WIP

.github/workflows/backend.yml

@@ -381,12 +381,24 @@ jobs:
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
           # sycl builds
-          - build-type: 'intel'
+          - build-type: 'sycl_f32'
             cuda-major-version: ""
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-rerankers'
+            tag-suffix: '-gpu-intel-sycl-f32-rerankers'
+            runs-on: 'ubuntu-latest'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            skip-drivers: 'false'
+            backend: "rerankers"
+            dockerfile: "./backend/Dockerfile.python"
+            context: "./backend"
+          - build-type: 'sycl_f16'
+            cuda-major-version: ""
+            cuda-minor-version: ""
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-gpu-intel-sycl-f16-rerankers'
             runs-on: 'ubuntu-latest'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
@@ -417,36 +429,60 @@ jobs:
             backend: "llama-cpp"
             dockerfile: "./backend/Dockerfile.llama-cpp"
             context: "./"
-          - build-type: 'intel'
+          - build-type: 'sycl_f32'
             cuda-major-version: ""
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-vllm'
+            tag-suffix: '-gpu-intel-sycl-f32-vllm'
             runs-on: 'ubuntu-latest'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "vllm"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: 'intel'
+          - build-type: 'sycl_f16'
             cuda-major-version: ""
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-transformers'
+            tag-suffix: '-gpu-intel-sycl-f16-vllm'
+            runs-on: 'ubuntu-latest'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            skip-drivers: 'false'
+            backend: "vllm"
+            dockerfile: "./backend/Dockerfile.python"
+            context: "./backend"
+          - build-type: 'sycl_f32'
+            cuda-major-version: ""
+            cuda-minor-version: ""
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-gpu-intel-sycl-f32-transformers'
             runs-on: 'ubuntu-latest'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "transformers"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: 'intel'
+          - build-type: 'sycl_f16'
             cuda-major-version: ""
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-diffusers'
+            tag-suffix: '-gpu-intel-sycl-f16-transformers'
+            runs-on: 'ubuntu-latest'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            skip-drivers: 'false'
+            backend: "transformers"
+            dockerfile: "./backend/Dockerfile.python"
+            context: "./backend"
+          - build-type: 'sycl_f32'
+            cuda-major-version: ""
+            cuda-minor-version: ""
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-gpu-intel-sycl-f32-diffusers'
             runs-on: 'ubuntu-latest'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
@@ -454,48 +490,96 @@ jobs:
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
           # SYCL additional backends
-          - build-type: 'intel'
+          - build-type: 'sycl_f32'
             cuda-major-version: ""
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-kokoro'
+            tag-suffix: '-gpu-intel-sycl-f32-kokoro'
             runs-on: 'ubuntu-latest'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "kokoro"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: 'intel'
+          - build-type: 'sycl_f16'
             cuda-major-version: ""
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-faster-whisper'
+            tag-suffix: '-gpu-intel-sycl-f16-kokoro'
+            runs-on: 'ubuntu-latest'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            skip-drivers: 'false'
+            backend: "kokoro"
+            dockerfile: "./backend/Dockerfile.python"
+            context: "./backend"
+          - build-type: 'sycl_f32'
+            cuda-major-version: ""
+            cuda-minor-version: ""
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-gpu-intel-sycl-f32-faster-whisper'
             runs-on: 'ubuntu-latest'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "faster-whisper"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: 'intel'
+          - build-type: 'sycl_f16'
             cuda-major-version: ""
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-coqui'
+            tag-suffix: '-gpu-intel-sycl-f16-faster-whisper'
+            runs-on: 'ubuntu-latest'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            skip-drivers: 'false'
+            backend: "faster-whisper"
+            dockerfile: "./backend/Dockerfile.python"
+            context: "./backend"
+          - build-type: 'sycl_f32'
+            cuda-major-version: ""
+            cuda-minor-version: ""
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-gpu-intel-sycl-f32-coqui'
             runs-on: 'ubuntu-latest'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "coqui"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: 'intel'
+          - build-type: 'sycl_f16'
             cuda-major-version: ""
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-bark'
+            tag-suffix: '-gpu-intel-sycl-f16-coqui'
+            runs-on: 'ubuntu-latest'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            skip-drivers: 'false'
+            backend: "coqui"
+            dockerfile: "./backend/Dockerfile.python"
+            context: "./backend"
+          - build-type: 'sycl_f32'
+            cuda-major-version: ""
+            cuda-minor-version: ""
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-gpu-intel-sycl-f32-bark'
+            runs-on: 'ubuntu-latest'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            skip-drivers: 'false'
+            backend: "bark"
+            dockerfile: "./backend/Dockerfile.python"
+            context: "./backend"
+          - build-type: 'sycl_f16'
+            cuda-major-version: ""
+            cuda-minor-version: ""
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-gpu-intel-sycl-f16-bark'
             runs-on: 'ubuntu-latest'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
@@ -513,7 +597,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "piper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           # bark-cpp
           - build-type: ''
@@ -526,7 +610,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "bark-cpp"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: ''
             cuda-major-version: ""
@@ -575,7 +659,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "stablediffusion-ggml"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'cublas'
             cuda-major-version: "12"
@@ -587,7 +671,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "stablediffusion-ggml"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'cublas'
             cuda-major-version: "11"
@@ -599,7 +683,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "stablediffusion-ggml"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'sycl_f32'
             cuda-major-version: ""
@@ -611,7 +695,7 @@ jobs:
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "stablediffusion-ggml"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'sycl_f16'
             cuda-major-version: ""
@@ -623,7 +707,7 @@ jobs:
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "stablediffusion-ggml"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'vulkan'
             cuda-major-version: ""
@@ -635,7 +719,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "stablediffusion-ggml"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'cublas'
             cuda-major-version: "12"
@@ -647,7 +731,7 @@ jobs:
             base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
             runs-on: 'ubuntu-24.04-arm'
             backend: "stablediffusion-ggml"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           # whisper
           - build-type: ''
@@ -660,7 +744,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "whisper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'cublas'
             cuda-major-version: "12"
@@ -672,7 +756,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "whisper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'cublas'
             cuda-major-version: "11"
@@ -684,7 +768,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "whisper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'sycl_f32'
             cuda-major-version: ""
@@ -696,7 +780,7 @@ jobs:
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "whisper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'sycl_f16'
             cuda-major-version: ""
@@ -708,7 +792,7 @@ jobs:
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             skip-drivers: 'false'
             backend: "whisper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'vulkan'
             cuda-major-version: ""
@@ -720,7 +804,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "whisper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'cublas'
             cuda-major-version: "12"
@@ -732,7 +816,7 @@ jobs:
             base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
             runs-on: 'ubuntu-24.04-arm'
             backend: "whisper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           - build-type: 'hipblas'
             cuda-major-version: ""
@@ -744,7 +828,7 @@ jobs:
             runs-on: 'ubuntu-latest'
             skip-drivers: 'false'
             backend: "whisper"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           #silero-vad
           - build-type: ''
@@ -757,7 +841,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "silero-vad"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           # local-store
           - build-type: ''
@@ -770,7 +854,7 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "local-store"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
             context: "./"
           # huggingface
           - build-type: ''
@@ -783,143 +867,8 @@ jobs:
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             backend: "huggingface"
-            dockerfile: "./backend/Dockerfile.golang"
+            dockerfile: "./backend/Dockerfile.go"
-            context: "./"
+            context: "./"  
-          # rfdetr
-          - build-type: ''
-            cuda-major-version: ""
-            cuda-minor-version: ""
-            platforms: 'linux/amd64,linux/arm64'
-            tag-latest: 'auto'
-            tag-suffix: '-cpu-rfdetr'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'false'
-            backend: "rfdetr"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'cublas'
-            cuda-major-version: "12"
-            cuda-minor-version: "0"
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-cuda-12-rfdetr'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'false'
-            backend: "rfdetr"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'cublas'
-            cuda-major-version: "11"
-            cuda-minor-version: "7"
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-cuda-11-rfdetr'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'false'
-            backend: "rfdetr"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'intel'
-            cuda-major-version: ""
-            cuda-minor-version: ""
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-rfdetr'
-            runs-on: 'ubuntu-latest'
-            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
-            skip-drivers: 'false'
-            backend: "rfdetr"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'cublas'
-            cuda-major-version: "12"
-            cuda-minor-version: "0"
-            platforms: 'linux/arm64'
-            skip-drivers: 'true'
-            tag-latest: 'auto'
-            tag-suffix: '-nvidia-l4t-arm64-rfdetr'
-            base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
-            runs-on: 'ubuntu-24.04-arm'
-            backend: "rfdetr"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          # exllama2
-          - build-type: ''
-            cuda-major-version: ""
-            cuda-minor-version: ""
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-cpu-exllama2'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'false'
-            backend: "exllama2"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'cublas'
-            cuda-major-version: "12"
-            cuda-minor-version: "0"
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-cuda-12-exllama2'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'false'
-            backend: "exllama2"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'cublas'
-            cuda-major-version: "11"
-            cuda-minor-version: "7"
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-cuda-11-exllama2'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'false'
-            backend: "exllama2"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'intel'
-            cuda-major-version: ""
-            cuda-minor-version: ""
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-intel-exllama2'
-            runs-on: 'ubuntu-latest'
-            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
-            skip-drivers: 'false'
-            backend: "exllama2"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'hipblas'
-            cuda-major-version: ""
-            cuda-minor-version: ""
-            platforms: 'linux/amd64'
-            skip-drivers: 'true'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-hipblas-exllama2'
-            base-image: "rocm/dev-ubuntu-22.04:6.1"
-            runs-on: 'ubuntu-latest'
-            backend: "exllama2"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          # runs out of space on the runner
-          # - build-type: 'hipblas'
-          #   cuda-major-version: ""
-          #   cuda-minor-version: ""
-          #   platforms: 'linux/amd64'
-          #   tag-latest: 'auto'
-          #   tag-suffix: '-gpu-hipblas-rfdetr'
-          #   base-image: "rocm/dev-ubuntu-22.04:6.1"
-          #   runs-on: 'ubuntu-latest'
-          #   skip-drivers: 'false'
-          #   backend: "rfdetr"
-          #   dockerfile: "./backend/Dockerfile.python"
-          #   context: "./backend"
   llama-cpp-darwin:
     runs-on: macOS-14
     strategy:

.github/workflows/bump_deps.yaml

@@ -21,7 +21,7 @@ jobs:
             variable: "BARKCPP_VERSION"
             branch: "main"
             file: "Makefile"
-          - repository: "leejet/stable-diffusion.cpp"
+          - repository: "richiejp/stable-diffusion.cpp"
             variable: "STABLEDIFFUSION_GGML_VERSION"
             branch: "master"
             file: "backend/go/stablediffusion-ggml/Makefile"

.github/workflows/image-pr.yml

@@ -39,7 +39,7 @@ jobs:
             cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            tag-suffix: '-gpu-nvidia-cuda-12'
+            tag-suffix: '-gpu-nvidia-cuda12'
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
             makeflags: "--jobs=3 --output-sync=target"
@@ -51,12 +51,12 @@ jobs:
             grpc-base-image: "ubuntu:22.04"
             runs-on: 'ubuntu-latest'
             makeflags: "--jobs=3 --output-sync=target"
-          - build-type: 'sycl'
+          - build-type: 'sycl_f16'
             platforms: 'linux/amd64'
             tag-latest: 'false'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             grpc-base-image: "ubuntu:22.04"
-            tag-suffix: 'sycl'
+            tag-suffix: 'sycl-f16'
             runs-on: 'ubuntu-latest'
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'vulkan'

