test-models \
 	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --label-filter="gpt4all" --flake-attempts 5 -v -r ./api ./pkg
@@ -349,99 +378,141 @@ protogen: protogen-go protogen-python
 
 protogen-go:
 	protoc --go_out=. --go_opt=paths=source_relative --go-grpc_out=. --go-grpc_opt=paths=source_relative \
-    pkg/grpc/proto/backend.proto
+    backend/backend.proto
 
 protogen-python:
-	python3 -m grpc_tools.protoc -Ipkg/grpc/proto/ --python_out=extra/grpc/huggingface/ --grpc_python_out=extra/grpc/huggingface/ pkg/grpc/proto/backend.proto
-	python3 -m grpc_tools.protoc -Ipkg/grpc/proto/ --python_out=extra/grpc/autogptq/ --grpc_python_out=extra/grpc/autogptq/ pkg/grpc/proto/backend.proto
-	python3 -m grpc_tools.protoc -Ipkg/grpc/proto/ --python_out=extra/grpc/exllama/ --grpc_python_out=extra/grpc/exllama/ pkg/grpc/proto/backend.proto
-	python3 -m grpc_tools.protoc -Ipkg/grpc/proto/ --python_out=extra/grpc/bark/ --grpc_python_out=extra/grpc/bark/ pkg/grpc/proto/backend.proto
-	python3 -m grpc_tools.protoc -Ipkg/grpc/proto/ --python_out=extra/grpc/diffusers/ --grpc_python_out=extra/grpc/diffusers/ pkg/grpc/proto/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/sentencetransformers/ --grpc_python_out=backend/python/sentencetransformers/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/transformers/ --grpc_python_out=backend/python/transformers/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/autogptq/ --grpc_python_out=backend/python/autogptq/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/exllama/ --grpc_python_out=backend/python/exllama/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/bark/ --grpc_python_out=backend/python/bark/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/diffusers/ --grpc_python_out=backend/python/diffusers/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/vall-e-x/ --grpc_python_out=backend/python/vall-e-x/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/vllm/ --grpc_python_out=backend/python/vllm/ backend/backend.proto
+	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/petals/ --grpc_python_out=backend/python/petals/ backend/backend.proto
 
 ## GRPC
+# Note: it is duplicated in the Dockerfile
+prepare-extra-conda-environments:
+	$(MAKE) -C backend/python/autogptq
+	$(MAKE) -C backend/python/bark
+	$(MAKE) -C backend/python/diffusers
+	$(MAKE) -C backend/python/vllm
+	$(MAKE) -C backend/python/sentencetransformers
+	$(MAKE) -C backend/python/transformers
+	$(MAKE) -C backend/python/vall-e-x
+	$(MAKE) -C backend/python/exllama
+	$(MAKE) -C backend/python/petals
+
 
 backend-assets/grpc:
 	mkdir -p backend-assets/grpc
 
-backend-assets/grpc/falcon: backend-assets/grpc go-ggllm/libggllm.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggllm LIBRARY_PATH=$(shell pwd)/go-ggllm \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/falcon ./cmd/grpc/falcon/
-
-backend-assets/grpc/llama: backend-assets/grpc go-llama/libbinding.a
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(shell pwd)/go-llama
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-llama LIBRARY_PATH=$(shell pwd)/go-llama \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama ./cmd/grpc/llama/
-# TODO: every binary should have its own folder instead, so can have different metal implementations
+backend-assets/grpc/llama: backend-assets/grpc sources/go-llama/libbinding.a
+	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(shell pwd)/sources/go-llama
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-llama LIBRARY_PATH=$(shell pwd)/sources/go-llama \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama ./backend/go/llm/llama/
+# TODO: every binary should have its own folder instead, so can have different  implementations
 ifeq ($(BUILD_TYPE),metal)
-	cp go-llama/build/bin/ggml-metal.metal backend-assets/grpc/
+	cp backend/cpp/llama/llama.cpp/ggml-metal.metal backend-assets/grpc/
 endif
 
-backend-assets/grpc/llama-stable: backend-assets/grpc go-llama-stable/libbinding.a
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(shell pwd)/go-llama-stable
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-llama-stable LIBRARY_PATH=$(shell pwd)/go-llama \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama-stable ./cmd/grpc/llama-stable/
+## BACKEND CPP LLAMA START
+# Sets the variables in case it has to build the gRPC locally.
+INSTALLED_PACKAGES=$(CURDIR)/backend/cpp/grpc/installed_packages
+INSTALLED_LIB_CMAKE=$(INSTALLED_PACKAGES)/lib/cmake
+ADDED_CMAKE_ARGS=-Dabsl_DIR=${INSTALLED_LIB_CMAKE}/absl \
+                 -DProtobuf_DIR=${INSTALLED_LIB_CMAKE}/protobuf \
+                 -Dutf8_range_DIR=${INSTALLED_LIB_CMAKE}/utf8_range \
+                 -DgRPC_DIR=${INSTALLED_LIB_CMAKE}/grpc \
+                 -DCMAKE_CXX_STANDARD_INCLUDE_DIRECTORIES=${INSTALLED_PACKAGES}/include
 
-backend-assets/grpc/gpt4all: backend-assets/grpc backend-assets/gpt4all gpt4all/gpt4all-bindings/golang/libgpt4all.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/gpt4all/gpt4all-bindings/golang/ LIBRARY_PATH=$(shell pwd)/gpt4all/gpt4all-bindings/golang/ \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gpt4all ./cmd/grpc/gpt4all/
+backend/cpp/llama/grpc-server:
+ifdef BUILD_GRPC_FOR_BACKEND_LLAMA
+	backend/cpp/grpc/script/build_grpc.sh ${INSTALLED_PACKAGES}
+	export _PROTOBUF_PROTOC=${INSTALLED_PACKAGES}/bin/proto && \
+	export _GRPC_CPP_PLUGIN_EXECUTABLE=${INSTALLED_PACKAGES}/bin/grpc_cpp_plugin && \
+	export PATH=${PATH}:${INSTALLED_PACKAGES}/bin && \
+	CMAKE_ARGS="${CMAKE_ARGS} ${ADDED_CMAKE_ARGS}" LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama grpc-server
+else
+	echo "BUILD_GRPC_FOR_BACKEND_LLAMA is not defined."
+	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama grpc-server			
+endif
+## BACKEND CPP LLAMA END
+		
+##
+backend-assets/grpc/llama-cpp: backend-assets/grpc backend/cpp/llama/grpc-server
+	cp -rfv backend/cpp/llama/grpc-server backend-assets/grpc/llama-cpp
+# TODO: every binary should have its own folder instead, so can have different metal implementations
+ifeq ($(BUILD_TYPE),metal)
+	cp backend/cpp/llama/llama.cpp/build/bin/ggml-metal.metal backend-assets/grpc/
+endif
 
-backend-assets/grpc/dolly: backend-assets/grpc go-ggml-transformers/libtransformers.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/go-ggml-transformers \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/dolly ./cmd/grpc/dolly/
+backend-assets/grpc/llama-ggml: backend-assets/grpc sources/go-llama-ggml/libbinding.a
+	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(shell pwd)/sources/go-llama-ggml
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-llama-ggml LIBRARY_PATH=$(shell pwd)/sources/go-llama-ggml \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama-ggml ./backend/go/llm/llama-ggml/
 
-backend-assets/grpc/gpt2: backend-assets/grpc go-ggml-transformers/libtransformers.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/go-ggml-transformers \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gpt2 ./cmd/grpc/gpt2/
+backend-assets/grpc/gpt4all: backend-assets/grpc backend-assets/gpt4all sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/gpt4all/gpt4all-bindings/golang/ LIBRARY_PATH=$(shell pwd)/sources/gpt4all/gpt4all-bindings/golang/ \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gpt4all ./backend/go/llm/gpt4all/
 
-backend-assets/grpc/gptj: backend-assets/grpc go-ggml-transformers/libtransformers.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/go-ggml-transformers \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gptj ./cmd/grpc/gptj/
+backend-assets/grpc/dolly: backend-assets/grpc sources/go-ggml-transformers/libtransformers.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/sources/go-ggml-transformers \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/dolly ./backend/go/llm/dolly/
 
-backend-assets/grpc/gptneox: backend-assets/grpc go-ggml-transformers/libtransformers.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/go-ggml-transformers \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gptneox ./cmd/grpc/gptneox/
+backend-assets/grpc/gpt2: backend-assets/grpc sources/go-ggml-transformers/libtransformers.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/sources/go-ggml-transformers \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gpt2 ./backend/go/llm/gpt2/
 
-backend-assets/grpc/mpt: backend-assets/grpc go-ggml-transformers/libtransformers.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/go-ggml-transformers \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/mpt ./cmd/grpc/mpt/
+backend-assets/grpc/gptj: backend-assets/grpc sources/go-ggml-transformers/libtransformers.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/sources/go-ggml-transformers \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gptj ./backend/go/llm/gptj/
 
-backend-assets/grpc/replit: backend-assets/grpc go-ggml-transformers/libtransformers.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/go-ggml-transformers \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/replit ./cmd/grpc/replit/
+backend-assets/grpc/gptneox: backend-assets/grpc sources/go-ggml-transformers/libtransformers.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/sources/go-ggml-transformers \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gptneox ./backend/go/llm/gptneox/
 
-backend-assets/grpc/falcon-ggml: backend-assets/grpc go-ggml-transformers/libtransformers.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/go-ggml-transformers \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/falcon-ggml ./cmd/grpc/falcon-ggml/
+backend-assets/grpc/mpt: backend-assets/grpc sources/go-ggml-transformers/libtransformers.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/sources/go-ggml-transformers \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/mpt ./backend/go/llm/mpt/
 
-backend-assets/grpc/starcoder: backend-assets/grpc go-ggml-transformers/libtransformers.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/go-ggml-transformers \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/starcoder ./cmd/grpc/starcoder/
+backend-assets/grpc/replit: backend-assets/grpc sources/go-ggml-transformers/libtransformers.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/sources/go-ggml-transformers \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/replit ./backend/go/llm/replit/
 
-backend-assets/grpc/rwkv: backend-assets/grpc go-rwkv/librwkv.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-rwkv LIBRARY_PATH=$(shell pwd)/go-rwkv \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/rwkv ./cmd/grpc/rwkv/
+backend-assets/grpc/falcon-ggml: backend-assets/grpc sources/go-ggml-transformers/libtransformers.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/sources/go-ggml-transformers \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/falcon-ggml ./backend/go/llm/falcon-ggml/
 
-backend-assets/grpc/bloomz: backend-assets/grpc bloomz/libbloomz.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/bloomz LIBRARY_PATH=$(shell pwd)/bloomz \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/bloomz ./cmd/grpc/bloomz/
+backend-assets/grpc/starcoder: backend-assets/grpc sources/go-ggml-transformers/libtransformers.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-ggml-transformers LIBRARY_PATH=$(shell pwd)/sources/go-ggml-transformers \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/starcoder ./backend/go/llm/starcoder/
 
-backend-assets/grpc/bert-embeddings: backend-assets/grpc go-bert/libgobert.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-bert LIBRARY_PATH=$(shell pwd)/go-bert \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/bert-embeddings ./cmd/grpc/bert-embeddings/
+backend-assets/grpc/rwkv: backend-assets/grpc sources/go-rwkv/librwkv.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-rwkv LIBRARY_PATH=$(shell pwd)/sources/go-rwkv \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/rwkv ./backend/go/llm/rwkv
+
+backend-assets/grpc/bert-embeddings: backend-assets/grpc sources/go-bert/libgobert.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-bert LIBRARY_PATH=$(shell pwd)/sources/go-bert \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/bert-embeddings ./backend/go/llm/bert/
 
 backend-assets/grpc/langchain-huggingface: backend-assets/grpc
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/langchain-huggingface ./cmd/grpc/langchain-huggingface/
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/langchain-huggingface ./backend/go/llm/langchain/
 
-backend-assets/grpc/stablediffusion: backend-assets/grpc go-stable-diffusion/libstablediffusion.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/go-stable-diffusion/ LIBRARY_PATH=$(shell pwd)/go-stable-diffusion/ \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/stablediffusion ./cmd/grpc/stablediffusion/
+backend-assets/grpc/stablediffusion: backend-assets/grpc
+	if [ ! -f backend-assets/grpc/stablediffusion ]; then \
+		$(MAKE) sources/go-stable-diffusion/libstablediffusion.a; \
+		CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/go-stable-diffusion/ LIBRARY_PATH=$(shell pwd)/sources/go-stable-diffusion/ \
+		$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/stablediffusion ./backend/go/image/; \
+	fi
 
-backend-assets/grpc/piper: backend-assets/grpc backend-assets/espeak-ng-data go-piper/libpiper_binding.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" LIBRARY_PATH=$(shell pwd)/go-piper \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/piper ./cmd/grpc/piper/
+backend-assets/grpc/piper: backend-assets/grpc backend-assets/espeak-ng-data sources/go-piper/libpiper_binding.a
+	CGO_CXXFLAGS="$(PIPER_CGO_CXXFLAGS)" CGO_LDFLAGS="$(PIPER_CGO_LDFLAGS)" LIBRARY_PATH=$(shell pwd)/sources/go-piper \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/piper ./backend/go/tts/
 
-backend-assets/grpc/whisper: backend-assets/grpc whisper.cpp/libwhisper.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/whisper.cpp LIBRARY_PATH=$(shell pwd)/whisper.cpp \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/whisper ./cmd/grpc/whisper/
+backend-assets/grpc/whisper: backend-assets/grpc sources/whisper.cpp/libwhisper.a
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(shell pwd)/sources/whisper.cpp LIBRARY_PATH=$(shell pwd)/sources/whisper.cpp \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/whisper ./backend/go/transcribe/
 
 grpcs: prepare $(GRPC_BACKENDS)

README.md

@@ -22,14 +22,13 @@
 
 > :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/) [📣 News](https://localai.io/basics/news/) [ 🛫 Examples ](https://github.com/go-skynet/LocalAI/tree/master/examples/) [ 🖼️ Models ](https://localai.io/models/)
-
+> [💻 Quickstart](https://localai.io/basics/getting_started/) [📣 News](https://localai.io/basics/news/) [ 🛫 Examples ](https://github.com/go-skynet/LocalAI/tree/master/examples/) [ 🖼️ Models ](https://localai.io/models/) [ 🚀 Roadmap ](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap)
 
 [![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)
 
-**LocalAI** is a drop-in replacement REST API that's compatible with OpenAI API specifications for local inferencing. It allows you to run LLMs (and not only) locally or on-prem with consumer grade hardware, supporting multiple model families that are compatible with the ggml format. Does not require GPU.
+**LocalAI** is the free, Open Source OpenAI alternative. LocalAI act as a drop-in replacement REST API that’s compatible with OpenAI API specifications for local inferencing. It allows you to run LLMs, generate images, audio (and not only) locally or on-prem with consumer grade hardware, supporting multiple model families. Does not require GPU.
 