.github/workflows/image.yml

@@ -83,7 +83,7 @@ jobs:
             cuda-minor-version: "7"
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-cuda-11'
+            tag-suffix: '-gpu-nvidia-cuda11'
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
             makeflags: "--jobs=4 --output-sync=target"
@@ -94,7 +94,7 @@ jobs:
             cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-cuda-12'
+            tag-suffix: '-gpu-nvidia-cuda12'
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
@@ -103,21 +103,30 @@ jobs:
           - build-type: 'vulkan'
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-vulkan'
+            tag-suffix: '-vulkan'
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
             makeflags: "--jobs=4 --output-sync=target"
             aio: "-aio-gpu-vulkan"
-          - build-type: 'sycl'
+          - build-type: 'sycl_f16'
             platforms: 'linux/amd64'
             tag-latest: 'auto'
             base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
             grpc-base-image: "ubuntu:22.04"
-            tag-suffix: '-gpu-intel'
+            tag-suffix: '-gpu-intel-f16'
             runs-on: 'ubuntu-latest'
             makeflags: "--jobs=3 --output-sync=target"
-            aio: "-aio-gpu-intel"
+            aio: "-aio-gpu-intel-f16"
+          - build-type: 'sycl_f32'
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
+            tag-suffix: '-gpu-intel-f32'
+            runs-on: 'ubuntu-latest'
+            makeflags: "--jobs=3 --output-sync=target"
+            aio: "-aio-gpu-intel-f32"
 
   gh-runner:
     uses: ./.github/workflows/image_build.yml

.gitignore

@@ -12,7 +12,6 @@ prepare-sources
 /backends
 /backend-images
 /result.yaml
-protoc
 
 *.log
 

Dockerfile

@@ -72,12 +72,6 @@ RUN <<EOT bash
     fi
 EOT
 
-RUN <<EOT bash
-    if [ "${BUILD_TYPE}" = "cublas" ] && [ "${TARGETARCH}" = "arm64" ]; then
-        echo "nvidia-l4t" > /run/localai/capability
-    fi
-EOT
-
 # If we are building with clblas support, we need the libraries for the builds
 RUN if [ "${BUILD_TYPE}" = "clblas" ] && [ "${SKIP_DRIVERS}" = "false" ]; then \
         apt-get update && \
@@ -100,11 +94,6 @@ RUN if [ "${BUILD_TYPE}" = "hipblas" ] && [ "${SKIP_DRIVERS}" = "false" ]; then
         ldconfig \
     ; fi
 
-RUN expr "${BUILD_TYPE}" : sycl && \
-      echo "intel" > /run/localai/capability || \
-      echo "Not Intel"
-
-
 # Cuda
 ENV PATH=/usr/local/cuda/bin:${PATH}
 

Makefile

@@ -5,6 +5,8 @@ BINARY_NAME=local-ai
 
 GORELEASER?=
 
+ONEAPI_VERSION?=2025.2
+
 export BUILD_TYPE?=
 
 GO_TAGS?=
@@ -143,7 +145,7 @@ backends/stablediffusion-ggml: docker-build-stablediffusion-ggml docker-save-sta
 
 backends/whisper: docker-build-whisper docker-save-whisper build
 	./local-ai backends install "ocifile://$(abspath ./backend-images/whisper.tar)"
-
+	
 backends/silero-vad: docker-build-silero-vad docker-save-silero-vad build
 	./local-ai backends install "ocifile://$(abspath ./backend-images/silero-vad.tar)"
 
@@ -153,9 +155,6 @@ backends/local-store: docker-build-local-store docker-save-local-store build
 backends/huggingface: docker-build-huggingface docker-save-huggingface build
 	./local-ai backends install "ocifile://$(abspath ./backend-images/huggingface.tar)"
 
-backends/rfdetr: docker-build-rfdetr docker-save-rfdetr build
-	./local-ai backends install "ocifile://$(abspath ./backend-images/rfdetr.tar)"
-
 ########################################################
 ## AIO tests
 ########################################################
@@ -243,7 +242,10 @@ help: ## Show this help.
 ########################################################
 
 .PHONY: protogen
-protogen: protogen-go
+protogen: protogen-go protogen-python
+
+.PHONY: protogen-clean
+protogen-clean: protogen-go-clean protogen-python-clean
 
 protoc:
 	@OS_NAME=$$(uname -s | tr '[:upper:]' '[:lower:]'); \
@@ -288,6 +290,93 @@ protogen-go-clean:
 	$(RM) pkg/grpc/proto/backend.pb.go pkg/grpc/proto/backend_grpc.pb.go
 	$(RM) bin/*
 
+.PHONY: protogen-python
+protogen-python: bark-protogen coqui-protogen chatterbox-protogen diffusers-protogen exllama2-protogen rerankers-protogen transformers-protogen kokoro-protogen vllm-protogen faster-whisper-protogen
+
+.PHONY: protogen-python-clean
+protogen-python-clean: bark-protogen-clean coqui-protogen-clean chatterbox-protogen-clean diffusers-protogen-clean  exllama2-protogen-clean rerankers-protogen-clean transformers-protogen-clean kokoro-protogen-clean vllm-protogen-clean faster-whisper-protogen-clean
+
+.PHONY: bark-protogen
+bark-protogen:
+	$(MAKE) -C backend/python/bark protogen
+
+.PHONY: bark-protogen-clean
+bark-protogen-clean:
+	$(MAKE) -C backend/python/bark protogen-clean
+
+.PHONY: coqui-protogen
+coqui-protogen:
+	$(MAKE) -C backend/python/coqui protogen
+
+.PHONY: coqui-protogen-clean
+coqui-protogen-clean:
+	$(MAKE) -C backend/python/coqui protogen-clean
+
+.PHONY: diffusers-protogen
+diffusers-protogen:
+	$(MAKE) -C backend/python/diffusers protogen
+
+.PHONY: chatterbox-protogen
+chatterbox-protogen:
+	$(MAKE) -C backend/python/chatterbox protogen
+
+.PHONY: diffusers-protogen-clean
+diffusers-protogen-clean:
+	$(MAKE) -C backend/python/diffusers protogen-clean
+
+.PHONY: chatterbox-protogen-clean
+chatterbox-protogen-clean:
+	$(MAKE) -C backend/python/chatterbox protogen-clean
+
+.PHONY: faster-whisper-protogen
+faster-whisper-protogen:
+	$(MAKE) -C backend/python/faster-whisper protogen
+
+.PHONY: faster-whisper-protogen-clean
+faster-whisper-protogen-clean:
+	$(MAKE) -C backend/python/faster-whisper protogen-clean
+
+.PHONY: exllama2-protogen
+exllama2-protogen:
+	$(MAKE) -C backend/python/exllama2 protogen
+
+.PHONY: exllama2-protogen-clean
+exllama2-protogen-clean:
+	$(MAKE) -C backend/python/exllama2 protogen-clean
+
+.PHONY: rerankers-protogen
+rerankers-protogen:
+	$(MAKE) -C backend/python/rerankers protogen
+
+.PHONY: rerankers-protogen-clean
+rerankers-protogen-clean:
+	$(MAKE) -C backend/python/rerankers protogen-clean
+
+.PHONY: transformers-protogen
+transformers-protogen:
+	$(MAKE) -C backend/python/transformers protogen
+
+.PHONY: transformers-protogen-clean
+transformers-protogen-clean:
+	$(MAKE) -C backend/python/transformers protogen-clean
+
+.PHONY: kokoro-protogen
+kokoro-protogen:
+	$(MAKE) -C backend/python/kokoro protogen
+
+.PHONY: kokoro-protogen-clean
+kokoro-protogen-clean:
+	$(MAKE) -C backend/python/kokoro protogen-clean
+
+.PHONY: vllm-protogen
+vllm-protogen:
+	$(MAKE) -C backend/python/vllm protogen
+
+.PHONY: vllm-protogen-clean
+vllm-protogen-clean:
+	$(MAKE) -C backend/python/vllm protogen-clean
+
+
 prepare-test-extra: protogen-python
 	$(MAKE) -C backend/python/transformers
 	$(MAKE) -C backend/python/diffusers
@@ -323,7 +412,7 @@ docker-cuda11:
 		--build-arg GO_TAGS="$(GO_TAGS)" \
 		--build-arg MAKEFLAGS="$(DOCKER_MAKEFLAGS)" \
 		--build-arg BUILD_TYPE=$(BUILD_TYPE) \
-		-t $(DOCKER_IMAGE)-cuda-11 .
+		-t $(DOCKER_IMAGE)-cuda11 .
 
 docker-aio:
 	@echo "Building AIO image with base $(BASE_IMAGE) as $(DOCKER_AIO_IMAGE)"
@@ -338,11 +427,19 @@ docker-aio-all:
 
 docker-image-intel:
 	docker build \
-		--build-arg BASE_IMAGE=quay.io/go-skynet/intel-oneapi-base:latest \
+		--build-arg BASE_IMAGE=intel/oneapi-basekit:${ONEAPI_VERSION}.0-0-devel-ubuntu24.04 \
 		--build-arg IMAGE_TYPE=$(IMAGE_TYPE) \
 		--build-arg GO_TAGS="$(GO_TAGS)" \
 		--build-arg MAKEFLAGS="$(DOCKER_MAKEFLAGS)" \
-		--build-arg BUILD_TYPE=sycl -t $(DOCKER_IMAGE) .
+		--build-arg BUILD_TYPE=sycl_f32 -t $(DOCKER_IMAGE) .
+
+docker-image-intel-xpu:
+	docker build \
+		--build-arg BASE_IMAGE=intel/oneapi-basekit:${ONEAPI_VERSION}.0-0-devel-ubuntu22.04 \
+		--build-arg IMAGE_TYPE=$(IMAGE_TYPE) \
+		--build-arg GO_TAGS="$(GO_TAGS)" \
+		--build-arg MAKEFLAGS="$(DOCKER_MAKEFLAGS)" \
+		--build-arg BUILD_TYPE=sycl_f32 -t $(DOCKER_IMAGE) .
 
 ########################################################
 ## Backends
@@ -352,25 +449,19 @@ backend-images:
 	mkdir -p backend-images
 
 docker-build-llama-cpp:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:llama-cpp -f backend/Dockerfile.llama-cpp .
+	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg IMAGE_BASE=$(IMAGE_BASE) -t local-ai-backend:llama-cpp -f backend/Dockerfile.llama-cpp .
 
 docker-build-bark-cpp:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:bark-cpp -f backend/Dockerfile.golang --build-arg BACKEND=bark-cpp .
+	docker build -t local-ai-backend:bark-cpp -f backend/Dockerfile.go --build-arg BACKEND=bark-cpp .
 
 docker-build-piper:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:piper -f backend/Dockerfile.golang --build-arg BACKEND=piper .
+	docker build -t local-ai-backend:piper -f backend/Dockerfile.go --build-arg BACKEND=piper .
 
 docker-build-local-store:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:local-store -f backend/Dockerfile.golang --build-arg BACKEND=local-store .
+	docker build -t local-ai-backend:local-store -f backend/Dockerfile.go --build-arg BACKEND=local-store .
 
 docker-build-huggingface:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:huggingface -f backend/Dockerfile.golang --build-arg BACKEND=huggingface .
+	docker build -t local-ai-backend:huggingface -f backend/Dockerfile.go --build-arg BACKEND=huggingface .
-
-docker-build-rfdetr:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:rfdetr -f backend/Dockerfile.python --build-arg BACKEND=rfdetr ./backend
-
-docker-save-rfdetr: backend-images
-	docker save local-ai-backend:rfdetr -o backend-images/rfdetr.tar
 
 docker-save-huggingface: backend-images
 	docker save local-ai-backend:huggingface -o backend-images/huggingface.tar
@@ -379,7 +470,7 @@ docker-save-local-store: backend-images
 	docker save local-ai-backend:local-store -o backend-images/local-store.tar
 
 docker-build-silero-vad:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:silero-vad -f backend/Dockerfile.golang --build-arg BACKEND=silero-vad .
+	docker build -t local-ai-backend:silero-vad -f backend/Dockerfile.go --build-arg BACKEND=silero-vad .
 
 docker-save-silero-vad: backend-images
 	docker save local-ai-backend:silero-vad -o backend-images/silero-vad.tar
@@ -394,46 +485,46 @@ docker-save-bark-cpp: backend-images
 	docker save local-ai-backend:bark-cpp -o backend-images/bark-cpp.tar
 
 docker-build-stablediffusion-ggml:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:stablediffusion-ggml -f backend/Dockerfile.golang --build-arg BACKEND=stablediffusion-ggml .
+	docker build -t local-ai-backend:stablediffusion-ggml -f backend/Dockerfile.go --build-arg BACKEND=stablediffusion-ggml .
 