-<p align="center"><b>Follow LocalAI </b></p>
+<p  align="center"><b>Follow LocalAI </b></p>
 
 <p align="center">
 <a href="https://twitter.com/LocalAI_API" target="blank">
@@ -39,7 +38,7 @@
 <img src="https://dcbadge.vercel.app/api/server/uJAeKSAGDy?style=flat-square&theme=default-inverted" alt="Join LocalAI Discord Community"/>
 </a>
 
-<p align="center"><b>Connect with the Creator </b></p>
+<p  align="center"><b>Connect with the Creator </b></p>
 
 <p align="center">
 <a href="https://twitter.com/mudler_it" target="blank">
@@ -50,7 +49,7 @@
 </a>
 </p>
 
-<p align="center"><b>Share LocalAI Repository</b></p>
+<p  align="center"><b>Share LocalAI Repository</b></p>
 
 <p align="center">
 
@@ -64,6 +63,22 @@
 
 </p>
 
+## 💻 [Getting started](https://localai.io/basics/getting_started/index.html)
+
+## 🔥🔥 Hot topics / Roadmap
+
+[Roadmap](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap)
+
+🆕 New! [LLM finetuning guide](https://localai.io/advanced/fine-tuning/)
+
+Hot topics (looking for contributors):
+- Backends v2: https://github.com/mudler/LocalAI/issues/1126
+- Improving UX v2: https://github.com/mudler/LocalAI/issues/1373
+
+If you want to help and contribute, issues up for grabs: https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3A%22up+for+grabs%22
+
+
+
 <hr>
 
 In a nutshell:
@@ -79,8 +94,6 @@ LocalAI was created by [Ettore Di Giacinto](https://github.com/mudler/) and is a
 
 Note that this started just as a [fun weekend project](https://localai.io/#backstory) in order to try to create the necessary pieces for a full AI assistant like `ChatGPT`: the community is growing fast and we are working hard to make it better and more stable. If you want to help, please consider contributing (see below)!
 
-## 🔥🔥 [Hot topics / Roadmap](https://localai.io/#-hot-topics--roadmap)
-
 ## 🚀 [Features](https://localai.io/features/)
 
 - 📖 [Text generation with GPTs](https://localai.io/features/text-generation/) (`llama.cpp`, `gpt4all.cpp`, ... [:book: and more](https://localai.io/model-compatibility/index.html#model-compatibility-table))
@@ -91,8 +104,32 @@ Note that this started just as a [fun weekend project](https://localai.io/#backs
 - 🧠 [Embeddings generation for vector databases](https://localai.io/features/embeddings/)
 - ✍️ [Constrained grammars](https://localai.io/features/constrained_grammars/)
 - 🖼️ [Download Models directly from Huggingface ](https://localai.io/models/)
+- 🆕 [Vision API](https://localai.io/features/gpt-vision/)
 
+## 💻 Usage
 
+Check out the [Getting started](https://localai.io/basics/getting_started/index.html) section in our documentation.
+
+### 🔗 Community and integrations
+
+WebUIs:
+- https://github.com/Jirubizu/localai-admin
+- https://github.com/go-skynet/LocalAI-frontend
+
+Model galleries
+- https://github.com/go-skynet/model-gallery
+
+Other:
+- Helm chart https://github.com/go-skynet/helm-charts
+
+### 🔗 Resources
+
+- 🆕 New! [LLM finetuning guide](https://localai.io/advanced/fine-tuning/)
+- [How to build locally](https://localai.io/basics/build/index.html)
+- [How to install in Kubernetes](https://localai.io/basics/getting_started/index.html#run-localai-in-kubernetes)
+- [Projects integrating LocalAI](https://localai.io/integrations/)
+- [How tos section](https://localai.io/howtos/) (curated by our community)
+  
 ## :book: 🎥 [Media, Blogs, Social](https://localai.io/basics/news/#media-blogs-social)
 
 - [Create a slackbot for teams and OSS projects that answer to documentation](https://mudler.pm/posts/smart-slackbot-for-teams/)
@@ -100,19 +137,19 @@ Note that this started just as a [fun weekend project](https://localai.io/#backs
 - [Question Answering on Documents locally with LangChain, LocalAI, Chroma, and GPT4All](https://mudler.pm/posts/localai-question-answering/)
 - [Tutorial to use k8sgpt with LocalAI](https://medium.com/@tyler_97636/k8sgpt-localai-unlock-kubernetes-superpowers-for-free-584790de9b65)
 
-## 💻 Usage
+## Citation
 
-Check out the [Getting started](https://localai.io/basics/getting_started/index.html) section in our documentation.
+If you utilize this repository, data in a downstream project, please consider citing it with:
 
-### 💡 Example: Use GPT4ALL-J model
-
-See the [documentation](https://localai.io/basics/getting_started/#example-use-gpt4all-j-model-with-docker-compose)
-
-### 🔗 Resources
-
-- [How to build locally](https://localai.io/basics/build/index.html)
-- [How to install in Kubernetes](https://localai.io/basics/getting_started/index.html#run-localai-in-kubernetes)
-- [Projects integrating LocalAI](https://localai.io/integrations/)
+```
+@misc{localai,
+  author = {Ettore Di Giacinto},
+  title = {LocalAI: The free, Open source OpenAI alternative},
+  year = {2023},
+  publisher = {GitHub},
+  journal = {GitHub repository},
+  howpublished = {\url{https://github.com/go-skynet/LocalAI}},
+```
 
 ## ❤️ Sponsors
 
@@ -127,6 +164,11 @@ A huge thank you to our generous sponsors who support this project:
 |  [Spectro Cloud](https://www.spectrocloud.com/)  |  
 |  Spectro Cloud kindly supports LocalAI by providing GPU and computing resources to run tests on lamdalabs!  |
 
+And a huge shout-out to individuals sponsoring the project by donating hardware or backing the project. 
+
+- [Sponsor list](https://github.com/sponsors/mudler)
+- JDAM00 (donating HW for the CI)
+
 ## 🌟 Star history
 
 [![LocalAI Star history Chart](https://api.star-history.com/svg?repos=go-skynet/LocalAI&type=Date)](https://star-history.com/#go-skynet/LocalAI&Date)

api/api.go

@@ -11,7 +11,9 @@ import (
 	"github.com/go-skynet/LocalAI/api/options"
 	"github.com/go-skynet/LocalAI/api/schema"
 	"github.com/go-skynet/LocalAI/internal"
+	"github.com/go-skynet/LocalAI/metrics"
 	"github.com/go-skynet/LocalAI/pkg/assets"
+	"github.com/go-skynet/LocalAI/pkg/model"
 
 	"github.com/gofiber/fiber/v2"
 	"github.com/gofiber/fiber/v2/middleware/cors"
@@ -78,6 +80,22 @@ func Startup(opts ...options.AppOption) (*options.Option, *config.ConfigLoader,
 		options.Loader.StopAllGRPC()
 	}()
 
+	if options.WatchDog {
+		wd := model.NewWatchDog(
+			options.Loader,
+			options.WatchDogBusyTimeout,
+			options.WatchDogIdleTimeout,
+			options.WatchDogBusy,
+			options.WatchDogIdle)
+		options.Loader.SetWatchDog(wd)
+		go wd.Run()
+		go func() {
+			<-options.Context.Done()
+			log.Debug().Msgf("Context canceled, shutting down")
+			wd.Shutdown()
+		}()
+	}
+
 	return options, cl, nil
 }
 
@@ -120,6 +138,9 @@ func App(opts ...options.AppOption) (*fiber.App, error) {
 
 	// Default middleware config
 	app.Use(recover.New())
+	if options.Metrics != nil {
+		app.Use(metrics.APIMiddleware(options.Metrics))
+	}
 
 	// Auth middleware checking if API key is valid. If no API key is set, no auth is required.
 	auth := func(c *fiber.Ctx) error {
@@ -168,9 +189,14 @@ func App(opts ...options.AppOption) (*fiber.App, error) {
 		}{Version: internal.PrintableVersion()})
 	})
 
-	app.Post("/models/apply", auth, localai.ApplyModelGalleryEndpoint(options.Loader.ModelPath, cl, galleryService.C, options.Galleries))
-	app.Get("/models/available", auth, localai.ListModelFromGalleryEndpoint(options.Galleries, options.Loader.ModelPath))
-	app.Get("/models/jobs/:uuid", auth, localai.GetOpStatusEndpoint(galleryService))
+	modelGalleryService := localai.CreateModelGalleryService(options.Galleries, options.Loader.ModelPath, galleryService)
+	app.Post("/models/apply", auth, modelGalleryService.ApplyModelGalleryEndpoint())
+	app.Get("/models/available", auth, modelGalleryService.ListModelFromGalleryEndpoint())
+	app.Get("/models/galleries", auth, modelGalleryService.ListModelGalleriesEndpoint())
+	app.Post("/models/galleries", auth, modelGalleryService.AddModelGalleryEndpoint())
+	app.Delete("/models/galleries", auth, modelGalleryService.RemoveModelGalleryEndpoint())
+	app.Get("/models/jobs/:uuid", auth, modelGalleryService.GetOpStatusEndpoint())
+	app.Get("/models/jobs", auth, modelGalleryService.GetAllStatusEndpoint())
 
 	// openAI compatible API endpoint
 
@@ -224,5 +250,7 @@ func App(opts ...options.AppOption) (*fiber.App, error) {
 	app.Get("/v1/models", auth, openai.ListModelsEndpoint(options.Loader, cl))
 	app.Get("/models", auth, openai.ListModelsEndpoint(options.Loader, cl))
 
+	app.Get("/metrics", metrics.MetricsHandler())
+
 	return app, nil
 }

api/api_test.go

@@ -15,6 +15,7 @@ import (
 
 	. "github.com/go-skynet/LocalAI/api"
 	"github.com/go-skynet/LocalAI/api/options"
+	"github.com/go-skynet/LocalAI/metrics"
 	"github.com/go-skynet/LocalAI/pkg/gallery"
 	"github.com/go-skynet/LocalAI/pkg/model"
 	"github.com/go-skynet/LocalAI/pkg/utils"
@@ -162,8 +163,12 @@ var _ = Describe("API test", func() {
 				},
 			}
 
+			metricsService, err := metrics.SetupMetrics()
+			Expect(err).ToNot(HaveOccurred())
+
 			app, err = App(
 				append(commonOpts,
+					options.WithMetrics(metricsService),
 					options.WithContext(c),
 					options.WithGalleries(galleries),
 					options.WithModelLoader(modelLoader), options.WithBackendAssets(backendAssets), options.WithBackendAssetsOutput(tmpdir))...)
@@ -296,7 +301,7 @@ var _ = Describe("API test", func() {
 				response := postModelApplyRequest("http://127.0.0.1:9090/models/apply", modelApplyRequest{
 					URL:       "github:go-skynet/model-gallery/openllama_3b.yaml",
 					Name:      "openllama_3b",
-					Overrides: map[string]interface{}{"backend": "llama-stable", "mmap": true, "f16": true, "context_size": 128},
+					Overrides: map[string]interface{}{"backend": "llama-ggml", "mmap": true, "f16": true, "context_size": 128},
 				})
 
 				Expect(response["uuid"]).ToNot(BeEmpty(), fmt.Sprint(response))
@@ -363,9 +368,10 @@ var _ = Describe("API test", func() {
 				if runtime.GOOS != "linux" {
 					Skip("test supported only on linux")
 				}
+				modelName := "codellama"
 				response := postModelApplyRequest("http://127.0.0.1:9090/models/apply", modelApplyRequest{
-					URL:       "github:go-skynet/model-gallery/openllama-3b-gguf.yaml",
-					Name:      "openllama_3b_gguf",
+					URL:       "github:go-skynet/model-gallery/codellama-7b-instruct.yaml",
+					Name:      modelName,
 					Overrides: map[string]interface{}{"backend": "llama", "mmap": true, "f16": true, "context_size": 128},
 				})
 
@@ -378,17 +384,22 @@ var _ = Describe("API test", func() {
 					return response["processed"].(bool)
 				}, "360s", "10s").Should(Equal(true))
 
-				By("testing completion")
-				resp, err := client.CreateCompletion(context.TODO(), openai.CompletionRequest{Model: "openllama_3b_gguf", Prompt: "Count up to five: one, two, three, four, "})
+				By("testing chat")
+				resp, err := client.CreateChatCompletion(context.TODO(), openai.ChatCompletionRequest{Model: modelName, Messages: []openai.ChatCompletionMessage{
+					{
+						Role:    "user",
+						Content: "How much is 2+2?",
+					},
+				}})
 				Expect(err).ToNot(HaveOccurred())
 				Expect(len(resp.Choices)).To(Equal(1))
-				Expect(resp.Choices[0].Text).To(ContainSubstring("five"))
+				Expect(resp.Choices[0].Message.Content).To(Or(ContainSubstring("4"), ContainSubstring("four")))
 
 				By("testing functions")
 				resp2, err := client.CreateChatCompletion(
 					context.TODO(),
 					openai.ChatCompletionRequest{
-						Model: "openllama_3b_gguf",
+						Model: modelName,
 						Messages: []openai.ChatCompletionMessage{
 							{
 								Role:    "user",
@@ -424,7 +435,7 @@ var _ = Describe("API test", func() {
 				var res map[string]string
 				err = json.Unmarshal([]byte(resp2.Choices[0].Message.FunctionCall.Arguments), &res)
 				Expect(err).ToNot(HaveOccurred())
-				Expect(res["location"]).To(Equal("San Francisco, California"), fmt.Sprint(res))
+				Expect(res["location"]).To(Equal("San Francisco"), fmt.Sprint(res))
 				Expect(res["unit"]).To(Equal("celcius"), fmt.Sprint(res))
 				Expect(string(resp2.Choices[0].FinishReason)).To(Equal("function_call"), fmt.Sprint(resp2.Choices[0].FinishReason))
 			})
@@ -446,7 +457,7 @@ var _ = Describe("API test", func() {
 				Eventually(func() bool {
 					response := getModelStatus("http://127.0.0.1:9090/models/jobs/" + uuid)
 					return response["processed"].(bool)
-				}, "360s", "10s").Should(Equal(true))
+				}, "960s", "10s").Should(Equal(true))
 
 				resp, err := client.CreateChatCompletion(context.TODO(), openai.ChatCompletionRequest{Model: "gpt4all-j", Messages: []openai.ChatCompletionMessage{openai.ChatCompletionMessage{Role: "user", Content: "How are you?"}}})
 				Expect(err).ToNot(HaveOccurred())
@@ -473,9 +484,13 @@ var _ = Describe("API test", func() {
 				},
 			}
 