 docker-save-stablediffusion-ggml: backend-images
 	docker save local-ai-backend:stablediffusion-ggml -o backend-images/stablediffusion-ggml.tar
 
 docker-build-rerankers:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:rerankers -f backend/Dockerfile.python --build-arg BACKEND=rerankers .
+	docker build -t local-ai-backend:rerankers -f backend/Dockerfile.python --build-arg BACKEND=rerankers .
 
 docker-build-vllm:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:vllm -f backend/Dockerfile.python --build-arg BACKEND=vllm .
+	docker build -t local-ai-backend:vllm -f backend/Dockerfile.python --build-arg BACKEND=vllm .
 
 docker-build-transformers:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:transformers -f backend/Dockerfile.python --build-arg BACKEND=transformers .
+	docker build -t local-ai-backend:transformers -f backend/Dockerfile.python --build-arg BACKEND=transformers .
 
 docker-build-diffusers:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:diffusers -f backend/Dockerfile.python --build-arg BACKEND=diffusers .
+	docker build -t local-ai-backend:diffusers -f backend/Dockerfile.python --build-arg BACKEND=diffusers .
 
 docker-build-kokoro:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:kokoro -f backend/Dockerfile.python --build-arg BACKEND=kokoro .
+	docker build -t local-ai-backend:kokoro -f backend/Dockerfile.python --build-arg BACKEND=kokoro .
 
 docker-build-whisper:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:whisper -f backend/Dockerfile.golang --build-arg BACKEND=whisper  .
+	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:whisper -f backend/Dockerfile.go --build-arg BACKEND=whisper  .
 
 docker-save-whisper: backend-images
 	docker save local-ai-backend:whisper -o backend-images/whisper.tar
 
 docker-build-faster-whisper:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:faster-whisper -f backend/Dockerfile.python --build-arg BACKEND=faster-whisper .
+	docker build -t local-ai-backend:faster-whisper -f backend/Dockerfile.python --build-arg BACKEND=faster-whisper .
 
 docker-build-coqui:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:coqui -f backend/Dockerfile.python --build-arg BACKEND=coqui .
+	docker build -t local-ai-backend:coqui -f backend/Dockerfile.python --build-arg BACKEND=coqui .
 
 docker-build-bark:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:bark -f backend/Dockerfile.python --build-arg BACKEND=bark .
+	docker build -t local-ai-backend:bark -f backend/Dockerfile.python --build-arg BACKEND=bark .
 
 docker-build-chatterbox:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:chatterbox -f backend/Dockerfile.python --build-arg BACKEND=chatterbox .
+	docker build -t local-ai-backend:chatterbox -f backend/Dockerfile.python --build-arg BACKEND=chatterbox .
 
 docker-build-exllama2:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:exllama2 -f backend/Dockerfile.python --build-arg BACKEND=exllama2 .
+	docker build -t local-ai-backend:exllama2 -f backend/Dockerfile.python --build-arg BACKEND=exllama2 .
 
 docker-build-backends: docker-build-llama-cpp docker-build-rerankers docker-build-vllm docker-build-transformers docker-build-diffusers docker-build-kokoro docker-build-faster-whisper docker-build-coqui docker-build-bark docker-build-chatterbox docker-build-exllama2
 

README.md

@@ -140,7 +140,11 @@ docker run -ti --name local-ai -p 8080:8080 --device=/dev/kfd --device=/dev/dri
 ### Intel GPU Images (oneAPI):
 
 ```bash
-docker run -ti --name local-ai -p 8080:8080 --device=/dev/dri/card1 --device=/dev/dri/renderD128 localai/localai:latest-gpu-intel
+# Intel GPU with FP16 support
+docker run -ti --name local-ai -p 8080:8080 --device=/dev/dri/card1 --device=/dev/dri/renderD128 localai/localai:latest-gpu-intel-f16
+
+# Intel GPU with FP32 support
+docker run -ti --name local-ai -p 8080:8080 --device=/dev/dri/card1 --device=/dev/dri/renderD128 localai/localai:latest-gpu-intel-f32
 ```
 
 ### Vulkan GPU Images:
@@ -162,7 +166,7 @@ docker run -ti --name local-ai -p 8080:8080 --gpus all localai/localai:latest-ai
 docker run -ti --name local-ai -p 8080:8080 --gpus all localai/localai:latest-aio-gpu-nvidia-cuda-11
 
 # Intel GPU version
-docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-aio-gpu-intel
+docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-aio-gpu-intel-f16
 