+			metricsService, err := metrics.SetupMetrics()
+			Expect(err).ToNot(HaveOccurred())
+
 			app, err = App(
 				append(commonOpts,
 					options.WithContext(c),
+					options.WithMetrics(metricsService),
 					options.WithAudioDir(tmpdir),
 					options.WithImageDir(tmpdir),
 					options.WithGalleries(galleries),
@@ -577,12 +592,15 @@ var _ = Describe("API test", func() {
 			modelLoader = model.NewModelLoader(os.Getenv("MODELS_PATH"))
 			c, cancel = context.WithCancel(context.Background())
 
-			var err error
+			metricsService, err := metrics.SetupMetrics()
+			Expect(err).ToNot(HaveOccurred())
+
 			app, err = App(
 				append(commonOpts,
 					options.WithExternalBackend("huggingface", os.Getenv("HUGGINGFACE_GRPC")),
 					options.WithContext(c),
 					options.WithModelLoader(modelLoader),
+					options.WithMetrics(metricsService),
 				)...)
 			Expect(err).ToNot(HaveOccurred())
 			go app.Listen("127.0.0.1:9090")
@@ -669,7 +687,7 @@ var _ = Describe("API test", func() {
 					Input: []string{"sun", "cat"},
 				},
 			)
-			Expect(err).ToNot(HaveOccurred())
+			Expect(err).ToNot(HaveOccurred(), err)
 			Expect(len(resp.Data[0].Embedding)).To(BeNumerically("==", 384))
 			Expect(len(resp.Data[1].Embedding)).To(BeNumerically("==", 384))
 
@@ -686,7 +704,7 @@ var _ = Describe("API test", func() {
 		})
 
 		Context("External gRPC calls", func() {
-			It("calculate embeddings with huggingface", func() {
+			It("calculate embeddings with sentencetransformers", func() {
 				if runtime.GOOS != "linux" {
 					Skip("test supported only on linux")
 				}
@@ -786,10 +804,13 @@ var _ = Describe("API test", func() {
 			modelLoader = model.NewModelLoader(os.Getenv("MODELS_PATH"))
 			c, cancel = context.WithCancel(context.Background())
 
-			var err error
+			metricsService, err := metrics.SetupMetrics()
+			Expect(err).ToNot(HaveOccurred())
+
 			app, err = App(
 				append(commonOpts,
 					options.WithContext(c),
+					options.WithMetrics(metricsService),
 					options.WithModelLoader(modelLoader),
 					options.WithConfigFile(os.Getenv("CONFIG_FILE")))...,
 			)

api/backend/image.go

@@ -20,6 +20,9 @@ func ImageGeneration(height, width, mode, step, seed int, positive_prompt, negat
 			SchedulerType: c.Diffusers.SchedulerType,
 			PipelineType:  c.Diffusers.PipelineType,
 			CFGScale:      c.Diffusers.CFGScale,
+			LoraAdapter:   c.LoraAdapter,
+			LoraScale:     c.LoraScale,
+			LoraBase:      c.LoraBase,
 			IMG2IMG:       c.Diffusers.IMG2IMG,
 			CLIPModel:     c.Diffusers.ClipModel,
 			CLIPSubfolder: c.Diffusers.ClipSubFolder,

api/backend/llm.go

@@ -6,6 +6,7 @@ import (
 	"regexp"
 	"strings"
 	"sync"
+	"unicode/utf8"
 
 	config "github.com/go-skynet/LocalAI/api/config"
 	"github.com/go-skynet/LocalAI/api/options"
@@ -25,7 +26,7 @@ type TokenUsage struct {
 	Completion int
 }
 
-func ModelInference(ctx context.Context, s string, loader *model.ModelLoader, c config.Config, o *options.Option, tokenCallback func(string, TokenUsage) bool) (func() (LLMResponse, error), error) {
+func ModelInference(ctx context.Context, s string, images []string, loader *model.ModelLoader, c config.Config, o *options.Option, tokenCallback func(string, TokenUsage) bool) (func() (LLMResponse, error), error) {
 	modelFile := c.Model
 
 	grpcOpts := gRPCModelOpts(c)
@@ -71,6 +72,7 @@ func ModelInference(ctx context.Context, s string, loader *model.ModelLoader, c
 	fn := func() (LLMResponse, error) {
 		opts := gRPCPredictOpts(c, loader.ModelPath)
 		opts.Prompt = s
+		opts.Images = images
 
 		tokenUsage := TokenUsage{}
 
@@ -97,9 +99,23 @@ func ModelInference(ctx context.Context, s string, loader *model.ModelLoader, c
 
 		if tokenCallback != nil {
 			ss := ""
-			err := inferenceModel.PredictStream(ctx, opts, func(s []byte) {
-				tokenCallback(string(s), tokenUsage)
-				ss += string(s)
+
+			var partialRune []byte
+			err := inferenceModel.PredictStream(ctx, opts, func(chars []byte) {
+				partialRune = append(partialRune, chars...)
+
+				for len(partialRune) > 0 {
+					r, size := utf8.DecodeRune(partialRune)
+					if r == utf8.RuneError {
+						// incomplete rune, wait for more bytes
+						break
+					}
+
+					tokenCallback(string(r), tokenUsage)
+					ss += string(r)
+
+					partialRune = partialRune[size:]
+				}
 			})
 			return LLMResponse{
 				Response: ss,

api/backend/options.go

@@ -16,6 +16,10 @@ func modelOpts(c config.Config, o *options.Option, opts []model.Option) []model.
 		opts = append(opts, model.WithSingleActiveBackend())
 	}
 
+	if o.ParallelBackendRequests {
+		opts = append(opts, model.EnableParallelRequests)
+	}
+
 	if c.GRPC.Attempts != 0 {
 		opts = append(opts, model.WithGRPCAttempts(c.GRPC.Attempts))
 	}
@@ -38,26 +42,35 @@ func gRPCModelOpts(c config.Config) *pb.ModelOptions {
 	}
 
 	return &pb.ModelOptions{
-		ContextSize:   int32(c.ContextSize),
-		Seed:          int32(c.Seed),
-		NBatch:        int32(b),
-		NoMulMatQ:     c.NoMulMatQ,
-		LoraAdapter:   c.LoraAdapter,
-		LoraBase:      c.LoraBase,
-		NGQA:          c.NGQA,
-		RMSNormEps:    c.RMSNormEps,
-		F16Memory:     c.F16,
-		MLock:         c.MMlock,
-		RopeFreqBase:  c.RopeFreqBase,
-		RopeFreqScale: c.RopeFreqScale,
-		NUMA:          c.NUMA,
-		Embeddings:    c.Embeddings,
-		LowVRAM:       c.LowVRAM,
-		NGPULayers:    int32(c.NGPULayers),
-		MMap:          c.MMap,
-		MainGPU:       c.MainGPU,
-		Threads:       int32(c.Threads),
-		TensorSplit:   c.TensorSplit,
+		ContextSize:    int32(c.ContextSize),
+		Seed:           int32(c.Seed),
+		NBatch:         int32(b),
+		NoMulMatQ:      c.NoMulMatQ,
+		DraftModel:     c.DraftModel,
+		AudioPath:      c.VallE.AudioPath,
+		Quantization:   c.Quantization,
+		MMProj:         c.MMProj,
+		YarnExtFactor:  c.YarnExtFactor,
+		YarnAttnFactor: c.YarnAttnFactor,
+		YarnBetaFast:   c.YarnBetaFast,
+		YarnBetaSlow:   c.YarnBetaSlow,
+		LoraAdapter:    c.LoraAdapter,
+		LoraBase:       c.LoraBase,
+		LoraScale:      c.LoraScale,
+		NGQA:           c.NGQA,
+		RMSNormEps:     c.RMSNormEps,
+		F16Memory:      c.F16,
+		MLock:          c.MMlock,
+		RopeFreqBase:   c.RopeFreqBase,
+		RopeFreqScale:  c.RopeFreqScale,
+		NUMA:           c.NUMA,
+		Embeddings:     c.Embeddings,
+		LowVRAM:        c.LowVRAM,
+		NGPULayers:     int32(c.NGPULayers),
+		MMap:           c.MMap,
+		MainGPU:        c.MainGPU,
+		Threads:        int32(c.Threads),
+		TensorSplit:    c.TensorSplit,
 		// AutoGPTQ
 		ModelBaseName:    c.AutoGPTQ.ModelBaseName,
 		Device:           c.AutoGPTQ.Device,
@@ -78,6 +91,7 @@ func gRPCPredictOpts(c config.Config, modelPath string) *pb.PredictOptions {
 	return &pb.PredictOptions{
 		Temperature:         float32(c.Temperature),
 		TopP:                float32(c.TopP),
+		NDraft:              c.NDraft,
 		TopK:                int32(c.TopK),
 		Tokens:              int32(c.Maxtokens),
 		Threads:             int32(c.Threads),

api/config/config.go

@@ -43,6 +43,13 @@ type Config struct {
 
 	// GRPC Options
 	GRPC GRPC `yaml:"grpc"`
+
+	// Vall-e-x
+	VallE VallE `yaml:"vall-e"`
+}
+
+type VallE struct {
+	AudioPath string `yaml:"audio_path"`
 }
 
 type FeatureFlag map[string]*bool
@@ -93,7 +100,18 @@ type LLMConfig struct {
 	NUMA            bool     `yaml:"numa"`
 	LoraAdapter     string   `yaml:"lora_adapter"`
 	LoraBase        string   `yaml:"lora_base"`
+	LoraScale       float32  `yaml:"lora_scale"`
 	NoMulMatQ       bool     `yaml:"no_mulmatq"`
+	DraftModel      string   `yaml:"draft_model"`
+	NDraft          int32    `yaml:"n_draft"`
+	Quantization    string   `yaml:"quantization"`
+	MMProj          string   `yaml:"mmproj"`
+
+	RopeScaling    string  `yaml:"rope_scaling"`
+	YarnExtFactor  float32 `yaml:"yarn_ext_factor"`
+	YarnAttnFactor float32 `yaml:"yarn_attn_factor"`
+	YarnBetaFast   float32 `yaml:"yarn_beta_fast"`
+	YarnBetaSlow   float32 `yaml:"yarn_beta_slow"`
 }
 
 type AutoGPTQ struct {
@@ -259,7 +277,7 @@ func (cm *ConfigLoader) LoadConfigs(path string) error {
 	}
 	for _, file := range files {
 		// Skip templates, YAML and .keep files
-		if !strings.Contains(file.Name(), ".yaml") {
+		if !strings.Contains(file.Name(), ".yaml") && !strings.Contains(file.Name(), ".yml") {
 			continue
 		}
 		c, err := ReadConfig(filepath.Join(path, file.Name()))

api/localai/backend_monitor.go

@@ -123,13 +123,12 @@ func BackendMonitorEndpoint(bm BackendMonitor) func(c *fiber.Ctx) error {
 			return err
 		}
 
-		client := bm.options.Loader.CheckIsLoaded(backendId)
-
-		if client == nil {
+		model := bm.options.Loader.CheckIsLoaded(backendId)
+		if model == "" {
 			return fmt.Errorf("backend %s is not currently loaded", backendId)
 		}
 
-		status, rpcErr := client.Status(context.TODO())
+		status, rpcErr := model.GRPC(false, nil).Status(context.TODO())
 		if rpcErr != nil {
 			log.Warn().Msgf("backend %s experienced an error retrieving status info: %s", backendId, rpcErr.Error())
 			val, slbErr := bm.SampleLocalBackendProcess(backendId)

api/localai/gallery.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"os"
+	"slices"
 	"strings"
 	"sync"
 
@@ -27,6 +28,7 @@ type galleryOp struct {
 }
 
 type galleryOpStatus struct {
+	FileName           string  `json:"file_name"`
 	Error              error   `json:"error"`
 	Processed          bool    `json:"processed"`
 	Message            string  `json:"message"`
@@ -50,7 +52,6 @@ func NewGalleryService(modelPath string) *galleryApplier {
 	}
 }
 
-// prepareModel applies a
 func prepareModel(modelPath string, req gallery.GalleryModel, cm *config.ConfigLoader, downloadStatus func(string, string, string, float64)) error {
 
 	config, err := gallery.GetGalleryConfigFromURL(req.URL)
@@ -76,6 +77,13 @@ func (g *galleryApplier) getStatus(s string) *galleryOpStatus {
 	return g.statuses[s]
 }
 
+func (g *galleryApplier) getAllStatus() map[string]*galleryOpStatus {
+	g.Lock()
+	defer g.Unlock()
+
+	return g.statuses
+}
+
 func (g *galleryApplier) Start(c context.Context, cm *config.ConfigLoader) {
 	go func() {
 		for {
@@ -94,7 +102,7 @@ func (g *galleryApplier) Start(c context.Context, cm *config.ConfigLoader) {
 
 				// displayDownload displays the download progress
 				progressCallback := func(fileName string, current string, total string, percentage float64) {
-					g.updateStatus(op.id, &galleryOpStatus{Message: "processing", Progress: percentage, TotalFileSize: total, DownloadedFileSize: current})
+					g.updateStatus(op.id, &galleryOpStatus{Message: "processing", FileName: fileName, Progress: percentage, TotalFileSize: total, DownloadedFileSize: current})
 					utils.DisplayDownloadFunction(fileName, current, total, percentage)
 				}
 
@@ -176,18 +184,12 @@ func ApplyGalleryFromString(modelPath, s string, cm *config.ConfigLoader, galler
 	return processRequests(modelPath, s, cm, galleries, requests)
 }
 
-/// Endpoints
+/// Endpoint Service
 
-func GetOpStatusEndpoint(g *galleryApplier) func(c *fiber.Ctx) error {
-	return func(c *fiber.Ctx) error {
-
-		status := g.getStatus(c.Params("uuid"))
-		if status == nil {
-			return fmt.Errorf("could not find any status for ID")
-		}
-
-		return c.JSON(status)
-	}
+type ModelGalleryService struct {
+	galleries      []gallery.Gallery
+	modelPath      string
+	galleryApplier *galleryApplier
 }
 
 type GalleryModel struct {
@@ -195,7 +197,31 @@ type GalleryModel struct {
 	gallery.GalleryModel
 }
 
-func ApplyModelGalleryEndpoint(modelPath string, cm *config.ConfigLoader, g chan galleryOp, galleries []gallery.Gallery) func(c *fiber.Ctx) error {
+func CreateModelGalleryService(galleries []gallery.Gallery, modelPath string, galleryApplier *galleryApplier) ModelGalleryService {
+	return ModelGalleryService{
+		galleries:      galleries,
+		modelPath:      modelPath,
+		galleryApplier: galleryApplier,
+	}
+}
+
+func (mgs *ModelGalleryService) GetOpStatusEndpoint() func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		status := mgs.galleryApplier.getStatus(c.Params("uuid"))
+		if status == nil {
+			return fmt.Errorf("could not find any status for ID")
+		}
+		return c.JSON(status)
+	}
+}
+
+func (mgs *ModelGalleryService) GetAllStatusEndpoint() func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		return c.JSON(mgs.galleryApplier.getAllStatus())
+	}
+}
+
+func (mgs *ModelGalleryService) ApplyModelGalleryEndpoint() func(c *fiber.Ctx) error {
 	return func(c *fiber.Ctx) error {
 		input := new(GalleryModel)
 		// Get input data from the request body
@@ -207,11 +233,11 @@ func ApplyModelGalleryEndpoint(modelPath string, cm *config.ConfigLoader, g chan
 		if err != nil {
 			return err
 		}
-		g <- galleryOp{
+		mgs.galleryApplier.C <- galleryOp{
 			req:         input.GalleryModel,
 			id:          uuid.String(),
 			galleryName: input.ID,
-			galleries:   galleries,
+			galleries:   mgs.galleries,
 		}
 		return c.JSON(struct {
 			ID        string `json:"uuid"`
@@ -220,11 +246,11 @@ func ApplyModelGalleryEndpoint(modelPath string, cm *config.ConfigLoader, g chan
 	}
 }
 
-func ListModelFromGalleryEndpoint(galleries []gallery.Gallery, basePath string) func(c *fiber.Ctx) error {
+func (mgs *ModelGalleryService) ListModelFromGalleryEndpoint() func(c *fiber.Ctx) error {
 	return func(c *fiber.Ctx) error {
-		log.Debug().Msgf("Listing models from galleries: %+v", galleries)
+		log.Debug().Msgf("Listing models from galleries: %+v", mgs.galleries)
 
-		models, err := gallery.AvailableGalleryModels(galleries, basePath)
+		models, err := gallery.AvailableGalleryModels(mgs.galleries, mgs.modelPath)
 		if err != nil {
 			return err
 		}
@@ -239,3 +265,56 @@ func ListModelFromGalleryEndpoint(galleries []gallery.Gallery, basePath string)
 		return c.Send(dat)
 	}
 }
+
+// NOTE: This is different (and much simpler!) than above! This JUST lists the model galleries that have been loaded, not their contents!
+func (mgs *ModelGalleryService) ListModelGalleriesEndpoint() func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		log.Debug().Msgf("Listing model galleries %+v", mgs.galleries)
+		dat, err := json.Marshal(mgs.galleries)
+		if err != nil {
+			return err
+		}
+		return c.Send(dat)
+	}
+}
+
+func (mgs *ModelGalleryService) AddModelGalleryEndpoint() func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		input := new(gallery.Gallery)
+		// Get input data from the request body
+		if err := c.BodyParser(input); err != nil {
+			return err
+		}
+		if slices.ContainsFunc(mgs.galleries, func(gallery gallery.Gallery) bool {
+			return gallery.Name == input.Name
+		}) {
+			return fmt.Errorf("%s already exists", input.Name)
+		}
+		dat, err := json.Marshal(mgs.galleries)
+		if err != nil {
+			return err
+		}
+		log.Debug().Msgf("Adding %+v to gallery list", *input)
+		mgs.galleries = append(mgs.galleries, *input)
+		return c.Send(dat)
+	}
+}
+
+func (mgs *ModelGalleryService) RemoveModelGalleryEndpoint() func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		input := new(gallery.Gallery)
+		// Get input data from the request body
+		if err := c.BodyParser(input); err != nil {
+			return err
+		}
+		if !slices.ContainsFunc(mgs.galleries, func(gallery gallery.Gallery) bool {
+			return gallery.Name == input.Name
+		}) {
+			return fmt.Errorf("%s is not currently registered", input.Name)
+		}
+		mgs.galleries = slices.DeleteFunc(mgs.galleries, func(gallery gallery.Gallery) bool {
+			return gallery.Name == input.Name
+		})
+		return c.Send(nil)
+	}
+}

api/openai/chat.go

@@ -6,6 +6,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"strings"
+	"time"
 
 	"github.com/go-skynet/LocalAI/api/backend"
 	config "github.com/go-skynet/LocalAI/api/config"
@@ -15,15 +16,20 @@ import (
 	model "github.com/go-skynet/LocalAI/pkg/model"
 	"github.com/go-skynet/LocalAI/pkg/utils"
 	"github.com/gofiber/fiber/v2"
+	"github.com/google/uuid"
 	"github.com/rs/zerolog/log"
 	"github.com/valyala/fasthttp"
 )
 
 func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx) error {
 	emptyMessage := ""
+	id := uuid.New().String()
+	created := int(time.Now().Unix())
 
 	process := func(s string, req *schema.OpenAIRequest, config *config.Config, loader *model.ModelLoader, responses chan schema.OpenAIResponse) {
 		initialMessage := schema.OpenAIResponse{
+			ID:      id,
+			Created: created,
 			Model:   req.Model, // we have to return what the user sent here, due to OpenAI spec.
 			Choices: []schema.Choice{{Delta: &schema.Message{Role: "assistant", Content: &emptyMessage}}},
 			Object:  "chat.completion.chunk",
@@ -32,6 +38,8 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 
 		ComputeChoices(req, s, config, o, loader, func(s string, c *[]schema.Choice) {}, func(s string, usage backend.TokenUsage) bool {
 			resp := schema.OpenAIResponse{
+				ID:      id,
+				Created: created,
 				Model:   req.Model, // we have to return what the user sent here, due to OpenAI spec.
 				Choices: []schema.Choice{{Delta: &schema.Message{Content: &s}, Index: 0}},
 				Object:  "chat.completion.chunk",
@@ -73,6 +81,10 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 			noActionDescription = config.FunctionsConfig.NoActionDescriptionName
 		}
 
+		if input.ResponseFormat.Type == "json_object" {
+			input.Grammar = grammar.JSONBNF
+		}
+
 		// process functions if we have any defined or if we have a function call string
 		if len(input.Functions) > 0 && config.ShouldUseFunctions() {
 			log.Debug().Msgf("Response needs to process functions")
@@ -132,14 +144,14 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 				}
 			}
 			r := config.Roles[role]
-			contentExists := i.Content != nil && *i.Content != ""
+			contentExists := i.Content != nil && i.StringContent != ""
 			// First attempt to populate content via a chat message specific template
 			if config.TemplateConfig.ChatMessage != "" {
 				chatMessageData := model.ChatMessageTemplateData{
 					SystemPrompt: config.SystemPrompt,
 					Role:         r,
 					RoleName:     role,
-					Content:      *i.Content,
+					Content:      i.StringContent,
 					MessageIndex: messageIndex,
 				}
 				templatedChatMessage, err := o.Loader.EvaluateTemplateForChatMessage(config.TemplateConfig.ChatMessage, chatMessageData)
@@ -158,7 +170,7 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 			if content == "" {
 				if r != "" {
 					if contentExists {
-						content = fmt.Sprint(r, " ", *i.Content)
+						content = fmt.Sprint(r, i.StringContent)
 					}
 					if i.FunctionCall != nil {
 						j, err := json.Marshal(i.FunctionCall)
@@ -172,7 +184,7 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 					}
 				} else {
 					if contentExists {
-						content = fmt.Sprint(*i.Content)
+						content = fmt.Sprint(i.StringContent)
 					}
 					if i.FunctionCall != nil {
 						j, err := json.Marshal(i.FunctionCall)
@@ -261,7 +273,9 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 				}
 
 				resp := &schema.OpenAIResponse{
-					Model: input.Model, // we have to return what the user sent here, due to OpenAI spec.
+					ID:      id,
+					Created: created,
+					Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
 					Choices: []schema.Choice{
 						{
 							FinishReason: "stop",
@@ -324,7 +338,11 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 					// Otherwise ask the LLM to understand the JSON output and the context, and return a message
 					// Note: This costs (in term of CPU) another computation
 					config.Grammar = ""
-					predFunc, err := backend.ModelInference(input.Context, predInput, o.Loader, *config, o, nil)
+					images := []string{}
+					for _, m := range input.Messages {
+						images = append(images, m.StringImages...)
+					}
+					predFunc, err := backend.ModelInference(input.Context, predInput, images, o.Loader, *config, o, nil)
 					if err != nil {
 						log.Error().Msgf("inference error: %s", err.Error())
 						return
@@ -355,6 +373,8 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 		}
 
 		resp := &schema.OpenAIResponse{
+			ID:      id,
+			Created: created,
 			Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
 			Choices: result,
 			Object:  "chat.completion",

api/openai/completion.go

@@ -6,23 +6,31 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
+	"time"
 
 	"github.com/go-skynet/LocalAI/api/backend"
 	config "github.com/go-skynet/LocalAI/api/config"
 	"github.com/go-skynet/LocalAI/api/options"
 	"github.com/go-skynet/LocalAI/api/schema"
+	"github.com/go-skynet/LocalAI/pkg/grammar"
 	model "github.com/go-skynet/LocalAI/pkg/model"
 	"github.com/gofiber/fiber/v2"
+	"github.com/google/uuid"
 	"github.com/rs/zerolog/log"
 	"github.com/valyala/fasthttp"
 )
 
 // https://platform.openai.com/docs/api-reference/completions
 func CompletionEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx) error {
+	id := uuid.New().String()
+	created := int(time.Now().Unix())
+
 	process := func(s string, req *schema.OpenAIRequest, config *config.Config, loader *model.ModelLoader, responses chan schema.OpenAIResponse) {
 		ComputeChoices(req, s, config, o, loader, func(s string, c *[]schema.Choice) {}, func(s string, usage backend.TokenUsage) bool {
 			resp := schema.OpenAIResponse{
-				Model: req.Model, // we have to return what the user sent here, due to OpenAI spec.
+				ID:      id,
+				Created: created,
+				Model:   req.Model, // we have to return what the user sent here, due to OpenAI spec.
 				Choices: []schema.Choice{
 					{
 						Index: 0,
@@ -57,6 +65,10 @@ func CompletionEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fibe
 			return fmt.Errorf("failed reading parameters from request:%w", err)
 		}
 
+		if input.ResponseFormat.Type == "json_object" {
+			input.Grammar = grammar.JSONBNF
+		}
+
 		log.Debug().Msgf("Parameter Config: %+v", config)
 
 		if input.Stream {
@@ -108,7 +120,9 @@ func CompletionEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fibe
 				}
 
 				resp := &schema.OpenAIResponse{
-					Model: input.Model, // we have to return what the user sent here, due to OpenAI spec.
+					ID:      id,
+					Created: created,
+					Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
 					Choices: []schema.Choice{
 						{
 							Index:        0,
@@ -156,6 +170,8 @@ func CompletionEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fibe
 		}
 
 		resp := &schema.OpenAIResponse{
+			ID:      id,
+			Created: created,
 			Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
 			Choices: result,
 			Object:  "text_completion",

api/openai/edit.go

@@ -3,6 +3,7 @@ package openai
 import (
 	"encoding/json"
 	"fmt"
+	"time"
 
 	"github.com/go-skynet/LocalAI/api/backend"
 	config "github.com/go-skynet/LocalAI/api/config"
@@ -10,6 +11,7 @@ import (
 	"github.com/go-skynet/LocalAI/api/schema"
 	model "github.com/go-skynet/LocalAI/pkg/model"
 	"github.com/gofiber/fiber/v2"
+	"github.com/google/uuid"
 
 	"github.com/rs/zerolog/log"
 )
@@ -62,7 +64,11 @@ func EditEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 			result = append(result, r...)
 		}
 
+		id := uuid.New().String()
+		created := int(time.Now().Unix())
 		resp := &schema.OpenAIResponse{
+			ID:      id,
+			Created: created,
 			Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
 			Choices: result,
 			Object:  "edit",

api/openai/embeddings.go

@@ -3,10 +3,12 @@ package openai
 import (
 	"encoding/json"
 	"fmt"
+	"time"
 
 	"github.com/go-skynet/LocalAI/api/backend"
 	config "github.com/go-skynet/LocalAI/api/config"
 	"github.com/go-skynet/LocalAI/api/schema"
+	"github.com/google/uuid"
 
 	"github.com/go-skynet/LocalAI/api/options"
 	"github.com/gofiber/fiber/v2"
@@ -57,10 +59,14 @@ func EmbeddingsEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fibe
 			items = append(items, schema.Item{Embedding: embeddings, Index: i, Object: "embedding"})
 		}
 
+		id := uuid.New().String()
+		created := int(time.Now().Unix())
 		resp := &schema.OpenAIResponse{
-			Model:  input.Model, // we have to return what the user sent here, due to OpenAI spec.
-			Data:   items,
-			Object: "list",
+			ID:      id,
+			Created: created,
+			Model:   input.Model, // we have to return what the user sent here, due to OpenAI spec.
+			Data:    items,
+			Object:  "list",
 		}
 
 		jsonResult, _ := json.Marshal(resp)

api/openai/image.go

@@ -5,11 +5,14 @@ import (
 	"encoding/base64"
 	"encoding/json"
 	"fmt"
-	"github.com/go-skynet/LocalAI/api/schema"
 	"os"
 	"path/filepath"
 	"strconv"
 	"strings"
+	"time"
+
+	"github.com/go-skynet/LocalAI/api/schema"
+	"github.com/google/uuid"
 
 	"github.com/go-skynet/LocalAI/api/backend"
 	config "github.com/go-skynet/LocalAI/api/config"
@@ -97,7 +100,7 @@ func ImageEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx
 		}
 
 		b64JSON := false
-		if input.ResponseFormat == "b64_json" {
+		if input.ResponseFormat.Type == "b64_json" {
 			b64JSON = true
 		}
 		// src and clip_skip
@@ -174,8 +177,12 @@ func ImageEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx
 			}
 		}
 