 # AMD GPU version
 docker run -ti --name local-ai -p 8080:8080 --device=/dev/kfd --device=/dev/dri --group-add=video localai/localai:latest-aio-gpu-hipblas
@@ -185,14 +189,10 @@ local-ai run https://gist.githubusercontent.com/.../phi-2.yaml
 local-ai run oci://localai/phi-2:latest
 ```
 
-> ⚡ **Automatic Backend Detection**: When you install models from the gallery or YAML files, LocalAI automatically detects your system's GPU capabilities (NVIDIA, AMD, Intel) and downloads the appropriate backend. For advanced configuration options, see [GPU Acceleration](https://localai.io/features/gpu-acceleration/#automatic-backend-detection).
-
 For more information, see [💻 Getting started](https://localai.io/basics/getting_started/index.html)
 
 ## 📰 Latest project news
 
-- July/August 2025: 🔍 [Object Detection](https://localai.io/features/object-detection/) added to the API featuring [rf-detr](https://github.com/roboflow/rf-detr)
-- July 2025: All backends migrated outside of the main binary. LocalAI is now more lightweight, small, and automatically downloads the required backend to run the model. [Read the release notes](https://github.com/mudler/LocalAI/releases/tag/v3.2.0)
 - June 2025: [Backend management](https://github.com/mudler/LocalAI/pull/5607) has been added. Attention: extras images are going to be deprecated from the next release! Read [the backend management PR](https://github.com/mudler/LocalAI/pull/5607).
 - May 2025: [Audio input](https://github.com/mudler/LocalAI/pull/5466) and [Reranking](https://github.com/mudler/LocalAI/pull/5396) in llama.cpp backend, [Realtime API](https://github.com/mudler/LocalAI/pull/5392),  Support to Gemma, SmollVLM, and more multimodal models (available in the gallery).
 - May 2025: Important: image name changes [See release](https://github.com/mudler/LocalAI/releases/tag/v2.29.0)
@@ -225,7 +225,6 @@ Roadmap items: [List of issues](https://github.com/mudler/LocalAI/issues?q=is%3A
 - ✍️ [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/)
-- 🔍 [Object Detection](https://localai.io/features/object-detection/)
 - 📈 [Reranker API](https://localai.io/features/reranker/)
 - 🆕🖧 [P2P Inferencing](https://localai.io/features/distribute/)
 - [Agentic capabilities](https://github.com/mudler/LocalAGI)

backend/Dockerfile.go

@@ -96,6 +96,17 @@ RUN if [ "${BUILD_TYPE}" = "hipblas" ] && [ "${SKIP_DRIVERS}" = "false" ]; then
         ldconfig \
     ; fi
 
+# Intel oneAPI requirements
+RUN <<EOT bash
+    if [[ "${BUILD_TYPE}" == sycl* ]] && [ "${SKIP_DRIVERS}" = "false" ]; then
+        apt-get update && \
+        apt-get install -y --no-install-recommends \
+            intel-oneapi-runtime-libs && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/*
+    fi
+EOT
+
 # Install Go
 RUN curl -L -s https://go.dev/dl/go${GO_VERSION}.linux-${TARGETARCH}.tar.gz | tar -C /usr/local -xz
 ENV PATH=$PATH:/root/go/bin:/usr/local/go/bin:/usr/local/bin

backend/Dockerfile.llama-cpp

@@ -11,6 +11,7 @@ ARG GRPC_MAKEFLAGS="-j4 -Otarget"
 ARG GRPC_VERSION=v1.65.0
 ARG CMAKE_FROM_SOURCE=false
 ARG CMAKE_VERSION=3.26.4
+ARG PROTOBUF_VERSION=v21.12
 
 ENV MAKEFLAGS=${GRPC_MAKEFLAGS}
 
@@ -49,6 +50,14 @@ RUN git clone --recurse-submodules --jobs 4 -b ${GRPC_VERSION} --depth 1 --shall
     make install && \
     rm -rf /build
 
+RUN git clone --recurse-submodules --branch ${PROTOBUF_VERSION} https://github.com/protocolbuffers/protobuf.git && \
+    mkdir -p /build/protobuf/build && \
+    cd /build/protobuf/build && \
+    cmake -Dprotobuf_BUILD_SHARED_LIBS=ON -Dprotobuf_BUILD_TESTS=OFF .. && \
+    make && \
+    make install && \
+    rm -rf /build
+
 FROM ${BASE_IMAGE} AS builder
 ARG BACKEND=rerankers
 ARG BUILD_TYPE
@@ -180,21 +189,9 @@ COPY --from=grpc /opt/grpc /usr/local
 
 COPY . /LocalAI
 
-## Otherwise just run the normal build
+RUN make -C /LocalAI/backend/cpp/llama-cpp llama-cpp
-RUN <<EOT bash
+RUN make -C /LocalAI/backend/cpp/llama-cpp llama-cpp-grpc
-if [ "${TARGETARCH}" = "arm64" ] || [ "${BUILD_TYPE}" = "hipblas" ]; then \
+RUN make -C /LocalAI/backend/cpp/llama-cpp llama-cpp-rpc-server
-        cd /LocalAI/backend/cpp/llama-cpp && make llama-cpp-fallback && \
-        make llama-cpp-grpc && make llama-cpp-rpc-server; \
-    else \
-        cd /LocalAI/backend/cpp/llama-cpp && make llama-cpp-avx && \
-        make llama-cpp-avx2 && \
-        make llama-cpp-avx512 && \
-        make llama-cpp-fallback && \
-        make llama-cpp-grpc && \
-        make llama-cpp-rpc-server; \
-    fi  
-EOT
-
 
 # Copy libraries using a script to handle architecture differences
 RUN make -C /LocalAI/backend/cpp/llama-cpp package

backend/backend.proto

@@ -20,7 +20,6 @@ service Backend {
   rpc SoundGeneration(SoundGenerationRequest) returns (Result) {}
   rpc TokenizeString(PredictOptions) returns (TokenizationResponse) {}
   rpc Status(HealthMessage) returns (StatusResponse) {}
-  rpc Detect(DetectOptions) returns (DetectResponse) {}
 
   rpc StoresSet(StoresSetOptions) returns (Result) {}
   rpc StoresDelete(StoresDeleteOptions) returns (Result) {}
@@ -305,9 +304,6 @@ message GenerateImageRequest {
   // Diffusers
   string EnableParameters = 10;
   int32 CLIPSkip = 11;
-  
-  // Reference images for models that support them (e.g., Flux Kontext)
-  repeated string ref_images = 12;
 }
 
 message GenerateVideoRequest {
@@ -380,20 +376,3 @@ message Message {
   string role = 1;
   string content = 2;
 }
-
-message DetectOptions {
-  string src = 1;
-}
-
-message Detection {
-  float x = 1;
-  float y = 2;
-  float width = 3;
-  float height = 4;
-  float confidence = 5;
-  string class_name = 6;
-}
-
-message DetectResponse {
-  repeated Detection Detections = 1;
-}

backend/cpp/llama-cpp/CMakeLists.txt

@@ -17,6 +17,8 @@ if (${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
     include_directories("${HOMEBREW_DEFAULT_PREFIX}/include")
 endif()
 
+set(Protobuf_USE_STATIC_LIBS OFF)
+set(gRPC_USE_STATIC_LIBS OFF)
 find_package(absl CONFIG REQUIRED)
 find_package(Protobuf CONFIG REQUIRED)
 find_package(gRPC CONFIG REQUIRED)

backend/cpp/llama-cpp/Makefile

@@ -1,5 +1,5 @@
 
-LLAMA_VERSION?=daf2dd788066b8b239cb7f68210e090c2124c199
+LLAMA_VERSION?=acd6cb1c41676f6bbb25c2a76fa5abeb1719301e
 LLAMA_REPO?=https://github.com/ggerganov/llama.cpp
 
 CMAKE_ARGS?=
@@ -7,10 +7,9 @@ BUILD_TYPE?=
 NATIVE?=false
 ONEAPI_VARS?=/opt/intel/oneapi/setvars.sh
 TARGET?=--target grpc-server
-JOBS?=$(shell nproc)
 
 # Disable Shared libs as we are linking on static gRPC and we can't mix shared and static
-CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF -DLLAMA_CURL=OFF
+CMAKE_ARGS+=-DBUILD_SHARED_LIBS=ON -DLLAMA_CURL=OFF -DGGML_CPU_ALL_VARIANTS=ON -DGGML_BACKEND_DL=ON
 
 CURRENT_MAKEFILE_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))
 ifeq ($(NATIVE),false)
@@ -26,7 +25,7 @@ else ifeq ($(BUILD_TYPE),openblas)
 # If build type is clblas (openCL) we set -DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
 else ifeq ($(BUILD_TYPE),clblas)
 	CMAKE_ARGS+=-DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
-# If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++
+# If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++ 
 else ifeq ($(BUILD_TYPE),hipblas)
 	ROCM_HOME ?= /opt/rocm
 	ROCM_PATH ?= /opt/rocm
@@ -90,33 +89,12 @@ else
 	LLAMA_VERSION=$(LLAMA_VERSION) $(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../$(VARIANT) grpc-server
 endif
 
-llama-cpp-avx2: llama.cpp
+llama-cpp: llama.cpp
-	cp -rf $(CURRENT_MAKEFILE_DIR)/../llama-cpp $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx2-build
+	cp -rf $(CURRENT_MAKEFILE_DIR)/../llama-cpp $(CURRENT_MAKEFILE_DIR)/../llama-cpp-build
-	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx2-build purge
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../llama-cpp-build purge
-	$(info ${GREEN}I llama-cpp build info:avx2${RESET})
+	$(info ${GREEN}I llama-cpp build info:${RESET})
-	CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=on -DGGML_AVX2=on -DGGML_AVX512=off -DGGML_FMA=on -DGGML_F16C=on" $(MAKE) VARIANT="llama-cpp-avx2-build" build-llama-cpp-grpc-server
+	CMAKE_ARGS="$(CMAKE_ARGS)" $(MAKE) VARIANT="llama-cpp-build" build-llama-cpp-grpc-server
-	cp -rfv $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx2-build/grpc-server llama-cpp-avx2
+	cp -rfv $(CURRENT_MAKEFILE_DIR)/../llama-cpp-build/grpc-server llama-cpp
-
-llama-cpp-avx512: llama.cpp
-	cp -rf $(CURRENT_MAKEFILE_DIR)/../llama-cpp $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx512-build
-	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx512-build purge
-	$(info ${GREEN}I llama-cpp build info:avx512${RESET})
-	CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=on -DGGML_AVX2=off -DGGML_AVX512=on -DGGML_FMA=on -DGGML_F16C=on" $(MAKE) VARIANT="llama-cpp-avx512-build" build-llama-cpp-grpc-server
-	cp -rfv $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx512-build/grpc-server llama-cpp-avx512
-
-llama-cpp-avx: llama.cpp
-	cp -rf $(CURRENT_MAKEFILE_DIR)/../llama-cpp $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx-build
-	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx-build purge
-	$(info ${GREEN}I llama-cpp build info:avx${RESET})
-	CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=on -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off" $(MAKE) VARIANT="llama-cpp-avx-build" build-llama-cpp-grpc-server
-	cp -rfv $(CURRENT_MAKEFILE_DIR)/../llama-cpp-avx-build/grpc-server llama-cpp-avx
-
-llama-cpp-fallback: llama.cpp
-	cp -rf $(CURRENT_MAKEFILE_DIR)/../llama-cpp $(CURRENT_MAKEFILE_DIR)/../llama-cpp-fallback-build
-	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../llama-cpp-fallback-build purge
-	$(info ${GREEN}I llama-cpp build info:fallback${RESET})
-	CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off" $(MAKE) VARIANT="llama-cpp-fallback-build" build-llama-cpp-grpc-server
-	cp -rfv $(CURRENT_MAKEFILE_DIR)/../llama-cpp-fallback-build/grpc-server llama-cpp-fallback
 
 llama-cpp-grpc: llama.cpp
 	cp -rf $(CURRENT_MAKEFILE_DIR)/../llama-cpp $(CURRENT_MAKEFILE_DIR)/../llama-cpp-grpc-build
@@ -161,8 +139,8 @@ grpc-server: llama.cpp llama.cpp/tools/grpc-server
 	@echo "Building grpc-server with $(BUILD_TYPE) build type and $(CMAKE_ARGS)"
 ifneq (,$(findstring sycl,$(BUILD_TYPE)))
 	+bash -c "source $(ONEAPI_VARS); \
-	cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && cmake --build . --config Release -j $(JOBS) $(TARGET)"
+	cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && cmake --build . --config Release $(TARGET)"
 else
-	+cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && cmake --build . --config Release -j $(JOBS) $(TARGET)
+	+cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && cmake --build . --config Release $(TARGET)
 endif
 	cp llama.cpp/build/bin/grpc-server .

backend/cpp/llama-cpp/run.sh

@@ -6,34 +6,9 @@ CURDIR=$(dirname "$(realpath $0)")
 
 cd /
 
-echo "CPU info:"
+BINARY=llama-cpp
-grep -e "model\sname" /proc/cpuinfo | head -1
-grep -e "flags" /proc/cpuinfo | head -1
-
-BINARY=llama-cpp-fallback
-
-if grep -q -e "\savx\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX    found OK"
-	if [ -e $CURDIR/llama-cpp-avx ]; then
-		BINARY=llama-cpp-avx
-	fi
-fi
-
-if grep -q -e "\savx2\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX2   found OK"
-	if [ -e $CURDIR/llama-cpp-avx2 ]; then
-		BINARY=llama-cpp-avx2
-	fi
-fi
-
-# Check avx 512
-if grep -q -e "\savx512f\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX512F found OK"
-	if [ -e $CURDIR/llama-cpp-avx512 ]; then
-		BINARY=llama-cpp-avx512
-	fi
-fi
 
+## P2P/GRPC mode
 if [ -n "$LLAMACPP_GRPC_SERVERS" ]; then
 	if [ -e $CURDIR/llama-cpp-grpc ]; then
 		BINARY=llama-cpp-grpc
@@ -56,6 +31,3 @@ fi
 
 echo "Using binary: $BINARY"
 exec $CURDIR/$BINARY "$@"
-
-# In case we fail execing, just run fallback
-exec $CURDIR/llama-cpp-fallback "$@"

backend/go/stablediffusion-ggml/Makefile

@@ -18,8 +18,8 @@ GO_TAGS?=
 LD_FLAGS?=
 
 # stablediffusion.cpp (ggml)
-STABLEDIFFUSION_GGML_REPO?=https://github.com/leejet/stable-diffusion.cpp
+STABLEDIFFUSION_GGML_REPO?=https://github.com/richiejp/stable-diffusion.cpp
-STABLEDIFFUSION_GGML_VERSION?=f6b9aa1a4373e322ff12c15b8a0749e6dd6f0253
+STABLEDIFFUSION_GGML_VERSION?=53e3b17eb3d0b5760ced06a1f98320b68b34aaae
 
 # Disable Shared libs as we are linking on static gRPC and we can't mix shared and static
 CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
@@ -91,18 +91,23 @@ endif
 # (ggml can have different backends cpu, cuda, etc., each backend generates a .a archive)
 GGML_ARCHIVE_DIR := build/ggml/src/
 ALL_ARCHIVES := $(shell find $(GGML_ARCHIVE_DIR) -type f -name '*.a')
-ALL_OBJS := $(shell find $(GGML_ARCHIVE_DIR) -type f -name '*.o')
 
 # Name of the single merged library
 COMBINED_LIB := libggmlall.a
 
-# Instead of using the archives generated by GGML, use the object files directly to avoid overwriting objects with the same base name
+# Rule to merge all the .a files into one
 $(COMBINED_LIB): $(ALL_ARCHIVES)
-	@echo "Merging all .o into $(COMBINED_LIB): $(ALL_OBJS)"
+	@echo "Merging all .a into $(COMBINED_LIB)"
 	rm -f $@
-	ar -qc $@ $(ALL_OBJS)
+	mkdir -p merge-tmp
+	for a in $(ALL_ARCHIVES); do \
+		( cd merge-tmp && ar x ../$$a ); \
+	done
+	( cd merge-tmp && ar rcs ../$@ *.o )
 	# Ensure we have a proper index
 	ranlib $@
+	# Clean up
+	rm -rf merge-tmp
 
 build/libstable-diffusion.a:
 	@echo "Building SD with $(BUILD_TYPE) build type and $(CMAKE_ARGS)"

backend/go/stablediffusion-ggml/gosd.cpp

@@ -5,7 +5,6 @@
 #include <random>
 #include <string>
 #include <vector>
-#include <filesystem>
 #include "gosd.h"
 
 // #include "preprocessing.hpp"
@@ -54,43 +53,9 @@ sd_ctx_t* sd_c;
 
 sample_method_t sample_method;
 
-// Copied from the upstream CLI
+int load_model(char *model, char* options[], int threads, int diff) {
-void sd_log_cb(enum sd_log_level_t level, const char* log, void* data) {
-    //SDParams* params = (SDParams*)data;
-    const char* level_str;
-
-    if (!log /*|| (!params->verbose && level <= SD_LOG_DEBUG)*/) {
-        return;
-    }
-
-    switch (level) {
-        case SD_LOG_DEBUG:
-            level_str = "DEBUG";
-            break;
-        case SD_LOG_INFO:
-            level_str = "INFO";
-            break;