+		id := uuid.New().String()
+		created := int(time.Now().Unix())
 		resp := &schema.OpenAIResponse{
-			Data: result,
+			ID:      id,
+			Created: created,
+			Data:    result,
 		}
 
 		jsonResult, _ := json.Marshal(resp)

api/openai/inference.go

@@ -23,8 +23,13 @@ func ComputeChoices(
 		n = 1
 	}
 
+	images := []string{}
+	for _, m := range req.Messages {
+		images = append(images, m.StringImages...)
+	}
+
 	// get the model function to call for the result
-	predFunc, err := backend.ModelInference(req.Context, predInput, loader, *config, o, tokenCallback)
+	predFunc, err := backend.ModelInference(req.Context, predInput, images, loader, *config, o, tokenCallback)
 	if err != nil {
 		return result, backend.TokenUsage{}, err
 	}

api/openai/request.go

@@ -2,8 +2,11 @@ package openai
 
 import (
 	"context"
+	"encoding/base64"
 	"encoding/json"
 	"fmt"
+	"io/ioutil"
+	"net/http"
 	"os"
 	"path/filepath"
 	"strings"
@@ -24,7 +27,7 @@ func readInput(c *fiber.Ctx, o *options.Option, randomModel bool) (string, *sche
 	input.Cancel = cancel
 	// Get input data from the request body
 	if err := c.BodyParser(input); err != nil {
-		return "", nil, err
+		return "", nil, fmt.Errorf("failed parsing request body: %w", err)
 	}
 
 	modelFile := input.Model
@@ -61,6 +64,37 @@ func readInput(c *fiber.Ctx, o *options.Option, randomModel bool) (string, *sche
 	return modelFile, input, nil
 }
 
+// this function check if the string is an URL, if it's an URL downloads the image in memory
+// encodes it in base64 and returns the base64 string
+func getBase64Image(s string) (string, error) {
+	if strings.HasPrefix(s, "http") {
+		// download the image
+		resp, err := http.Get(s)
+		if err != nil {
+			return "", err
+		}
+		defer resp.Body.Close()
+
+		// read the image data into memory
+		data, err := ioutil.ReadAll(resp.Body)
+		if err != nil {
+			return "", err
+		}
+
+		// encode the image data in base64
+		encoded := base64.StdEncoding.EncodeToString(data)
+
+		// return the base64 string
+		return encoded, nil
+	}
+
+	// if the string instead is prefixed with "data:image/jpeg;base64,", drop it
+	if strings.HasPrefix(s, "data:image/jpeg;base64,") {
+		return strings.ReplaceAll(s, "data:image/jpeg;base64,", ""), nil
+	}
+	return "", fmt.Errorf("not valid string")
+}
+
 func updateConfig(config *config.Config, input *schema.OpenAIRequest) {
 	if input.Echo {
 		config.Echo = input.Echo
@@ -129,6 +163,35 @@ func updateConfig(config *config.Config, input *schema.OpenAIRequest) {
 		}
 	}
 
+	// Decode each request's message content
+	index := 0
+	for i, m := range input.Messages {
+		switch content := m.Content.(type) {
+		case string:
+			input.Messages[i].StringContent = content
+		case []interface{}:
+			dat, _ := json.Marshal(content)
+			c := []schema.Content{}
+			json.Unmarshal(dat, &c)
+			for _, pp := range c {
+				if pp.Type == "text" {
+					input.Messages[i].StringContent = pp.Text
+				} else if pp.Type == "image_url" {
+					// Detect if pp.ImageURL is an URL, if it is download the image and encode it in base64:
+					base64, err := getBase64Image(pp.ImageURL.URL)
+					if err == nil {
+						input.Messages[i].StringImages = append(input.Messages[i].StringImages, base64) // TODO: make sure that we only return base64 stuff
+						// set a placeholder for each image
+						input.Messages[i].StringContent = fmt.Sprintf("[img-%d]", index) + input.Messages[i].StringContent
+						index++
+					} else {
+						fmt.Print("Failed encoding image", err)
+					}
+				}
+			}
+		}
+	}
+
 	if input.RepeatPenalty != 0 {
 		config.RepeatPenalty = input.RepeatPenalty
 	}

api/options/options.go

@@ -4,7 +4,9 @@ import (
 	"context"
 	"embed"
 	"encoding/json"
+	"time"
 
+	"github.com/go-skynet/LocalAI/metrics"
 	"github.com/go-skynet/LocalAI/pkg/gallery"
 	model "github.com/go-skynet/LocalAI/pkg/model"
 	"github.com/rs/zerolog/log"
@@ -24,6 +26,7 @@ type Option struct {
 	PreloadModelsFromPath               string
 	CORSAllowOrigins                    string
 	ApiKeys                             []string
+	Metrics                             *metrics.Metrics
 
 	Galleries []gallery.Gallery
 
@@ -34,7 +37,13 @@ type Option struct {
 
 	AutoloadGalleries bool
 
-	SingleBackend bool
+	SingleBackend           bool
+	ParallelBackendRequests bool
+
+	WatchDogIdle                             bool
+	WatchDogBusy                             bool
+	WatchDog                                 bool
+	WatchDogBusyTimeout, WatchDogIdleTimeout time.Duration
 }
 
 type AppOption func(*Option)
@@ -60,10 +69,40 @@ func WithCors(b bool) AppOption {
 	}
 }
 
+var EnableWatchDog = func(o *Option) {
+	o.WatchDog = true
+}
+
+var EnableWatchDogIdleCheck = func(o *Option) {
+	o.WatchDog = true
+	o.WatchDogIdle = true
+}
+
+var EnableWatchDogBusyCheck = func(o *Option) {
+	o.WatchDog = true
+	o.WatchDogBusy = true
+}
+
+func SetWatchDogBusyTimeout(t time.Duration) AppOption {
+	return func(o *Option) {
+		o.WatchDogBusyTimeout = t
+	}
+}
+
+func SetWatchDogIdleTimeout(t time.Duration) AppOption {
+	return func(o *Option) {
+		o.WatchDogIdleTimeout = t
+	}
+}
+
 var EnableSingleBackend = func(o *Option) {
 	o.SingleBackend = true
 }
 
+var EnableParallelBackendRequests = func(o *Option) {
+	o.ParallelBackendRequests = true
+}
+
 var EnableGalleriesAutoload = func(o *Option) {
 	o.AutoloadGalleries = true
 }
@@ -99,6 +138,7 @@ func WithStringGalleries(galls string) AppOption {
 	return func(o *Option) {
 		if galls == "" {
 			log.Debug().Msgf("no galleries to load")
+			o.Galleries = []gallery.Gallery{}
 			return
 		}
 		var galleries []gallery.Gallery
@@ -197,3 +237,9 @@ func WithApiKeys(apiKeys []string) AppOption {
 		o.ApiKeys = apiKeys
 	}
 }
+
+func WithMetrics(meter *metrics.Metrics) AppOption {
+	return func(o *Option) {
+		o.Metrics = meter
+	}
+}

api/schema/openai.go

@@ -55,11 +55,25 @@ type Choice struct {
 	Text         string   `json:"text,omitempty"`
 }
 
+type Content struct {
+	Type     string     `json:"type" yaml:"type"`
+	Text     string     `json:"text" yaml:"text"`
+	ImageURL ContentURL `json:"image_url" yaml:"image_url"`
+}
+
+type ContentURL struct {
+	URL string `json:"url" yaml:"url"`
+}
+
 type Message struct {
 	// The message role
 	Role string `json:"role,omitempty" yaml:"role"`
 	// The message content
-	Content *string `json:"content" yaml:"content"`
+	Content interface{} `json:"content" yaml:"content"`
+
+	StringContent string   `json:"string_content,omitempty" yaml:"string_content,omitempty"`
+	StringImages  []string `json:"string_images,omitempty" yaml:"string_images,omitempty"`
+
 	// A result of a function call
 	FunctionCall interface{} `json:"function_call,omitempty" yaml:"function_call,omitempty"`
 }
@@ -69,6 +83,12 @@ type OpenAIModel struct {
 	Object string `json:"object"`
 }
 
+type ChatCompletionResponseFormatType string
+
+type ChatCompletionResponseFormat struct {
+	Type ChatCompletionResponseFormatType `json:"type,omitempty"`
+}
+
 type OpenAIRequest struct {
 	config.PredictionOptions
 
@@ -78,7 +98,7 @@ type OpenAIRequest struct {
 	// whisper
 	File string `json:"file" validate:"required"`
 	//whisper/image
-	ResponseFormat string `json:"response_format"`
+	ResponseFormat ChatCompletionResponseFormat `json:"response_format"`
 	// image
 	Size string `json:"size"`
 	// Prompt is read only by completion/image API calls

backend/backend.proto

@@ -63,6 +63,8 @@ message PredictOptions {
   float RopeFreqScale = 38;
   float NegativePromptScale = 39;
   string NegativePrompt = 40;
+  int32 NDraft = 41;
+  repeated string Images = 42;
 }
 
 // The response message containing the result
@@ -115,7 +117,23 @@ message ModelOptions {
   // LLM (llama.cpp)
   string LoraBase = 35;
   string LoraAdapter = 36;
+  float LoraScale = 42;
+
   bool NoMulMatQ = 37;
+  string DraftModel = 39;
+  
+  string AudioPath = 38;
+
+  // vllm
+  string Quantization = 40;
+
+  string MMProj = 41;
+
+  string RopeScaling = 43;
+  float YarnExtFactor = 44;
+  float YarnAttnFactor = 45;
+  float YarnBetaFast = 46;
+  float YarnBetaSlow = 47;
 }
 
 message Result {

backend/cpp/grpc/.gitignore

@@ -0,0 +1,3 @@
+installed_packages/
+grpc_build/
+grpc_repo/

backend/cpp/grpc/script/build_grpc.sh

@@ -0,0 +1,81 @@
+#!/bin/bash
+
+# Builds locally from sources the packages needed by the llama cpp backend.
+
+# Makes sure a few base packages exist.
+# sudo apt-get --no-upgrade -y install g++ gcc binutils cmake git build-essential autoconf libtool pkg-config 
+
+SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
+echo "Script directory: $SCRIPT_DIR"
+
+CPP_INSTALLED_PACKAGES_DIR=$1
+if [ -z ${CPP_INSTALLED_PACKAGES_DIR} ]; then 
+    echo "CPP_INSTALLED_PACKAGES_DIR env variable not set. Don't know where to install: failed."; 
+    echo
+    exit -1
+fi
+
+if [ -d "${CPP_INSTALLED_PACKAGES_DIR}" ]; then
+  echo "gRPC installation directory already exists. Nothing to do."
+  exit 0
+fi
+
+# The depth when cloning a git repo. 1 speeds up the clone when the repo history is not needed.
+GIT_CLONE_DEPTH=1
+
+NUM_BUILD_THREADS=$(nproc --ignore=1)
+
+# Google gRPC --------------------------------------------------------------------------------------
+TAG_LIB_GRPC="v1.59.0"
+GIT_REPO_LIB_GRPC="https://github.com/grpc/grpc.git"
+GRPC_REPO_DIR="${SCRIPT_DIR}/../grpc_repo"
+GRPC_BUILD_DIR="${SCRIPT_DIR}/../grpc_build"
+SRC_DIR_LIB_GRPC="${GRPC_REPO_DIR}/grpc"
+
+echo "SRC_DIR_LIB_GRPC: ${SRC_DIR_LIB_GRPC}"
+echo "GRPC_REPO_DIR: ${GRPC_REPO_DIR}"
+echo "GRPC_BUILD_DIR: ${GRPC_BUILD_DIR}"
+
+mkdir -pv ${GRPC_REPO_DIR}
+
+rm   -rf ${GRPC_BUILD_DIR}
+mkdir -pv ${GRPC_BUILD_DIR}
+
+mkdir -pv ${CPP_INSTALLED_PACKAGES_DIR}
+	
+if [ -d "${SRC_DIR_LIB_GRPC}" ]; then
+  echo "gRPC source already exists locally. Not cloned again."
+else  
+  ( cd ${GRPC_REPO_DIR} && \
+    git clone --depth ${GIT_CLONE_DEPTH} -b ${TAG_LIB_GRPC} ${GIT_REPO_LIB_GRPC} && \
+    cd ${SRC_DIR_LIB_GRPC} && \
+    git submodule update --init --recursive --depth ${GIT_CLONE_DEPTH} 
+  )    
+fi
+
+( cd ${GRPC_BUILD_DIR} && \
+  cmake -G "Unix Makefiles" \
+     -DCMAKE_BUILD_TYPE=Release \
+     -DgRPC_INSTALL=ON \
+     -DEXECUTABLE_OUTPUT_PATH=${CPP_INSTALLED_PACKAGES_DIR}/grpc/bin \
+     -DLIBRARY_OUTPUT_PATH=${CPP_INSTALLED_PACKAGES_DIR}/grpc/lib \
+     -DgRPC_BUILD_TESTS=OFF \
+     -DgRPC_BUILD_CSHARP_EXT=OFF \
+     -DgRPC_BUILD_GRPC_CPP_PLUGIN=ON \
+     -DgRPC_BUILD_GRPC_CSHARP_PLUGIN=OFF \
+     -DgRPC_BUILD_GRPC_NODE_PLUGIN=OFF \
+     -DgRPC_BUILD_GRPC_OBJECTIVE_C_PLUGIN=OFF \
+     -DgRPC_BUILD_GRPC_PHP_PLUGIN=OFF \
+     -DgRPC_BUILD_GRPC_PYTHON_PLUGIN=ON \
+     -DgRPC_BUILD_GRPC_RUBY_PLUGIN=OFF \
+     -Dprotobuf_WITH_ZLIB=ON \
+     -DRE2_BUILD_TESTING=OFF \
+     -DCMAKE_INSTALL_PREFIX=${CPP_INSTALLED_PACKAGES_DIR}/ \
+     ${SRC_DIR_LIB_GRPC}  && \
+  cmake --build .  -- -j ${NUM_BUILD_THREADS} && \
+  cmake --build .  --target install -- -j ${NUM_BUILD_THREADS} 
+)
+
+rm -rf ${GRPC_BUILD_DIR}
+rm -rf ${GRPC_REPO_DIR}
+

backend/cpp/llama/CMakeLists.txt

@@ -0,0 +1,74 @@
+
+## 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)
+install(TARGETS ${TARGET} LIBRARY)
+target_link_libraries(${TARGET} PRIVATE common ggml ${CMAKE_THREAD_LIBS_INIT})
+target_compile_features(${TARGET} PRIVATE cxx_std_11)
+if (NOT MSVC)
+    target_compile_options(${TARGET} PRIVATE -Wno-cast-qual) # stb_image.h
+endif()
+
+set(TARGET grpc-server)
+# END CLIP hack
+set(CMAKE_CXX_STANDARD 17)
+cmake_minimum_required(VERSION 3.15)
+set(TARGET grpc-server)
+set(_PROTOBUF_LIBPROTOBUF libprotobuf)
+set(_REFLECTION grpc++_reflection)
+if (${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
+    link_directories("/opt/homebrew/lib")
+    include_directories("/opt/homebrew/include")
+endif()
+
+find_package(absl CONFIG REQUIRED)
+find_package(Protobuf CONFIG REQUIRED)
+find_package(gRPC CONFIG REQUIRED)
+
+find_program(_PROTOBUF_PROTOC protoc)
+set(_GRPC_GRPCPP grpc++)
+find_program(_GRPC_CPP_PLUGIN_EXECUTABLE grpc_cpp_plugin)
+
+include_directories(${CMAKE_CURRENT_BINARY_DIR})
+include_directories(${Protobuf_INCLUDE_DIRS})
+
+message(STATUS "Using protobuf version ${Protobuf_VERSION} | Protobuf_INCLUDE_DIRS: ${Protobuf_INCLUDE_DIRS} | CMAKE_CURRENT_BINARY_DIR: ${CMAKE_CURRENT_BINARY_DIR}")
+
+# Proto file
+get_filename_component(hw_proto "../../../../../../backend/backend.proto" ABSOLUTE)
+get_filename_component(hw_proto_path "${hw_proto}" PATH)
+
+# Generated sources
+set(hw_proto_srcs "${CMAKE_CURRENT_BINARY_DIR}/backend.pb.cc")
+set(hw_proto_hdrs "${CMAKE_CURRENT_BINARY_DIR}/backend.pb.h")
+set(hw_grpc_srcs "${CMAKE_CURRENT_BINARY_DIR}/backend.grpc.pb.cc")
+set(hw_grpc_hdrs "${CMAKE_CURRENT_BINARY_DIR}/backend.grpc.pb.h")
+
+add_custom_command(
+      OUTPUT "${hw_proto_srcs}" "${hw_proto_hdrs}" "${hw_grpc_srcs}" "${hw_grpc_hdrs}"
+      COMMAND ${_PROTOBUF_PROTOC}
+      ARGS --grpc_out "${CMAKE_CURRENT_BINARY_DIR}"
+        --cpp_out "${CMAKE_CURRENT_BINARY_DIR}"
+        -I "${hw_proto_path}"
+        --plugin=protoc-gen-grpc="${_GRPC_CPP_PLUGIN_EXECUTABLE}"
+        "${hw_proto}"
+      DEPENDS "${hw_proto}")
+
+# hw_grpc_proto
+add_library(hw_grpc_proto
+  ${hw_grpc_srcs}
+  ${hw_grpc_hdrs}
+  ${hw_proto_srcs}
+  ${hw_proto_hdrs} )
+
+add_executable(${TARGET} grpc-server.cpp json.hpp )
+target_link_libraries(${TARGET} PRIVATE common llama myclip ${CMAKE_THREAD_LIBS_INIT} absl::flags hw_grpc_proto
+  absl::flags_parse
+  gRPC::${_REFLECTION}
+  gRPC::${_GRPC_GRPCPP}
+  protobuf::${_PROTOBUF_LIBPROTOBUF})
+target_compile_features(${TARGET} PRIVATE cxx_std_11)
+if(TARGET BUILD_INFO)
+  add_dependencies(${TARGET} BUILD_INFO)
+endif()

backend/cpp/llama/Makefile

@@ -0,0 +1,50 @@
+
+LLAMA_VERSION?=d9b33fe95bd257b36c84ee5769cc048230067d6f
+
+CMAKE_ARGS?=
+BUILD_TYPE?=
+
+# If build type is cublas, then we set -DLLAMA_CUBLAS=ON to CMAKE_ARGS automatically
+ifeq ($(BUILD_TYPE),cublas)
+	CMAKE_ARGS+=-DLLAMA_CUBLAS=ON
+# If build type is openblas then we set -DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS
+# to CMAKE_ARGS automatically
+else ifeq ($(BUILD_TYPE),openblas)
+	CMAKE_ARGS+=-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS
+# If build type is clblast (openCL) we set -DLLAMA_CLBLAST=ON -DCLBlast_DIR=/some/path
+else ifeq ($(BUILD_TYPE),clblast)
+	CMAKE_ARGS+=-DLLAMA_CLBLAST=ON -DCLBlast_DIR=/some/path
+# If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++ 
+else ifeq ($(BUILD_TYPE),hipblas)
+	CMAKE_ARGS+=-DLLAMA_HIPBLAS=ON
+endif
+
+llama.cpp:
+	git clone --recurse-submodules https://github.com/ggerganov/llama.cpp llama.cpp
+	cd llama.cpp && git checkout -b build $(LLAMA_VERSION) && git submodule update --init --recursive --depth 1
+
+llama.cpp/examples/grpc-server:
+	mkdir -p llama.cpp/examples/grpc-server
+	cp -r $(abspath ./)/CMakeLists.txt llama.cpp/examples/grpc-server/
+	cp -r $(abspath ./)/grpc-server.cpp llama.cpp/examples/grpc-server/
+	cp -rfv $(abspath ./)/json.hpp llama.cpp/examples/grpc-server/
+	echo "add_subdirectory(grpc-server)" >> llama.cpp/examples/CMakeLists.txt
+## 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/examples/llava/clip.h llama.cpp/examples/grpc-server/clip.h
+	cp -rfv llama.cpp/examples/llava/clip.cpp llama.cpp/examples/grpc-server/clip.cpp
+
+rebuild:
+	cp -rfv $(abspath ./)/CMakeLists.txt llama.cpp/examples/grpc-server/
+	cp -rfv $(abspath ./)/grpc-server.cpp llama.cpp/examples/grpc-server/
+	cp -rfv $(abspath ./)/json.hpp llama.cpp/examples/grpc-server/
+	rm -rf grpc-server
+	$(MAKE) grpc-server
+
+clean:
+	rm -rf llama.cpp
+	rm -rf grpc-server
+
+grpc-server: llama.cpp llama.cpp/examples/grpc-server
+	cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && cmake --build . --config Release
+	cp llama.cpp/build/bin/grpc-server .

backend/go/image/main.go

@@ -5,7 +5,6 @@ package main
 import (
 	"flag"
 
-	bert "github.com/go-skynet/LocalAI/pkg/backend/llm/bert"
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )
 
@@ -16,7 +15,7 @@ var (
 func main() {
 	flag.Parse()
 
-	if err := grpc.StartServer(*addr, &bert.Embeddings{}); err != nil {
+	if err := grpc.StartServer(*addr, &StableDiffusion{}); err != nil {
 		panic(err)
 	}
 }

backend/go/image/stablediffusion.go

@@ -1,4 +1,4 @@
-package image
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)

backend/go/llm/bert/bert.go

@@ -1,4 +1,4 @@
-package bert
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)

backend/go/llm/bert/main.go

@@ -5,8 +5,6 @@ package main
 import (
 	"flag"
 
-	rwkv "github.com/go-skynet/LocalAI/pkg/backend/llm/rwkv"
-
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )
 
@@ -17,7 +15,7 @@ var (
 func main() {
 	flag.Parse()
 
-	if err := grpc.StartServer(*addr, &rwkv.LLM{}); err != nil {
+	if err := grpc.StartServer(*addr, &Embeddings{}); err != nil {
 		panic(err)
 	}
 }

backend/go/llm/dolly/main.go

@@ -5,7 +5,7 @@ package main
 import (
 	"flag"
 
-	transformers "github.com/go-skynet/LocalAI/pkg/backend/llm/transformers"
+	transformers "github.com/go-skynet/LocalAI/backend/go/llm/transformers"
 
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )

backend/go/llm/falcon-ggml/main.go

@@ -5,7 +5,7 @@ package main
 import (
 	"flag"
 
-	transformers "github.com/go-skynet/LocalAI/pkg/backend/llm/transformers"
+	transformers "github.com/go-skynet/LocalAI/backend/go/llm/transformers"
 
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )

backend/go/llm/gpt2/main.go

@@ -5,7 +5,7 @@ package main
 import (
 	"flag"
 
-	transformers "github.com/go-skynet/LocalAI/pkg/backend/llm/transformers"
+	transformers "github.com/go-skynet/LocalAI/backend/go/llm/transformers"
 
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )

backend/go/llm/gpt4all/gpt4all.go

@@ -1,4 +1,4 @@
-package gpt4all
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)

backend/go/llm/gpt4all/main.go

@@ -5,8 +5,6 @@ package main
 import (
 	"flag"
 
-	bloomz "github.com/go-skynet/LocalAI/pkg/backend/llm/bloomz"
-
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )
 
@@ -17,7 +15,7 @@ var (
 func main() {
 	flag.Parse()
 
-	if err := grpc.StartServer(*addr, &bloomz.LLM{}); err != nil {
+	if err := grpc.StartServer(*addr, &LLM{}); err != nil {
 		panic(err)
 	}
 }

backend/go/llm/gptj/main.go

@@ -5,7 +5,7 @@ package main
 import (
 	"flag"
 
-	transformers "github.com/go-skynet/LocalAI/pkg/backend/llm/transformers"
+	transformers "github.com/go-skynet/LocalAI/backend/go/llm/transformers"
 
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )

backend/go/llm/gptneox/main.go

@@ -5,7 +5,7 @@ package main
 import (
 	"flag"
 
-	transformers "github.com/go-skynet/LocalAI/pkg/backend/llm/transformers"
+	transformers "github.com/go-skynet/LocalAI/backend/go/llm/transformers"
 
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )

backend/go/llm/langchain/langchain.go

@@ -1,4 +1,4 @@
-package langchain
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)

backend/go/llm/langchain/main.go

@@ -0,0 +1,21 @@
+package main
+
+// Note: this is started internally by LocalAI and a server is allocated for each model
+
+import (
+	"flag"
+
+	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
+)
+
+var (
+	addr = flag.String("addr", "localhost:50051", "the address to connect to")
+)
+
+func main() {
+	flag.Parse()
+
+	if err := grpc.StartServer(*addr, &LLM{}); err != nil {
+		panic(err)
+	}
+}

backend/go/llm/llama-ggml/llama.go

@@ -1,4 +1,4 @@
-package llama
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)

backend/go/llm/llama-ggml/main.go

@@ -3,8 +3,6 @@ package main
 import (
 	"flag"
 
-	llama "github.com/go-skynet/LocalAI/pkg/backend/llm/llama-stable"
-
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )
 
@@ -15,7 +13,7 @@ var (
 func main() {
 	flag.Parse()
 
-	if err := grpc.StartServer(*addr, &llama.LLM{}); err != nil {
+	if err := grpc.StartServer(*addr, &LLM{}); err != nil {
 		panic(err)
 	}
 }

backend/go/llm/llama/llama.go

@@ -1,9 +1,10 @@
-package llama
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)
 import (
 	"fmt"
+	"path/filepath"
 
 	"github.com/go-skynet/LocalAI/pkg/grpc/base"
 	pb "github.com/go-skynet/LocalAI/pkg/grpc/proto"
@@ -13,7 +14,8 @@ import (
 type LLM struct {
 	base.SingleThread
 
-	llama *llama.LLama
+	llama      *llama.LLama
+	draftModel *llama.LLama
 }
 
 func (llm *LLM) Load(opts *pb.ModelOptions) error {
@@ -36,12 +38,15 @@ func (llm *LLM) Load(opts *pb.ModelOptions) error {
 		llamaOpts = append(llamaOpts, llama.SetMulMatQ(false))
 	}
 
+	// Get base path of opts.ModelFile and use the same for lora (assume the same path)
+	basePath := filepath.Dir(opts.ModelFile)
+
 	if opts.LoraAdapter != "" {
-		llamaOpts = append(llamaOpts, llama.SetLoraAdapter(opts.LoraAdapter))
+		llamaOpts = append(llamaOpts, llama.SetLoraAdapter(filepath.Join(basePath, opts.LoraAdapter)))
 	}
 
 	if opts.LoraBase != "" {
-		llamaOpts = append(llamaOpts, llama.SetLoraBase(opts.LoraBase))
+		llamaOpts = append(llamaOpts, llama.SetLoraBase(filepath.Join(basePath, opts.LoraBase)))
 	}
 
 	if opts.ContextSize != 0 {
@@ -74,7 +79,27 @@ func (llm *LLM) Load(opts *pb.ModelOptions) error {
 		llamaOpts = append(llamaOpts, llama.EnabelLowVRAM)
 	}
 
+	if opts.DraftModel != "" {
+		// https://github.com/ggerganov/llama.cpp/blob/71ca2fad7d6c0ef95ef9944fb3a1a843e481f314/examples/speculative/speculative.cpp#L40
+		llamaOpts = append(llamaOpts, llama.SetPerplexity(true))
+	}
+
 	model, err := llama.New(opts.ModelFile, llamaOpts...)
+
+	if opts.DraftModel != "" {
+		// opts.DraftModel is relative to opts.ModelFile, so we need to get the basepath of opts.ModelFile
+		if !filepath.IsAbs(opts.DraftModel) {
+			dir := filepath.Dir(opts.ModelFile)
+			opts.DraftModel = filepath.Join(dir, opts.DraftModel)
+		}
+
+		draftModel, err := llama.New(opts.DraftModel, llamaOpts...)
+		if err != nil {
+			return err
+		}
+		llm.draftModel = draftModel
+	}
+
 	llm.llama = model
 
 	return err
@@ -158,6 +183,9 @@ func buildPredictOptions(opts *pb.PredictOptions) []llama.PredictOption {
 		predictOptions = append(predictOptions, llama.SetSeed(int(opts.Seed)))
 	}
 
+	if opts.NDraft != 0 {
+		predictOptions = append(predictOptions, llama.SetNDraft(int(opts.NDraft)))
+	}
 	//predictOptions = append(predictOptions, llama.SetLogitBias(c.Seed))
 
 	predictOptions = append(predictOptions, llama.SetFrequencyPenalty(opts.FrequencyPenalty))
@@ -171,6 +199,9 @@ func buildPredictOptions(opts *pb.PredictOptions) []llama.PredictOption {
 }
 
 func (llm *LLM) Predict(opts *pb.PredictOptions) (string, error) {
+	if llm.draftModel != nil {
+		return llm.llama.SpeculativeSampling(llm.draftModel, opts.Prompt, buildPredictOptions(opts)...)
+	}
 	return llm.llama.Predict(opts.Prompt, buildPredictOptions(opts)...)
 }
 