-        case SD_LOG_WARN:
-            level_str = "WARN";
-            break;
-        case SD_LOG_ERROR:
-            level_str = "ERROR";
-            break;
-        default: /* Potential future-proofing */
-            level_str = "?????";
-            break;
-    }
-
-    fprintf(stderr, "[%-5s] ", level_str);
-    fputs(log, stderr);
-    fflush(stderr);
-}
-
-int load_model(char *model, char *model_path, char* options[], int threads, int diff) {
     fprintf (stderr, "Loading model!\n");
 
-    sd_set_log_callback(sd_log_cb, NULL);
-
     char *stableDiffusionModel = "";
     if (diff == 1 ) {
         stableDiffusionModel = model;
@@ -104,10 +69,6 @@ int load_model(char *model, char *model_path, char* options[], int threads, int
     char *vae_path  = "";
     char *scheduler = "";
     char *sampler = "";
-    char *lora_dir = model_path;
-    bool lora_dir_allocated = false;
-
-    fprintf(stderr, "parsing options\n");
 
     // If options is not NULL, parse options
     for (int i = 0; options[i] != NULL; i++) {
@@ -135,29 +96,12 @@ int load_model(char *model, char *model_path, char* options[], int threads, int
         if (!strcmp(optname, "sampler")) {
             sampler = optval;
         }
-        if (!strcmp(optname, "lora_dir")) {
-            // Path join with model dir
-            if (model_path && strlen(model_path) > 0) {
-                std::filesystem::path model_path_str(model_path);
-                std::filesystem::path lora_path(optval);
-                std::filesystem::path full_lora_path = model_path_str / lora_path;
-                lora_dir = strdup(full_lora_path.string().c_str());
-                lora_dir_allocated = true;
-                fprintf(stderr, "Lora dir resolved to: %s\n", lora_dir);
-            } else {
-                lora_dir = optval;
-                fprintf(stderr, "No model path provided, using lora dir as-is: %s\n", lora_dir);
-            }
-        }
     }
 
-    fprintf(stderr, "parsed options\n");
-
     int sample_method_found = -1;
-    for (int m = 0; m < SAMPLE_METHOD_COUNT; m++) {
+    for (int m = 0; m < N_SAMPLE_METHODS; m++) {
         if (!strcmp(sampler, sample_method_str[m])) {
             sample_method_found = m;
-            fprintf(stderr, "Found sampler: %s\n", sampler);
         }
     }
     if (sample_method_found == -1) {
@@ -167,7 +111,7 @@ int load_model(char *model, char *model_path, char* options[], int threads, int
     sample_method = (sample_method_t)sample_method_found;
 
     int schedule_found            = -1;
-    for (int d = 0; d < SCHEDULE_COUNT; d++) {
+    for (int d = 0; d < N_SCHEDULES; d++) {
         if (!strcmp(scheduler, schedule_str[d])) {
             schedule_found = d;
                 fprintf (stderr, "Found scheduler: %s\n", scheduler);
@@ -181,50 +125,43 @@ int load_model(char *model, char *model_path, char* options[], int threads, int
     }
 
     schedule_t schedule = (schedule_t)schedule_found;
-
+    
     fprintf (stderr, "Creating context\n");
-    sd_ctx_params_t ctx_params;
+    sd_ctx_t* sd_ctx = new_sd_ctx(model,
-    sd_ctx_params_init(&ctx_params);
+                                  clip_l_path,
-    ctx_params.model_path = model;
+                                  clip_g_path,
-    ctx_params.clip_l_path = clip_l_path;
+                                  t5xxl_path,
-    ctx_params.clip_g_path = clip_g_path;
+                                  stableDiffusionModel,
-    ctx_params.t5xxl_path = t5xxl_path;
+                                  vae_path,
-    ctx_params.diffusion_model_path = stableDiffusionModel;
+                                  "",
-    ctx_params.vae_path = vae_path;
+                                  "",
-    ctx_params.taesd_path = "";
+                                  "",
-    ctx_params.control_net_path = "";
+                                  "",
-    ctx_params.lora_model_dir = lora_dir;
+                                  "",
-    ctx_params.embedding_dir = "";
+                                  false,
-    ctx_params.stacked_id_embed_dir = "";
+                                  false,
-    ctx_params.vae_decode_only = false;
+                                  false,
-    ctx_params.vae_tiling = false;
+                                  threads,
-    ctx_params.free_params_immediately = false;
+                                  SD_TYPE_COUNT,
-    ctx_params.n_threads = threads;
+                                  STD_DEFAULT_RNG,
-    ctx_params.rng_type = STD_DEFAULT_RNG;
+                                  schedule,
-    ctx_params.schedule = schedule;
+                                  false,
-    sd_ctx_t* sd_ctx = new_sd_ctx(&ctx_params);
+                                  false,
+                                  false,
+                                  false);
 
     if (sd_ctx == NULL) {
         fprintf (stderr, "failed loading model (generic error)\n");
-        // Clean up allocated memory
-        if (lora_dir_allocated && lora_dir) {
-            free(lora_dir);
-        }
         return 1;
     }
     fprintf (stderr, "Created context: OK\n");
 
     sd_c = sd_ctx;
 
-    // Clean up allocated memory
-    if (lora_dir_allocated && lora_dir) {
-        free(lora_dir);
-    }
-
     return 0;
 }
 
-int gen_image(char *text, char *negativeText, int width, int height, int steps, int seed , char *dst, float cfg_scale, char *src_image, float strength, char *mask_image, char **ref_images, int ref_images_count) {
+int gen_image(char *text, char *negativeText, int width, int height, int steps, int seed , char *dst, float cfg_scale) {
 
     sd_image_t* results;
 
@@ -232,202 +169,37 @@ int gen_image(char *text, char *negativeText, int width, int height, int steps,
 
     fprintf (stderr, "Generating image\n");
 
-    sd_img_gen_params_t p;
+    results = txt2img(sd_c,
-    sd_img_gen_params_init(&p);
+                            text,
-
+                            negativeText,
-    p.prompt = text;
+                            -1, //clip_skip
-    p.negative_prompt = negativeText;
+                            cfg_scale, // sfg_scale
-    p.guidance.txt_cfg = cfg_scale;
+                            3.5f,
-    p.guidance.slg.layers = skip_layers.data();
+			    0, // eta
-    p.guidance.slg.layer_count = skip_layers.size();
+                            width,
-    p.width = width;
+                            height,
-    p.height = height;
+                            sample_method, 
-    p.sample_method = sample_method;
+                            steps,
-    p.sample_steps = steps;
+                            seed,
-    p.seed = seed;
+                            1,
-    p.input_id_images_path = "";
+                            NULL,
-
+                            0.9f,
-    // Handle input image for img2img
+                            20.f,
-    bool has_input_image = (src_image != NULL && strlen(src_image) > 0);
+                            false,
-    bool has_mask_image = (mask_image != NULL && strlen(mask_image) > 0);
+                            "",
-    
+                            skip_layers.data(),
-    uint8_t* input_image_buffer = NULL;
+                            skip_layers.size(),
-    uint8_t* mask_image_buffer = NULL;
+                            0,
-    std::vector<uint8_t> default_mask_image_vec;
+                            0.01,
-    
+                            0.2);
-    if (has_input_image) {
-        fprintf(stderr, "Loading input image: %s\n", src_image);
-        
-        int c = 0;
-        int img_width = 0;
-        int img_height = 0;
-        input_image_buffer = stbi_load(src_image, &img_width, &img_height, &c, 3);
-        if (input_image_buffer == NULL) {
-            fprintf(stderr, "Failed to load input image from '%s'\n", src_image);
-            return 1;
-        }
-        if (c < 3) {
-            fprintf(stderr, "Input image must have at least 3 channels, got %d\n", c);
-            free(input_image_buffer);
-            return 1;
-        }
-        
-        // Resize input image if dimensions don't match
-        if (img_width != width || img_height != height) {
-            fprintf(stderr, "Resizing input image from %dx%d to %dx%d\n", img_width, img_height, width, height);
-            
-            uint8_t* resized_image_buffer = (uint8_t*)malloc(height * width * 3);
-            if (resized_image_buffer == NULL) {
-                fprintf(stderr, "Failed to allocate memory for resized image\n");
-                free(input_image_buffer);
-                return 1;
-            }
-            
-            stbir_resize(input_image_buffer, img_width, img_height, 0,
-                         resized_image_buffer, width, height, 0, STBIR_TYPE_UINT8,
-                         3, STBIR_ALPHA_CHANNEL_NONE, 0,
-                         STBIR_EDGE_CLAMP, STBIR_EDGE_CLAMP,
-                         STBIR_FILTER_BOX, STBIR_FILTER_BOX,
-                         STBIR_COLORSPACE_SRGB, nullptr);
-            
-            free(input_image_buffer);
-            input_image_buffer = resized_image_buffer;
-        }
-        
-        p.init_image = {(uint32_t)width, (uint32_t)height, 3, input_image_buffer};
-        p.strength = strength;
-        fprintf(stderr, "Using img2img with strength: %.2f\n", strength);
-    } else {
-        // No input image, use empty image for text-to-image
-        p.init_image = {(uint32_t)width, (uint32_t)height, 3, NULL};
-        p.strength = 0.0f;
-    }
-    
-    // Handle mask image for inpainting
-    if (has_mask_image) {
-        fprintf(stderr, "Loading mask image: %s\n", mask_image);
-        
-        int c = 0;
-        int mask_width = 0;
-        int mask_height = 0;
-        mask_image_buffer = stbi_load(mask_image, &mask_width, &mask_height, &c, 1);
-        if (mask_image_buffer == NULL) {
-            fprintf(stderr, "Failed to load mask image from '%s'\n", mask_image);
-            if (input_image_buffer) free(input_image_buffer);
-            return 1;
-        }
-        
-        // Resize mask if dimensions don't match
-        if (mask_width != width || mask_height != height) {
-            fprintf(stderr, "Resizing mask image from %dx%d to %dx%d\n", mask_width, mask_height, width, height);
-            
-            uint8_t* resized_mask_buffer = (uint8_t*)malloc(height * width);
-            if (resized_mask_buffer == NULL) {
-                fprintf(stderr, "Failed to allocate memory for resized mask\n");
-                free(mask_image_buffer);
-                if (input_image_buffer) free(input_image_buffer);
-                return 1;
-            }
-            
-            stbir_resize(mask_image_buffer, mask_width, mask_height, 0,
-                         resized_mask_buffer, width, height, 0, STBIR_TYPE_UINT8,
-                         1, STBIR_ALPHA_CHANNEL_NONE, 0,
-                         STBIR_EDGE_CLAMP, STBIR_EDGE_CLAMP,
-                         STBIR_FILTER_BOX, STBIR_FILTER_BOX,
-                         STBIR_COLORSPACE_SRGB, nullptr);
-            
-            free(mask_image_buffer);
-            mask_image_buffer = resized_mask_buffer;
-        }
-        
-        p.mask_image = {(uint32_t)width, (uint32_t)height, 1, mask_image_buffer};
-        fprintf(stderr, "Using inpainting with mask\n");
-    } else {
-        // No mask image, create default full mask
-        default_mask_image_vec.resize(width * height, 255);
-        p.mask_image = {(uint32_t)width, (uint32_t)height, 1, default_mask_image_vec.data()};
-    }
-
-    // Handle reference images
-    std::vector<sd_image_t> ref_images_vec;
-    std::vector<uint8_t*> ref_image_buffers;
-    
-    if (ref_images_count > 0 && ref_images != NULL) {
-        fprintf(stderr, "Loading %d reference images\n", ref_images_count);
-        
-        for (int i = 0; i < ref_images_count; i++) {
-            if (ref_images[i] == NULL || strlen(ref_images[i]) == 0) {
-                continue;
-            }
-            
-            fprintf(stderr, "Loading reference image %d: %s\n", i + 1, ref_images[i]);
-            
-            int c = 0;
-            int ref_width = 0;
-            int ref_height = 0;
-            uint8_t* ref_image_buffer = stbi_load(ref_images[i], &ref_width, &ref_height, &c, 3);
-            if (ref_image_buffer == NULL) {
-                fprintf(stderr, "Failed to load reference image from '%s'\n", ref_images[i]);
-                continue;
-            }
-            if (c < 3) {
-                fprintf(stderr, "Reference image must have at least 3 channels, got %d\n", c);
-                free(ref_image_buffer);
-                continue;
-            }
-            
-            // Resize reference image if dimensions don't match
-            if (ref_width != width || ref_height != height) {
-                fprintf(stderr, "Resizing reference image from %dx%d to %dx%d\n", ref_width, ref_height, width, height);
-                
-                uint8_t* resized_ref_buffer = (uint8_t*)malloc(height * width * 3);
-                if (resized_ref_buffer == NULL) {
-                    fprintf(stderr, "Failed to allocate memory for resized reference image\n");
-                    free(ref_image_buffer);
-                    continue;
-                }
-                
-                stbir_resize(ref_image_buffer, ref_width, ref_height, 0,
-                             resized_ref_buffer, width, height, 0, STBIR_TYPE_UINT8,
-                             3, STBIR_ALPHA_CHANNEL_NONE, 0,
-                             STBIR_EDGE_CLAMP, STBIR_EDGE_CLAMP,
-                             STBIR_FILTER_BOX, STBIR_FILTER_BOX,
-                             STBIR_COLORSPACE_SRGB, nullptr);
-                
-                free(ref_image_buffer);
-                ref_image_buffer = resized_ref_buffer;
-            }
-            
-            ref_image_buffers.push_back(ref_image_buffer);
-            ref_images_vec.push_back({(uint32_t)width, (uint32_t)height, 3, ref_image_buffer});
-        }
-        
-        if (!ref_images_vec.empty()) {
-            p.ref_images = ref_images_vec.data();
-            p.ref_images_count = ref_images_vec.size();
-            fprintf(stderr, "Using %zu reference images\n", ref_images_vec.size());
-        }
-    }
-
-    results = generate_image(sd_c, &p);
 
     if (results == NULL) {
         fprintf (stderr, "NO results\n");
-        if (input_image_buffer) free(input_image_buffer);
-        if (mask_image_buffer) free(mask_image_buffer);
-        for (auto buffer : ref_image_buffers) {
-            if (buffer) free(buffer);
-        }
         return 1;
     }
 
     if (results[0].data == NULL) {
         fprintf (stderr, "Results with no data\n");
-        if (input_image_buffer) free(input_image_buffer);
-        if (mask_image_buffer) free(mask_image_buffer);
-        for (auto buffer : ref_image_buffers) {
-            if (buffer) free(buffer);
-        }
         return 1;
     }
 
@@ -443,15 +215,11 @@ int gen_image(char *text, char *negativeText, int width, int height, int steps,
                        results[0].data, 0, NULL);
     fprintf (stderr, "Saved resulting image to '%s'\n", dst);
 
-    // Clean up
+    // TODO: free results. Why does it crash?
+
     free(results[0].data);
     results[0].data = NULL;
     free(results);
-    if (input_image_buffer) free(input_image_buffer);
-    if (mask_image_buffer) free(mask_image_buffer);
-    for (auto buffer : ref_image_buffers) {
-        if (buffer) free(buffer);
-    }
     fprintf (stderr, "gen_image is done", dst);
 
     return 0;

backend/go/stablediffusion-ggml/gosd.go

@@ -29,21 +29,16 @@ func (sd *SDGGML) Load(opts *pb.ModelOptions) error {
 
 	sd.threads = int(opts.Threads)
 
-	modelPath := opts.ModelPath
-
 	modelFile := C.CString(opts.ModelFile)
 	defer C.free(unsafe.Pointer(modelFile))
 
-	modelPathC := C.CString(modelPath)
-	defer C.free(unsafe.Pointer(modelPathC))
-
 	var options **C.char
 	// prepare the options array to pass to C
 
 	size := C.size_t(unsafe.Sizeof((*C.char)(nil)))
 	length := C.size_t(len(opts.Options))
-	options = (**C.char)(C.malloc((length + 1) * size))
+	options = (**C.char)(C.malloc(length * size))
-	view := (*[1 << 30]*C.char)(unsafe.Pointer(options))[0 : len(opts.Options)+1 : len(opts.Options)+1]
+	view := (*[1 << 30]*C.char)(unsafe.Pointer(options))[0:len(opts.Options):len(opts.Options)]
 
 	var diffusionModel int
 
@@ -71,11 +66,10 @@ func (sd *SDGGML) Load(opts *pb.ModelOptions) error {
 	for i, x := range oo {
 		view[i] = C.CString(x)
 	}
-	view[len(oo)] = nil
 
 	sd.cfgScale = opts.CFGScale
 
-	ret := C.load_model(modelFile, modelPathC, options, C.int(opts.Threads), C.int(diffusionModel))
+	ret := C.load_model(modelFile, options, C.int(opts.Threads), C.int(diffusionModel))
 	if ret != 0 {
 		return fmt.Errorf("could not load model")
 	}
@@ -93,56 +87,7 @@ func (sd *SDGGML) GenerateImage(opts *pb.GenerateImageRequest) error {
 	negative := C.CString(opts.NegativePrompt)
 	defer C.free(unsafe.Pointer(negative))
 
-	// Handle source image path
+	ret := C.gen_image(t, negative, C.int(opts.Width), C.int(opts.Height), C.int(opts.Step), C.int(opts.Seed), dst, C.float(sd.cfgScale))
-	var srcImage *C.char
-	if opts.Src != "" {
-		srcImage = C.CString(opts.Src)
-		defer C.free(unsafe.Pointer(srcImage))
-	}
-
-	// Handle mask image path
-	var maskImage *C.char
-	if opts.EnableParameters != "" {
-		// Parse EnableParameters for mask path if provided
-		// This is a simple approach - in a real implementation you might want to parse JSON
-		if strings.Contains(opts.EnableParameters, "mask:") {
-			parts := strings.Split(opts.EnableParameters, "mask:")
-			if len(parts) > 1 {
-				maskPath := strings.TrimSpace(parts[1])
-				if maskPath != "" {
-					maskImage = C.CString(maskPath)
-					defer C.free(unsafe.Pointer(maskImage))
-				}
-			}
-		}
-	}
-
-	// Handle reference images
-	var refImages **C.char
-	var refImagesCount C.int
-	if len(opts.RefImages) > 0 {
-		refImagesCount = C.int(len(opts.RefImages))
-		// Allocate array of C strings
-		size := C.size_t(unsafe.Sizeof((*C.char)(nil)))
-		refImages = (**C.char)(C.malloc((C.size_t(len(opts.RefImages)) + 1) * size))
-		view := (*[1 << 30]*C.char)(unsafe.Pointer(refImages))[0 : len(opts.RefImages)+1 : len(opts.RefImages)+1]
-
-		for i, refImagePath := range opts.RefImages {
-			view[i] = C.CString(refImagePath)
-			defer C.free(unsafe.Pointer(view[i]))
-		}
-		view[len(opts.RefImages)] = nil
-	}
-
-	// Default strength for img2img (0.75 is a good default)
-	strength := C.float(0.75)
-	if opts.Src != "" {
-		// If we have a source image, use img2img mode
-		// You could also parse strength from EnableParameters if needed
-		strength = C.float(0.75)
-	}
-
-	ret := C.gen_image(t, negative, C.int(opts.Width), C.int(opts.Height), C.int(opts.Step), C.int(opts.Seed), dst, C.float(sd.cfgScale), srcImage, strength, maskImage, refImages, refImagesCount)
 	if ret != 0 {
 		return fmt.Errorf("inference failed")
 	}

backend/go/stablediffusion-ggml/gosd.h

@@ -1,8 +1,8 @@
 #ifdef __cplusplus
 extern "C" {
 #endif
-int load_model(char *model, char *model_path, char* options[], int threads, int diffusionModel);
+int load_model(char *model, char* options[], int threads, int diffusionModel);
-int gen_image(char *text, char *negativeText, int width, int height, int steps, int seed, char *dst, float cfg_scale, char *src_image, float strength, char *mask_image, char **ref_images, int ref_images_count);
+int gen_image(char *text, char *negativeText, int width, int height, int steps, int seed, char *dst, float cfg_scale);
 #ifdef __cplusplus
 }
 #endif

backend/go/whisper/Makefile

@@ -6,7 +6,7 @@ CMAKE_ARGS?=
 
 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggml-org/whisper.cpp
-WHISPER_CPP_VERSION?=f7502dca872866a310fe69d30b163fa87d256319
+WHISPER_CPP_VERSION?=1f5cf0b2888402d57bb17b2029b2caa97e5f3baf
 
 export WHISPER_CMAKE_ARGS?=-DBUILD_SHARED_LIBS=OFF
 export WHISPER_DIR=$(abspath ./sources/whisper.cpp)

backend/index.yaml

@@ -73,28 +73,6 @@
     nvidia-l4t: "nvidia-l4t-arm64-stablediffusion-ggml"
     # metal: "metal-stablediffusion-ggml"
     # darwin-x86: "darwin-x86-stablediffusion-ggml"
-- &rfdetr
-  name: "rfdetr"
-  alias: "rfdetr"
-  license: apache-2.0
-  icon: https://avatars.githubusercontent.com/u/53104118?s=200&v=4
-  description: |
-    RF-DETR is a real-time, transformer-based object detection model architecture developed by Roboflow and released under the Apache 2.0 license.
-    RF-DETR is the first real-time model to exceed 60 AP on the Microsoft COCO benchmark alongside competitive performance at base sizes. It also achieves state-of-the-art performance on RF100-VL, an object detection benchmark that measures model domain adaptability to real world problems. RF-DETR is fastest and most accurate for its size when compared current real-time objection models.
-    RF-DETR is small enough to run on the edge using Inference, making it an ideal model for deployments that need both strong accuracy and real-time performance.
-  urls:
-    - https://github.com/roboflow/rf-detr
-  tags:
-    - object-detection
-    - rfdetr
-    - gpu
-    - cpu
-  capabilities:
-    nvidia: "cuda12-rfdetr"
-    intel: "intel-rfdetr"
-    #amd: "rocm-rfdetr"
-    nvidia-l4t: "nvidia-l4t-arm64-rfdetr"
-    default: "cpu-rfdetr"
 - &vllm
   name: "vllm"
   license: apache-2.0
@@ -126,13 +104,13 @@
   capabilities:
     nvidia: "cuda12-vllm"
     amd: "rocm-vllm"
-    intel: "intel-vllm"
+    intel: "intel-sycl-f16-vllm"
 - &rerankers
   name: "rerankers"
   alias: "rerankers"
   capabilities:
     nvidia: "cuda12-rerankers"
-    intel: "intel-rerankers"
+    intel: "intel-sycl-f16-rerankers"
     amd: "rocm-rerankers"
 - &transformers
   name: "transformers"
@@ -149,7 +127,7 @@
     - multimodal
   capabilities:
     nvidia: "cuda12-transformers"
-    intel: "intel-transformers"
+    intel: "intel-sycl-f16-transformers"
     amd: "rocm-transformers"
 - &diffusers
   name: "diffusers"
@@ -166,7 +144,7 @@
   alias: "diffusers"
   capabilities:
     nvidia: "cuda12-diffusers"
-    intel: "intel-diffusers"
+    intel: "intel-sycl-f32-diffusers"
     amd: "rocm-diffusers"
 - &exllama2
   name: "exllama2"
@@ -182,7 +160,8 @@
   alias: "exllama2"
   capabilities:
     nvidia: "cuda12-exllama2"
-    intel: "intel-exllama2"
+    intel: "intel-sycl-f32-exllama2"
+    amd: "rocm-exllama2"
 - &faster-whisper
   icon: https://avatars.githubusercontent.com/u/1520500?s=200&v=4
   description: |
@@ -197,7 +176,7 @@
   name: "faster-whisper"
   capabilities:
     nvidia: "cuda12-faster-whisper"
-    intel: "intel-faster-whisper"
+    intel: "intel-sycl-f32-faster-whisper"
     amd: "rocm-faster-whisper"
 - &kokoro
   icon: https://avatars.githubusercontent.com/u/166769057?v=4
@@ -215,7 +194,7 @@
   name: "kokoro"
   capabilities:
     nvidia: "cuda12-kokoro"
-    intel: "intel-kokoro"
+    intel: "intel-sycl-f32-kokoro"
     amd: "rocm-kokoro"
 - &coqui
   urls:
@@ -236,7 +215,7 @@
   alias: "coqui"
   capabilities:
     nvidia: "cuda12-coqui"
-    intel: "intel-coqui"
+    intel: "intel-sycl-f32-coqui"
     amd: "rocm-coqui"
   icon: https://avatars.githubusercontent.com/u/1338804?s=200&v=4
 - &bark
@@ -252,7 +231,7 @@
   alias: "bark"
   capabilities:
     cuda: "cuda12-bark"
-    intel: "intel-bark"
+    intel: "intel-sycl-f32-bark"
     rocm: "rocm-bark"
   icon: https://avatars.githubusercontent.com/u/99442120?s=200&v=4
 - &barkcpp
@@ -279,8 +258,6 @@
   icon: https://github.com/PABannier/bark.cpp/raw/main/assets/banner.png
   name: "bark-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-bark-cpp"
-  mirrors:
-    - localai/localai-backends:latest-bark-cpp
   alias: "bark-cpp"
 - &chatterbox
   urls:
@@ -303,8 +280,6 @@
   urls:
     - https://github.com/rhasspy/piper
     - https://github.com/mudler/go-piper
-  mirrors:
-    - localai/localai-backends:latest-piper
   license: MIT
   description: |
      A fast, local neural text to speech system
@@ -317,8 +292,6 @@
   icon: https://user-images.githubusercontent.com/12515440/89997349-b3523080-dc94-11ea-9906-ca2e8bc50535.png
   urls:
     - https://github.com/snakers4/silero-vad
-  mirrors:
-    - localai/localai-backends:latest-cpu-silero-vad
   description: |
     Silero VAD: pre-trained enterprise-grade Voice Activity Detector.
     Silero VAD is a voice activity detection model that can be used to detect whether a given audio contains speech or not.
@@ -330,8 +303,6 @@
 - &local-store
   name: "local-store"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-local-store"
-  mirrors:
-    - localai/localai-backends:latest-cpu-local-store
   urls:
     - https://github.com/mudler/LocalAI
   description: |
@@ -345,8 +316,6 @@
 - &huggingface
   name: "huggingface"
   uri: "quay.io/go-skynet/local-ai-backends:latest-huggingface"
-  mirrors:
-    - localai/localai-backends:latest-huggingface
   icon: https://huggingface.co/front/assets/huggingface_logo-noborder.svg
   urls:
     - https://huggingface.co/docs/hub/en/api
@@ -359,721 +328,469 @@
 - !!merge <<: *huggingface
   name: "huggingface-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-huggingface"
-  mirrors:
-    - localai/localai-backends:master-huggingface
 - !!merge <<: *local-store
   name: "local-store-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-local-store"
-  mirrors:
-    - localai/localai-backends:master-cpu-local-store
 - !!merge <<: *silero-vad
   name: "silero-vad-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-silero-vad"
-  mirrors:
-    - localai/localai-backends:master-cpu-silero-vad
 - !!merge <<: *piper
   name: "piper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-piper"
-  mirrors:
-    - localai/localai-backends:master-piper
 ## llama-cpp
 - !!merge <<: *llamacpp
   name: "darwin-x86-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-darwin-x86-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-darwin-x86-llama-cpp
 - !!merge <<: *llamacpp
   name: "darwin-x86-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-darwin-x86-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-darwin-x86-llama-cpp
 - !!merge <<: *llamacpp
   name: "nvidia-l4t-arm64-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-arm64-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-nvidia-l4t-arm64-llama-cpp
 - !!merge <<: *llamacpp
   name: "nvidia-l4t-arm64-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-arm64-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-nvidia-l4t-arm64-llama-cpp
 - !!merge <<: *llamacpp
   name: "cpu-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-cpu-llama-cpp
 - !!merge <<: *llamacpp
   name: "cpu-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-cpu-llama-cpp
 - !!merge <<: *llamacpp
   name: "cuda11-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-llama-cpp
 - !!merge <<: *llamacpp
   name: "cuda12-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-llama-cpp
 - !!merge <<: *llamacpp
   name: "rocm-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-llama-cpp
 - !!merge <<: *llamacpp
   name: "intel-sycl-f32-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-sycl-f32-llama-cpp
 - !!merge <<: *llamacpp
   name: "intel-sycl-f16-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-sycl-f16-llama-cpp
 - !!merge <<: *llamacpp
   name: "vulkan-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-vulkan-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-gpu-vulkan-llama-cpp
 - !!merge <<: *llamacpp
   name: "vulkan-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-vulkan-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-gpu-vulkan-llama-cpp
 - !!merge <<: *llamacpp
   name: "metal-llama-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-llama-cpp"
-  mirrors:
-    - localai/localai-backends:latest-metal-darwin-arm64-llama-cpp
 - !!merge <<: *llamacpp
   name: "metal-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-metal-darwin-arm64-llama-cpp
 - !!merge <<: *llamacpp
   name: "cuda11-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-llama-cpp
 - !!merge <<: *llamacpp
   name: "cuda12-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-llama-cpp
 - !!merge <<: *llamacpp
   name: "rocm-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-llama-cpp
 - !!merge <<: *llamacpp
   name: "intel-sycl-f32-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-gpu-intel-sycl-f32-llama-cpp
 - !!merge <<: *llamacpp