@@ -183,7 +214,13 @@ func (llm *LLM) PredictStream(opts *pb.PredictOptions, results chan string) erro
 	}))
 
 	go func() {
-		_, err := llm.llama.Predict(opts.Prompt, predictOptions...)
+		var err error
+		if llm.draftModel != nil {
+			_, err = llm.llama.SpeculativeSampling(llm.draftModel, opts.Prompt, buildPredictOptions(opts)...)
+		} else {
+			_, err = llm.llama.Predict(opts.Prompt, predictOptions...)
+		}
+
 		if err != nil {
 			fmt.Println("err: ", err)
 		}

backend/go/llm/llama/main.go

@@ -1,12 +1,12 @@
 package main
 
+// GRPC Falcon server
+
 // Note: this is started internally by LocalAI and a server is allocated for each model
 
 import (
 	"flag"
 
-	tts "github.com/go-skynet/LocalAI/pkg/backend/tts"
-
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )
 
@@ -17,7 +17,7 @@ var (
 func main() {
 	flag.Parse()
 
-	if err := grpc.StartServer(*addr, &tts.Piper{}); err != nil {
+	if err := grpc.StartServer(*addr, &LLM{}); err != nil {
 		panic(err)
 	}
 }

backend/go/llm/mpt/main.go

@@ -5,7 +5,7 @@ package main
 import (
 	"flag"
 
-	transformers "github.com/go-skynet/LocalAI/pkg/backend/llm/transformers"
+	transformers "github.com/go-skynet/LocalAI/backend/go/llm/transformers"
 
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )

backend/go/llm/replit/main.go

@@ -5,7 +5,7 @@ package main
 import (
 	"flag"
 
-	transformers "github.com/go-skynet/LocalAI/pkg/backend/llm/transformers"
+	transformers "github.com/go-skynet/LocalAI/backend/go/llm/transformers"
 
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )

backend/go/llm/rwkv/main.go

@@ -0,0 +1,21 @@
+package main
+
+// Note: this is started internally by LocalAI and a server is allocated for each model
+
+import (
+	"flag"
+
+	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
+)
+
+var (
+	addr = flag.String("addr", "localhost:50051", "the address to connect to")
+)
+
+func main() {
+	flag.Parse()
+
+	if err := grpc.StartServer(*addr, &LLM{}); err != nil {
+		panic(err)
+	}
+}

backend/go/llm/rwkv/rwkv.go

@@ -1,4 +1,4 @@
-package rwkv
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)

backend/go/llm/starcoder/main.go

@@ -5,7 +5,7 @@ package main
 import (
 	"flag"
 
-	transformers "github.com/go-skynet/LocalAI/pkg/backend/llm/transformers"
+	transformers "github.com/go-skynet/LocalAI/backend/go/llm/transformers"
 
 	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
 )

backend/go/transcribe/main.go

@@ -0,0 +1,21 @@
+package main
+
+// Note: this is started internally by LocalAI and a server is allocated for each model
+
+import (
+	"flag"
+
+	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
+)
+
+var (
+	addr = flag.String("addr", "localhost:50051", "the address to connect to")
+)
+
+func main() {
+	flag.Parse()
+
+	if err := grpc.StartServer(*addr, &Whisper{}); err != nil {
+		panic(err)
+	}
+}

backend/go/transcribe/transcript.go

@@ -1,4 +1,4 @@
-package transcribe
+package main
 
 import (
 	"fmt"

backend/go/transcribe/whisper.go

@@ -1,4 +1,4 @@
-package transcribe
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)

backend/go/tts/main.go

@@ -0,0 +1,21 @@
+package main
+
+// Note: this is started internally by LocalAI and a server is allocated for each model
+
+import (
+	"flag"
+
+	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
+)
+
+var (
+	addr = flag.String("addr", "localhost:50051", "the address to connect to")
+)
+
+func main() {
+	flag.Parse()
+
+	if err := grpc.StartServer(*addr, &Piper{}); err != nil {
+		panic(err)
+	}
+}

backend/go/tts/piper.go

@@ -1,4 +1,4 @@
-package tts
+package main
 
 // This is a wrapper to statisfy the GRPC service interface
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)

backend/python/README.md

@@ -0,0 +1,38 @@
+# Common commands about conda environment
+
+## Create a new empty conda environment
+
+```
+conda create --name <env-name> python=<your version> -y
+
+conda create --name autogptq python=3.11 -y
+```
+
+## To activate the environment
+
+As of conda 4.4
+```
+conda activate autogptq
+```
+
+The conda version older than 4.4
+
+```
+source activate autogptq
+```
+
+## Install the packages to your environment
+
+Sometimes you need to install the packages from the conda-forge channel
+
+By using `conda`
+```
+conda install <your-package-name>
+
+conda install -c conda-forge <your package-name>
+```
+
+Or by using `pip`
+```
+pip install <your-package-name>
+```

backend/python/autogptq/Makefile

@@ -0,0 +1,5 @@
+.PHONY: autogptq
+autogptq:
+	@echo "Creating virtual environment..."
+	@conda env create --name autogptq --file autogptq.yml
+	@echo "Virtual environment created."

backend/python/autogptq/README.md

@@ -0,0 +1,5 @@
+# Creating a separate environment for the autogptq project
+
+```
+make autogptq
+```

backend/python/autogptq/autogptq.py

@@ -1,20 +1,23 @@
 #!/usr/bin/env python3
-import grpc
 from concurrent import futures
-import time
-import backend_pb2
-import backend_pb2_grpc
 import argparse
 import signal
 import sys
 import os
-from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
-from pathlib import Path
+import time
+
+import grpc
+import backend_pb2
+import backend_pb2_grpc
+from auto_gptq import AutoGPTQForCausalLM
 from transformers import AutoTokenizer
 from transformers import TextGenerationPipeline
 
 _ONE_DAY_IN_SECONDS = 60 * 60 * 24
 
+# If MAX_WORKERS are specified in the environment use it, otherwise default to 1
+MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
+
 # Implement the BackendServicer class with the service methods
 class BackendServicer(backend_pb2_grpc.BackendServicer):
     def Health(self, request, context):
@@ -77,7 +80,7 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
 
 
 def serve(address):
-    server = grpc.server(futures.ThreadPoolExecutor(max_workers=1))
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS))
     backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
     server.add_insecure_port(address)
     server.start()

backend/python/autogptq/autogptq.yml

@@ -0,0 +1,86 @@
+name: autogptq
+channels:
+  - defaults
+dependencies:
+  - _libgcc_mutex=0.1=main
+  - _openmp_mutex=5.1=1_gnu
+  - bzip2=1.0.8=h7b6447c_0
+  - ca-certificates=2023.08.22=h06a4308_0
+  - ld_impl_linux-64=2.38=h1181459_1
+  - libffi=3.4.4=h6a678d5_0
+  - libgcc-ng=11.2.0=h1234567_1
+  - libgomp=11.2.0=h1234567_1
+  - libstdcxx-ng=11.2.0=h1234567_1
+  - libuuid=1.41.5=h5eee18b_0
+  - ncurses=6.4=h6a678d5_0
+  - openssl=3.0.11=h7f8727e_2
+  - pip=23.2.1=py311h06a4308_0
+  - python=3.11.5=h955ad1f_0
+  - readline=8.2=h5eee18b_0
+  - setuptools=68.0.0=py311h06a4308_0
+  - sqlite=3.41.2=h5eee18b_0
+  - tk=8.6.12=h1ccaba5_0
+  - wheel=0.41.2=py311h06a4308_0
+  - xz=5.4.2=h5eee18b_0
+  - zlib=1.2.13=h5eee18b_0
+  - pip:
+      - accelerate==0.23.0
+      - aiohttp==3.8.5
+      - aiosignal==1.3.1
+      - async-timeout==4.0.3
+      - attrs==23.1.0
+      - auto-gptq==0.4.2
+      - certifi==2023.7.22
+      - charset-normalizer==3.3.0
+      - datasets==2.14.5
+      - dill==0.3.7
+      - filelock==3.12.4
+      - frozenlist==1.4.0
+      - fsspec==2023.6.0
+      - grpcio==1.59.0
+      - huggingface-hub==0.16.4
+      - idna==3.4
+      - jinja2==3.1.2
+      - markupsafe==2.1.3
+      - mpmath==1.3.0
+      - multidict==6.0.4
+      - multiprocess==0.70.15
+      - networkx==3.1
+      - numpy==1.26.0
+      - nvidia-cublas-cu12==12.1.3.1
+      - nvidia-cuda-cupti-cu12==12.1.105
+      - nvidia-cuda-nvrtc-cu12==12.1.105
+      - nvidia-cuda-runtime-cu12==12.1.105
+      - nvidia-cudnn-cu12==8.9.2.26
+      - nvidia-cufft-cu12==11.0.2.54
+      - nvidia-curand-cu12==10.3.2.106
+      - nvidia-cusolver-cu12==11.4.5.107
+      - nvidia-cusparse-cu12==12.1.0.106
+      - nvidia-nccl-cu12==2.18.1
+      - nvidia-nvjitlink-cu12==12.2.140
+      - nvidia-nvtx-cu12==12.1.105
+      - packaging==23.2
+      - pandas==2.1.1
+      - peft==0.5.0
+      - protobuf==4.24.4
+      - psutil==5.9.5
+      - pyarrow==13.0.0
+      - python-dateutil==2.8.2
+      - pytz==2023.3.post1
+      - pyyaml==6.0.1
+      - regex==2023.10.3
+      - requests==2.31.0
+      - rouge==1.0.1
+      - safetensors==0.3.3
+      - six==1.16.0
+      - sympy==1.12
+      - tokenizers==0.14.0
+      - torch==2.1.0
+      - tqdm==4.66.1
+      - transformers==4.34.0
+      - triton==2.1.0
+      - typing-extensions==4.8.0
+      - tzdata==2023.3
+      - urllib3==2.0.6
+      - xxhash==3.4.1
+      - yarl==1.9.2

backend/python/autogptq/run.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+
+##
+## A bash script wrapper that runs the autogptq server with conda
+
+export PATH=$PATH:/opt/conda/bin
+
+# Activate conda environment
+source activate autogptq
+
+# get the directory where the bash script is located
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+python $DIR/autogptq.py $@

backend/python/bark/Makefile

@@ -0,0 +1,5 @@
+.PHONY: ttsbark
+ttsbark:
+	@echo "Creating virtual environment..."
+	@conda env create --name ttsbark --file ttsbark.yml
+	@echo "Virtual environment created."

backend/python/bark/README.md

@@ -0,0 +1,16 @@
+# Creating a separate environment for ttsbark project
+
+```
+make ttsbark
+```
+
+# Testing the gRPC server
+
+```
+<The path of your python interpreter> -m unittest test_ttsbark.py
+```
+
+For example
+```
+/opt/conda/envs/bark/bin/python -m unittest extra/grpc/bark/test_ttsbark.py
+``````

backend/python/bark/run.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+
+##
+## A bash script wrapper that runs the ttsbark server with conda
+
+export PATH=$PATH:/opt/conda/bin
+
+# Activate conda environment
+source activate ttsbark
+
+# get the directory where the bash script is located
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+python $DIR/ttsbark.py $@

backend/python/bark/test_ttsbark.py

@@ -0,0 +1,32 @@
+import unittest
+import subprocess
+import time
+import backend_pb2
+import backend_pb2_grpc
+
+import grpc
+
+class TestBackendServicer(unittest.TestCase):
+    """
+    TestBackendServicer is the class that tests the gRPC service
+    """
+    def setUp(self):
+        self.service = subprocess.Popen(["python3", "ttsbark.py", "--addr", "localhost:50051"])
+
+    def tearDown(self) -> None:
+        self.service.terminate()
+        self.service.wait()
+
+    def test_server_startup(self):
+        time.sleep(2)
+        try:
+            self.setUp()
+            with grpc.insecure_channel("localhost:50051") as channel:
+                stub = backend_pb2_grpc.BackendStub(channel)
+                response = stub.Health(backend_pb2.HealthMessage())
+                self.assertEqual(response.message, b'OK')
+        except Exception as err:
+            print(err)
+            self.fail("Server failed to start")
+        finally:
+            self.tearDown()

backend/python/bark/ttsbark.py

@@ -1,22 +1,32 @@
 #!/usr/bin/env python3
-import grpc
+"""
+This is an extra gRPC server of LocalAI for Bark TTS
+"""
 from concurrent import futures
 import time
-import backend_pb2
-import backend_pb2_grpc
 import argparse
 import signal
 import sys
 import os
-from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
-from pathlib import Path
-from bark import SAMPLE_RATE, generate_audio, preload_models
 from scipy.io.wavfile import write as write_wav
 
+import backend_pb2
+import backend_pb2_grpc
+from bark import SAMPLE_RATE, generate_audio, preload_models
+
+import grpc
+
+
 _ONE_DAY_IN_SECONDS = 60 * 60 * 24
 
+# If MAX_WORKERS are specified in the environment use it, otherwise default to 1
+MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
+
 # Implement the BackendServicer class with the service methods
 class BackendServicer(backend_pb2_grpc.BackendServicer):
+    """
+    BackendServicer is the class that implements the gRPC service
+    """
     def Health(self, request, context):
         return backend_pb2.Reply(message=bytes("OK", 'utf-8'))
     def LoadModel(self, request, context):
@@ -51,7 +61,7 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
         return backend_pb2.Result(success=True)
 
 def serve(address):
-    server = grpc.server(futures.ThreadPoolExecutor(max_workers=1))
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS))
     backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
     server.add_insecure_port(address)
     server.start()
@@ -80,4 +90,4 @@ if __name__ == "__main__":
     )
     args = parser.parse_args()
 
-    serve(args.addr)
+    serve(args.addr)

backend/python/bark/ttsbark.yml

@@ -0,0 +1,96 @@
+name: bark
+channels:
+  - defaults
+dependencies:
+  - _libgcc_mutex=0.1=main
+  - _openmp_mutex=5.1=1_gnu
+  - bzip2=1.0.8=h7b6447c_0
+  - ca-certificates=2023.08.22=h06a4308_0
+  - ld_impl_linux-64=2.38=h1181459_1
+  - libffi=3.4.4=h6a678d5_0
+  - libgcc-ng=11.2.0=h1234567_1
+  - libgomp=11.2.0=h1234567_1
+  - libstdcxx-ng=11.2.0=h1234567_1
+  - libuuid=1.41.5=h5eee18b_0
+  - ncurses=6.4=h6a678d5_0
+  - openssl=3.0.11=h7f8727e_2
+  - pip=23.2.1=py311h06a4308_0
+  - python=3.11.5=h955ad1f_0
+  - readline=8.2=h5eee18b_0
+  - setuptools=68.0.0=py311h06a4308_0
+  - sqlite=3.41.2=h5eee18b_0
+  - tk=8.6.12=h1ccaba5_0
+  - wheel=0.41.2=py311h06a4308_0
+  - xz=5.4.2=h5eee18b_0
+  - zlib=1.2.13=h5eee18b_0
+  - pip:
+      - accelerate==0.23.0
+      - aiohttp==3.8.5
+      - aiosignal==1.3.1
+      - async-timeout==4.0.3
+      - attrs==23.1.0
+      - bark==0.1.5
+      - boto3==1.28.61
+      - botocore==1.31.61
+      - certifi==2023.7.22
+      - charset-normalizer==3.3.0
+      - datasets==2.14.5
+      - dill==0.3.7
+      - einops==0.7.0
+      - encodec==0.1.1
+      - filelock==3.12.4
+      - frozenlist==1.4.0
+      - fsspec==2023.6.0
+      - funcy==2.0
+      - grpcio==1.59.0
+      - huggingface-hub==0.16.4
+      - idna==3.4
+      - jinja2==3.1.2
+      - jmespath==1.0.1
+      - markupsafe==2.1.3
+      - mpmath==1.3.0
+      - multidict==6.0.4
+      - multiprocess==0.70.15
+      - networkx==3.1
+      - numpy==1.26.0
+      - nvidia-cublas-cu12==12.1.3.1
+      - nvidia-cuda-cupti-cu12==12.1.105
+      - nvidia-cuda-nvrtc-cu12==12.1.105
+      - nvidia-cuda-runtime-cu12==12.1.105
+      - nvidia-cudnn-cu12==8.9.2.26
+      - nvidia-cufft-cu12==11.0.2.54
+      - nvidia-curand-cu12==10.3.2.106
+      - nvidia-cusolver-cu12==11.4.5.107
+      - nvidia-cusparse-cu12==12.1.0.106
+      - nvidia-nccl-cu12==2.18.1
+      - nvidia-nvjitlink-cu12==12.2.140
+      - nvidia-nvtx-cu12==12.1.105
+      - packaging==23.2
+      - pandas==2.1.1
+      - peft==0.5.0
+      - protobuf==4.24.4
+      - psutil==5.9.5
+      - pyarrow==13.0.0
+      - python-dateutil==2.8.2
+      - pytz==2023.3.post1
+      - pyyaml==6.0.1
+      - regex==2023.10.3
+      - requests==2.31.0
+      - rouge==1.0.1
+      - s3transfer==0.7.0
+      - safetensors==0.3.3
+      - scipy==1.11.3
+      - six==1.16.0
+      - sympy==1.12
+      - tokenizers==0.14.0
+      - torch==2.1.0
+      - torchaudio==2.1.0
+      - tqdm==4.66.1
+      - transformers==4.34.0
+      - triton==2.1.0
+      - typing-extensions==4.8.0
+      - tzdata==2023.3
+      - urllib3==1.26.17
+      - xxhash==3.4.1
+      - yarl==1.9.2
+prefix: /opt/conda/envs/bark

backend/python/diffusers/Makefile

@@ -0,0 +1,11 @@
+.PHONY: diffusers
+diffusers:
+	@echo "Creating virtual environment..."
+	@conda env create --name diffusers --file diffusers.yml
+	@echo "Virtual environment created."
+
+.PHONY: run
+run:
+	@echo "Running diffusers..."
+	bash run.sh
+	@echo "Diffusers run."

backend/python/diffusers/README.md

@@ -0,0 +1,5 @@
+# Creating a separate environment for the diffusers project
+
+```
+make diffusers
+```

backend/python/diffusers/backend_diffusers.py

@@ -1,30 +1,39 @@
 #!/usr/bin/env python3
-import grpc
 from concurrent import futures
-import time
-import backend_pb2
-import backend_pb2_grpc
+
 import argparse
+from collections import defaultdict
+from enum import Enum
 import signal
 import sys
+import time
 import os
 
-# import diffusers
-import torch
-from torch import autocast
-from diffusers import StableDiffusionXLPipeline, StableDiffusionDepth2ImgPipeline, DPMSolverMultistepScheduler, StableDiffusionPipeline, DiffusionPipeline, EulerAncestralDiscreteScheduler
-from diffusers.pipelines.stable_diffusion import safety_checker
-from compel import Compel
 from PIL import Image
-from io import BytesIO
+import torch
+
+import backend_pb2
+import backend_pb2_grpc
+
+import grpc
+
+from diffusers import StableDiffusionXLPipeline, StableDiffusionDepth2ImgPipeline, DPMSolverMultistepScheduler, StableDiffusionPipeline, DiffusionPipeline, EulerAncestralDiscreteScheduler
 from diffusers import StableDiffusionImg2ImgPipeline
+from diffusers.pipelines.stable_diffusion import safety_checker
+
+from compel import Compel
+
 from transformers import CLIPTextModel
-from enum import Enum
+from safetensors.torch import load_file
+
 
 _ONE_DAY_IN_SECONDS = 60 * 60 * 24
 COMPEL=os.environ.get("COMPEL", "1") == "1"
 CLIPSKIP=os.environ.get("CLIPSKIP", "1") == "1"
 
+# If MAX_WORKERS are specified in the environment use it, otherwise default to 1
+MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
+
 # https://github.com/CompVis/stable-diffusion/issues/239#issuecomment-1627615287
 def sc(self, clip_input, images) : return images, [False for i in images]
 # edit the StableDiffusionSafetyChecker class so that, when called, it just returns the images and an array of True values
@@ -213,11 +222,83 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             self.compel = Compel(tokenizer=self.pipe.tokenizer, text_encoder=self.pipe.text_encoder)
             if request.CUDA:
                 self.pipe.to('cuda')
+            # Assume directory from request.ModelFile.
+            # Only if request.LoraAdapter it's not an absolute path
+            if request.LoraAdapter and request.ModelFile != "" and not os.path.isabs(request.LoraAdapter) and request.LoraAdapter:
+                # get base path of modelFile
+                modelFileBase = os.path.dirname(request.ModelFile)
+                # modify LoraAdapter to be relative to modelFileBase
+                request.LoraAdapter = os.path.join(modelFileBase, request.LoraAdapter)
+            device = "cpu" if not request.CUDA else "cuda"
+            self.device = device
+            if request.LoraAdapter:
+                # Check if its a local file and not a directory ( we load lora differently for a safetensor file )
+                if os.path.exists(request.LoraAdapter) and not os.path.isdir(request.LoraAdapter):
+                    self.load_lora_weights(request.LoraAdapter, 1, device, torchType)
+                else:
+                    self.pipe.unet.load_attn_procs(request.LoraAdapter)
+
         except Exception as err:
             return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
         # Implement your logic here for the LoadModel service
         # Replace this with your desired response
         return backend_pb2.Result(message="Model loaded successfully", success=True)
+
+    # https://github.com/huggingface/diffusers/issues/3064
+    def load_lora_weights(self, checkpoint_path, multiplier, device, dtype):
+        LORA_PREFIX_UNET = "lora_unet"
+        LORA_PREFIX_TEXT_ENCODER = "lora_te"
+        # load LoRA weight from .safetensors
+        state_dict = load_file(checkpoint_path, device=device)
+
+        updates = defaultdict(dict)
+        for key, value in state_dict.items():
+            # it is suggested to print out the key, it usually will be something like below
+            # "lora_te_text_model_encoder_layers_0_self_attn_k_proj.lora_down.weight"
+
+            layer, elem = key.split('.', 1)
+            updates[layer][elem] = value
+
+        # directly update weight in diffusers model
+        for layer, elems in updates.items():
+
+            if "text" in layer:
+                layer_infos = layer.split(LORA_PREFIX_TEXT_ENCODER + "_")[-1].split("_")
+                curr_layer = self.pipe.text_encoder
+            else:
+                layer_infos = layer.split(LORA_PREFIX_UNET + "_")[-1].split("_")
+                curr_layer = self.pipe.unet
+
+            # find the target layer
+            temp_name = layer_infos.pop(0)
+            while len(layer_infos) > -1:
+                try:
+                    curr_layer = curr_layer.__getattr__(temp_name)
+                    if len(layer_infos) > 0:
+                        temp_name = layer_infos.pop(0)
+                    elif len(layer_infos) == 0:
+                        break
+                except Exception:
+                    if len(temp_name) > 0:
+                        temp_name += "_" + layer_infos.pop(0)
+                    else:
+                        temp_name = layer_infos.pop(0)
+
+            # get elements for this layer
+            weight_up = elems['lora_up.weight'].to(dtype)
+            weight_down = elems['lora_down.weight'].to(dtype)
+            alpha = elems['alpha'] if 'alpha' in elems else None
+            if alpha:
+                alpha = alpha.item() / weight_up.shape[1]
+            else:
+                alpha = 1.0
+
+            # update weight
+            if len(weight_up.shape) == 4:
+                curr_layer.weight.data += multiplier * alpha * torch.mm(weight_up.squeeze(3).squeeze(2), weight_down.squeeze(3).squeeze(2)).unsqueeze(2).unsqueeze(3)
+            else:
+                curr_layer.weight.data += multiplier * alpha * torch.mm(weight_up, weight_down)
+
     def GenerateImage(self, request, context):
 
         prompt = request.positive_prompt
@@ -246,6 +327,12 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
         # create a dictionary of parameters by using the keys from EnableParameters and the values from defaults
         kwargs = {key: options[key] for key in keys}
 
+        # Set seed
+        if request.seed > 0:
+            kwargs["generator"] = torch.Generator(device=self.device).manual_seed(
+                request.seed
+            )
+
         image = {}
         if COMPEL:
             conditioning = self.compel.build_conditioning_tensor(prompt)
@@ -267,7 +354,7 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
         return backend_pb2.Result(message="Model loaded successfully", success=True)
 
 def serve(address):
-    server = grpc.server(futures.ThreadPoolExecutor(max_workers=1))
+    server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS))
     backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
     server.add_insecure_port(address)
     server.start()

backend/python/diffusers/diffusers.yml

@@ -0,0 +1,74 @@
+name: diffusers
+channels:
+  - defaults
+dependencies:
+  - _libgcc_mutex=0.1=main
+  - _openmp_mutex=5.1=1_gnu
+  - bzip2=1.0.8=h7b6447c_0
+  - ca-certificates=2023.08.22=h06a4308_0
+  - ld_impl_linux-64=2.38=h1181459_1
+  - libffi=3.4.4=h6a678d5_0
+  - libgcc-ng=11.2.0=h1234567_1
+  - libgomp=11.2.0=h1234567_1
+  - libstdcxx-ng=11.2.0=h1234567_1
+  - libuuid=1.41.5=h5eee18b_0
+  - ncurses=6.4=h6a678d5_0
+  - openssl=3.0.11=h7f8727e_2
+  - pip=23.2.1=py311h06a4308_0
+  - python=3.11.5=h955ad1f_0
+  - readline=8.2=h5eee18b_0
+  - setuptools=68.0.0=py311h06a4308_0
+  - sqlite=3.41.2=h5eee18b_0
+  - tk=8.6.12=h1ccaba5_0
+  - tzdata=2023c=h04d1e81_0
+  - wheel=0.41.2=py311h06a4308_0
+  - xz=5.4.2=h5eee18b_0
+  - zlib=1.2.13=h5eee18b_0
+  - pip:
+      - accelerate==0.23.0
+      - certifi==2023.7.22
+      - charset-normalizer==3.3.0
+      - compel==2.0.2
+      - diffusers==0.21.4
+      - filelock==3.12.4
+      - fsspec==2023.9.2
+      - grpcio==1.59.0
+      - huggingface-hub==0.17.3
+      - idna==3.4
+      - importlib-metadata==6.8.0
+      - jinja2==3.1.2
+      - markupsafe==2.1.3
+      - mpmath==1.3.0
+      - networkx==3.1
+      - numpy==1.26.0
+      - nvidia-cublas-cu12==12.1.3.1
+      - nvidia-cuda-cupti-cu12==12.1.105
+      - nvidia-cuda-nvrtc-cu12==12.1.105
+      - nvidia-cuda-runtime-cu12==12.1.105
+      - nvidia-cudnn-cu12==8.9.2.26
+      - nvidia-cufft-cu12==11.0.2.54
+      - nvidia-curand-cu12==10.3.2.106
+      - nvidia-cusolver-cu12==11.4.5.107
+      - nvidia-cusparse-cu12==12.1.0.106
+      - nvidia-nccl-cu12==2.18.1
+      - nvidia-nvjitlink-cu12==12.2.140
+      - nvidia-nvtx-cu12==12.1.105
+      - packaging==23.2
+      - pillow==10.0.1
+      - protobuf==4.24.4
+      - psutil==5.9.5
+      - pyparsing==3.1.1
+      - pyyaml==6.0.1
+      - regex==2023.10.3
+      - requests==2.31.0
+      - safetensors==0.4.0
+      - sympy==1.12
+      - tokenizers==0.14.1
+      - torch==2.1.0
+      - tqdm==4.66.1
+      - transformers==4.34.0
+      - triton==2.1.0
+      - typing-extensions==4.8.0
+      - urllib3==2.0.6
+      - zipp==3.17.0
+prefix: /opt/conda/envs/diffusers

backend/python/diffusers/run.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+
+##
+## A bash script wrapper that runs the diffusers server with conda
+
+export PATH=$PATH:/opt/conda/bin
+
+# Activate conda environment
+source activate diffusers
+
+# get the directory where the bash script is located
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+python $DIR/backend_diffusers.py $@

backend/python/exllama/Makefile

@@ -0,0 +1,11 @@
+.PHONY: exllama
+exllama:
+	@echo "Creating virtual environment..."
+	@conda env create --name exllama --file exllama.yml
+	@echo "Virtual environment created."
+
+.PHONY: run
+run:
+	@echo "Running exllama..."
+	bash run.sh
+	@echo "exllama run."

backend/python/exllama/README.md

@@ -0,0 +1,5 @@
+# Creating a separate environment for the exllama project
+
+```
+make exllama
+```