   name: "intel-sycl-f16-llama-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-llama-cpp"
-  mirrors:
-    - localai/localai-backends:master-gpu-intel-sycl-f16-llama-cpp
 ## whisper
 - !!merge <<: *whispercpp
   name: "nvidia-l4t-arm64-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-arm64-whisper"
-  mirrors:
-    - localai/localai-backends:latest-nvidia-l4t-arm64-whisper
 - !!merge <<: *whispercpp
   name: "nvidia-l4t-arm64-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-arm64-whisper"
-  mirrors:
-    - localai/localai-backends:master-nvidia-l4t-arm64-whisper
 - !!merge <<: *whispercpp
   name: "cpu-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-whisper"
-  mirrors:
-    - localai/localai-backends:latest-cpu-whisper
 - !!merge <<: *whispercpp
   name: "cpu-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-whisper"
-  mirrors:
-    - localai/localai-backends:master-cpu-whisper
 - !!merge <<: *whispercpp
   name: "cuda11-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-whisper"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-whisper
 - !!merge <<: *whispercpp
   name: "cuda12-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-whisper"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-whisper
 - !!merge <<: *whispercpp
   name: "rocm-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-whisper"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-whisper
 - !!merge <<: *whispercpp
   name: "intel-sycl-f32-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-whisper"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-sycl-f32-whisper
 - !!merge <<: *whispercpp
   name: "intel-sycl-f16-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-whisper"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-sycl-f16-whisper
 - !!merge <<: *whispercpp
   name: "vulkan-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-vulkan-whisper"
-  mirrors:
-    - localai/localai-backends:latest-gpu-vulkan-whisper
 - !!merge <<: *whispercpp
   name: "vulkan-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-vulkan-whisper"
-  mirrors:
-    - localai/localai-backends:master-gpu-vulkan-whisper
 - !!merge <<: *whispercpp
   name: "metal-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-whisper"
-  mirrors:
-    - localai/localai-backends:latest-metal-darwin-arm64-whisper
 - !!merge <<: *whispercpp
   name: "metal-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-whisper"
-  mirrors:
-    - localai/localai-backends:master-metal-darwin-arm64-whisper
 - !!merge <<: *whispercpp
   name: "cuda11-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-whisper"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-whisper
 - !!merge <<: *whispercpp
   name: "cuda12-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-whisper"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-whisper
 - !!merge <<: *whispercpp
   name: "rocm-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-whisper"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-whisper
 - !!merge <<: *whispercpp
   name: "intel-sycl-f32-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-whisper"
-  mirrors:
-    - localai/localai-backends:master-gpu-intel-sycl-f32-whisper
 - !!merge <<: *whispercpp
   name: "intel-sycl-f16-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-whisper"
-  mirrors:
-    - localai/localai-backends:master-gpu-intel-sycl-f16-whisper
 ## stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "cpu-stablediffusion-ggml"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:latest-cpu-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "cpu-stablediffusion-ggml-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:master-cpu-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "vulkan-stablediffusion-ggml"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-vulkan-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:latest-gpu-vulkan-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "vulkan-stablediffusion-ggml-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-vulkan-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:master-gpu-vulkan-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "cuda12-stablediffusion-ggml"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "intel-sycl-f32-stablediffusion-ggml"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-stablediffusion-ggml"
 - !!merge <<: *stablediffusionggml
   name: "intel-sycl-f16-stablediffusion-ggml"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-sycl-f16-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "cuda11-stablediffusion-ggml"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "cuda12-stablediffusion-ggml-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "intel-sycl-f32-stablediffusion-ggml-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:master-gpu-intel-sycl-f32-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "intel-sycl-f16-stablediffusion-ggml-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:master-gpu-intel-sycl-f16-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "cuda11-stablediffusion-ggml-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "nvidia-l4t-arm64-stablediffusion-ggml-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-arm64-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:master-nvidia-l4t-arm64-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "nvidia-l4t-arm64-stablediffusion-ggml"
   uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-arm64-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:latest-nvidia-l4t-arm64-stablediffusion-ggml
 # vllm
 - !!merge <<: *vllm
   name: "vllm-development"
   capabilities:
     nvidia: "cuda12-vllm-development"
     amd: "rocm-vllm-development"
-    intel: "intel-vllm-development"
+    intel: "intel-sycl-f16-vllm-development"
 - !!merge <<: *vllm
   name: "cuda12-vllm"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-vllm"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-vllm
 - !!merge <<: *vllm
   name: "rocm-vllm"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-vllm"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-vllm
 - !!merge <<: *vllm
-  name: "intel-vllm"
+  name: "intel-sycl-f32-vllm"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-vllm"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-vllm"
-  mirrors:
+- !!merge <<: *vllm
-    - localai/localai-backends:latest-gpu-intel-vllm
+  name: "intel-sycl-f16-vllm"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-vllm"
 - !!merge <<: *vllm
   name: "cuda12-vllm-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-vllm"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-vllm
 - !!merge <<: *vllm
   name: "rocm-vllm-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-vllm"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-vllm
 - !!merge <<: *vllm
-  name: "intel-vllm-development"
+  name: "intel-sycl-f32-vllm-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-vllm"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-vllm"
-  mirrors:
+- !!merge <<: *vllm
-    - localai/localai-backends:master-gpu-intel-vllm
+  name: "intel-sycl-f16-vllm-development"
-# rfdetr
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-vllm"
-- !!merge <<: *rfdetr
-  name: "rfdetr-development"
-  capabilities:
-    nvidia: "cuda12-rfdetr-development"
-    intel: "intel-rfdetr-development"
-    #amd: "rocm-rfdetr-development"
-    nvidia-l4t: "nvidia-l4t-arm64-rfdetr-development"
-    default: "cpu-rfdetr-development"
-- !!merge <<: *rfdetr
-  name: "cuda12-rfdetr"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-rfdetr"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-rfdetr
-- !!merge <<: *rfdetr
-  name: "intel-rfdetr"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-rfdetr"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-rfdetr
-# - !!merge <<: *rfdetr
-#   name: "rocm-rfdetr"
-#   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-hipblas-rfdetr"
-#   mirrors:
-#     - localai/localai-backends:latest-gpu-hipblas-rfdetr
-- !!merge <<: *rfdetr
-  name: "nvidia-l4t-arm64-rfdetr"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-arm64-rfdetr"
-  mirrors:
-    - localai/localai-backends:latest-nvidia-l4t-arm64-rfdetr
-- !!merge <<: *rfdetr
-  name: "cpu-rfdetr"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-rfdetr"
-  mirrors:
-    - localai/localai-backends:latest-cpu-rfdetr
-- !!merge <<: *rfdetr
-  name: "cuda12-rfdetr-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-rfdetr"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-rfdetr
-- !!merge <<: *rfdetr
-  name: "intel-rfdetr-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-rfdetr"
-  mirrors:
-    - localai/localai-backends:master-gpu-intel-rfdetr
-# - !!merge <<: *rfdetr
-#   name: "rocm-rfdetr-development"
-#   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-hipblas-rfdetr"
-#   mirrors:
-#     - localai/localai-backends:master-gpu-hipblas-rfdetr
-- !!merge <<: *rfdetr
-  name: "cpu-rfdetr-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-cpu-rfdetr"
-  mirrors:
-    - localai/localai-backends:master-cpu-rfdetr
-- !!merge <<: *rfdetr
-  name: "intel-rfdetr"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-rfdetr"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-rfdetr
 ## Rerankers
 - !!merge <<: *rerankers
   name: "rerankers-development"
   capabilities:
     nvidia: "cuda12-rerankers-development"
-    intel: "intel-rerankers-development"
+    intel: "intel-sycl-f16-rerankers-development"
     amd: "rocm-rerankers-development"
 - !!merge <<: *rerankers
   name: "cuda11-rerankers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-rerankers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-rerankers
 - !!merge <<: *rerankers
   name: "cuda12-rerankers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-rerankers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-rerankers
 - !!merge <<: *rerankers
-  name: "intel-rerankers"
+  name: "intel-sycl-f32-rerankers"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-rerankers"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-rerankers"
-  mirrors:
+- !!merge <<: *rerankers
-    - localai/localai-backends:latest-gpu-intel-rerankers
+  name: "intel-sycl-f16-rerankers"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-rerankers"
 - !!merge <<: *rerankers
   name: "rocm-rerankers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-rerankers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-rerankers
 - !!merge <<: *rerankers
   name: "cuda11-rerankers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-rerankers"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-rerankers
 - !!merge <<: *rerankers
   name: "cuda12-rerankers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-rerankers"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-rerankers
 - !!merge <<: *rerankers
   name: "rocm-rerankers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-rerankers"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-rerankers
 - !!merge <<: *rerankers
-  name: "intel-rerankers-development"
+  name: "intel-sycl-f32-rerankers-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-rerankers"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-rerankers"
-  mirrors:
+- !!merge <<: *rerankers
-    - localai/localai-backends:master-gpu-intel-rerankers
+  name: "intel-sycl-f16-rerankers-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-rerankers"
 ## Transformers
 - !!merge <<: *transformers
   name: "transformers-development"
   capabilities:
     nvidia: "cuda12-transformers-development"
-    intel: "intel-transformers-development"
+    intel: "intel-sycl-f16-transformers-development"
     amd: "rocm-transformers-development"
 - !!merge <<: *transformers
   name: "cuda12-transformers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-transformers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-transformers
 - !!merge <<: *transformers
   name: "rocm-transformers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-transformers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-transformers
 - !!merge <<: *transformers
-  name: "intel-transformers"
+  name: "intel-sycl-f32-transformers"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-transformers"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-transformers"
-  mirrors:
+- !!merge <<: *transformers
-    - localai/localai-backends:latest-gpu-intel-transformers
+  name: "intel-sycl-f16-transformers"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-transformers"
 - !!merge <<: *transformers
   name: "cuda11-transformers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-transformers"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-transformers
 - !!merge <<: *transformers
   name: "cuda11-transformers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-transformers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-transformers
 - !!merge <<: *transformers
   name: "cuda12-transformers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-transformers"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-transformers
 - !!merge <<: *transformers
   name: "rocm-transformers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-transformers"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-transformers
 - !!merge <<: *transformers
-  name: "intel-transformers-development"
+  name: "intel-sycl-f32-transformers-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-transformers"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-transformers"
-  mirrors:
+- !!merge <<: *transformers
-    - localai/localai-backends:master-gpu-intel-transformers
+  name: "intel-sycl-f16-transformers-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-transformers"
 ## Diffusers
 - !!merge <<: *diffusers
   name: "diffusers-development"
   capabilities:
     nvidia: "cuda12-diffusers-development"
-    intel: "intel-diffusers-development"
+    intel: "intel-sycl-f32-diffusers-development"
     amd: "rocm-diffusers-development"
 - !!merge <<: *diffusers
   name: "cuda12-diffusers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-diffusers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-diffusers
 - !!merge <<: *diffusers
   name: "rocm-diffusers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-diffusers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-diffusers
 - !!merge <<: *diffusers
   name: "cuda11-diffusers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-diffusers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-diffusers
 - !!merge <<: *diffusers
-  name: "intel-diffusers"
+  name: "intel-sycl-f32-diffusers"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-diffusers"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-diffusers"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-diffusers
 - !!merge <<: *diffusers
   name: "cuda11-diffusers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-diffusers"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-diffusers
 - !!merge <<: *diffusers
   name: "cuda12-diffusers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-diffusers"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-diffusers
 - !!merge <<: *diffusers
   name: "rocm-diffusers-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-diffusers"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-diffusers
 - !!merge <<: *diffusers
-  name: "intel-diffusers-development"
+  name: "intel-sycl-f32-diffusers-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-diffusers"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-diffusers"
-  mirrors:
-    - localai/localai-backends:master-gpu-intel-diffusers
   ## exllama2
 - !!merge <<: *exllama2
   name: "exllama2-development"
   capabilities:
     nvidia: "cuda12-exllama2-development"
-    intel: "intel-exllama2-development"
+    intel: "intel-sycl-f32-exllama2-development"
+    amd: "rocm-exllama2-development"
 - !!merge <<: *exllama2
   name: "cuda11-exllama2"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-exllama2"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-exllama2
 - !!merge <<: *exllama2
   name: "cuda12-exllama2"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-exllama2"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-exllama2
 - !!merge <<: *exllama2
   name: "cuda11-exllama2-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-exllama2"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-exllama2
 - !!merge <<: *exllama2
   name: "cuda12-exllama2-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-exllama2"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-exllama2
 ## kokoro
 - !!merge <<: *kokoro
   name: "kokoro-development"
   capabilities:
     nvidia: "cuda12-kokoro-development"
-    intel: "intel-kokoro-development"
+    intel: "intel-sycl-f32-kokoro-development"
     amd: "rocm-kokoro-development"
 - !!merge <<: *kokoro
   name: "cuda11-kokoro-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-kokoro"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-kokoro
 - !!merge <<: *kokoro
   name: "cuda12-kokoro-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-kokoro"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-kokoro
 - !!merge <<: *kokoro
   name: "rocm-kokoro-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-kokoro"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-kokoro
 - !!merge <<: *kokoro
-  name: "intel-kokoro"
+  name: "sycl-f32-kokoro"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-kokoro"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-kokoro"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-kokoro
 - !!merge <<: *kokoro
-  name: "intel-kokoro-development"
+  name: "sycl-f16-kokoro"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-kokoro"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-kokoro"
-  mirrors:
+- !!merge <<: *kokoro
-    - localai/localai-backends:master-gpu-intel-kokoro
+  name: "sycl-f16-kokoro-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-kokoro"
+- !!merge <<: *kokoro
+  name: "sycl-f32-kokoro-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-kokoro"
 - !!merge <<: *kokoro
   name: "cuda11-kokoro"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-kokoro"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-kokoro
 - !!merge <<: *kokoro
   name: "cuda12-kokoro"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-kokoro"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-kokoro
 - !!merge <<: *kokoro
   name: "rocm-kokoro"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-kokoro"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-kokoro
 ## faster-whisper
 - !!merge <<: *faster-whisper
   name: "faster-whisper-development"
   capabilities:
     nvidia: "cuda12-faster-whisper-development"
-    intel: "intel-faster-whisper-development"
+    intel: "intel-sycl-f32-faster-whisper-development"
     amd: "rocm-faster-whisper-development"
 - !!merge <<: *faster-whisper
   name: "cuda11-faster-whisper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-faster-whisper"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-faster-whisper
 - !!merge <<: *faster-whisper
   name: "cuda12-faster-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-faster-whisper"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-faster-whisper
 - !!merge <<: *faster-whisper
   name: "rocm-faster-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-faster-whisper"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-faster-whisper
 - !!merge <<: *faster-whisper
-  name: "intel-faster-whisper"
+  name: "sycl-f32-faster-whisper"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-faster-whisper"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-faster-whisper"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-faster-whisper
 - !!merge <<: *faster-whisper
-  name: "intel-faster-whisper-development"
+  name: "sycl-f16-faster-whisper"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-faster-whisper"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-faster-whisper"
-  mirrors:
+- !!merge <<: *faster-whisper
-    - localai/localai-backends:master-gpu-intel-faster-whisper
+  name: "sycl-f32-faster-whisper-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-faster-whisper"
+- !!merge <<: *faster-whisper
+  name: "sycl-f16-faster-whisper-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-faster-whisper"
 ## coqui
 
 - !!merge <<: *coqui
   name: "coqui-development"
   capabilities:
     nvidia: "cuda12-coqui-development"
-    intel: "intel-coqui-development"
+    intel: "intel-sycl-f32-coqui-development"
     amd: "rocm-coqui-development"
 - !!merge <<: *coqui
   name: "cuda11-coqui"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-coqui"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-coqui
 - !!merge <<: *coqui
   name: "cuda12-coqui"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-coqui"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-coqui
 - !!merge <<: *coqui
   name: "cuda11-coqui-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-coqui"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-coqui
 - !!merge <<: *coqui
   name: "cuda12-coqui-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-coqui"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-coqui
 - !!merge <<: *coqui
   name: "rocm-coqui-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-coqui"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-coqui
 - !!merge <<: *coqui
-  name: "intel-coqui"
+  name: "sycl-f32-coqui"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-coqui"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-coqui"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-coqui
 - !!merge <<: *coqui
-  name: "intel-coqui-development"
+  name: "sycl-f16-coqui"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-coqui"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-coqui"
-  mirrors:
+- !!merge <<: *coqui
-    - localai/localai-backends:master-gpu-intel-coqui
+  name: "sycl-f32-coqui-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-coqui"
+- !!merge <<: *coqui
+  name: "sycl-f16-coqui-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-coqui"
 - !!merge <<: *coqui
   name: "rocm-coqui"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-coqui"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-coqui
 ## bark
 - !!merge <<: *bark
   name: "bark-development"
   capabilities:
     nvidia: "cuda12-bark-development"
-    intel: "intel-bark-development"
+    intel: "intel-sycl-f32-bark-development"
     amd: "rocm-bark-development"
 - !!merge <<: *bark
   name: "cuda11-bark-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-bark"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-bark
 - !!merge <<: *bark
   name: "cuda11-bark"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-bark"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-bark
 - !!merge <<: *bark
   name: "rocm-bark-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-bark"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-bark
 - !!merge <<: *bark
-  name: "intel-bark"
+  name: "sycl-f32-bark"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-bark"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f32-bark"
-  mirrors:
-    - localai/localai-backends:latest-gpu-intel-bark
 - !!merge <<: *bark
-  name: "intel-bark-development"
+  name: "sycl-f16-bark"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-bark"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-intel-sycl-f16-bark"
-  mirrors:
+- !!merge <<: *bark
-    - localai/localai-backends:master-gpu-intel-bark
+  name: "sycl-f32-bark-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f32-bark"
+- !!merge <<: *bark
+  name: "sycl-f16-bark-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-sycl-f16-bark"
 - !!merge <<: *bark
   name: "cuda12-bark"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-bark"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-bark
 - !!merge <<: *bark
   name: "rocm-bark"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-bark"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-bark
 - !!merge <<: *bark
   name: "cuda12-bark-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-bark"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-bark
 - !!merge <<: *barkcpp
   name: "bark-cpp-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-bark-cpp"
@@ -1086,20 +803,12 @@
 - !!merge <<: *chatterbox
   name: "cuda12-chatterbox-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-chatterbox"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-chatterbox
 - !!merge <<: *chatterbox
   name: "cuda11-chatterbox"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-chatterbox"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-11-chatterbox
 - !!merge <<: *chatterbox
   name: "cuda11-chatterbox-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-chatterbox"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-11-chatterbox
 - !!merge <<: *chatterbox
   name: "cuda12-chatterbox"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-chatterbox"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-chatterbox

backend/python/common/libbackend.sh

@@ -111,7 +111,7 @@ function ensureVenv() {
 #  - requirements-${BUILD_TYPE}.txt
 #  - requirements-${BUILD_PROFILE}.txt
 #
-# BUILD_PROFILE is a pore specific version of BUILD_TYPE, ex: cuda-11 or cuda-12
+# BUILD_PROFILE is a pore specific version of BUILD_TYPE, ex: cuda11 or cuda12
 # it can also include some options that we do not have BUILD_TYPES for, ex: intel
 #
 # NOTE: for BUILD_PROFILE==intel, this function does NOT automatically use the Intel python package index.

backend/python/common/template/protogen.sh

@@ -8,6 +8,4 @@ else
     source $backend_dir/../common/libbackend.sh
 fi
 
-ensureVenv
-
 python3 -m grpc_tools.protoc -I../.. -I./ --python_out=. --grpc_python_out=. backend.proto

backend/python/rfdetr/Makefile

@@ -1,20 +0,0 @@
-.DEFAULT_GOAL := install
-
-.PHONY: install
-install:
-	bash install.sh
-	$(MAKE) protogen
-
-.PHONY: protogen
-protogen: backend_pb2_grpc.py backend_pb2.py
-
-.PHONY: protogen-clean
-protogen-clean:
-	$(RM) backend_pb2_grpc.py backend_pb2.py
-
-backend_pb2_grpc.py backend_pb2.py:
-	bash protogen.sh
-
-.PHONY: clean
-clean: protogen-clean
-	rm -rf venv __pycache__

backend/python/rfdetr/backend.py

@@ -1,174 +0,0 @@
-#!/usr/bin/env python3
-"""
-gRPC server for RFDETR object detection models.
-"""
-from concurrent import futures
-
-import argparse
-import signal
-import sys
-import os
-import time
-import base64
-import backend_pb2
-import backend_pb2_grpc
-import grpc
-
-import requests
-
-import supervision as sv
-from inference import get_model
-from PIL import Image
-from io import BytesIO
-
-_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):
-    """
-    A gRPC servicer for the RFDETR backend service.
-
-    This class implements the gRPC methods for object detection using RFDETR models.
-    """
-    
-    def __init__(self):
-        self.model = None
-        self.model_name = None
-        
-    def Health(self, request, context):
-        """
-        A gRPC method that returns the health status of the backend service.
-
-        Args:
-            request: A HealthMessage object that contains the request parameters.
-            context: A grpc.ServicerContext object that provides information about the RPC.
-
-        Returns:
-            A Reply object that contains the health status of the backend service.
-        """
-        return backend_pb2.Reply(message=bytes("OK", 'utf-8'))
-
-    def LoadModel(self, request, context):
-        """
-        A gRPC method that loads a RFDETR model into memory.
-
-        Args:
-            request: A ModelOptions object that contains the model parameters.
-            context: A grpc.ServicerContext object that provides information about the RPC.
-
-        Returns:
-            A Result object that contains the result of the LoadModel operation.
-        """
-        model_name = request.Model
-        try:
-            # Load the RFDETR model
-            self.model = get_model(model_name)
-            self.model_name = model_name
-            print(f'Loaded RFDETR model: {model_name}')
-        except Exception as err:
-            return backend_pb2.Result(success=False, message=f"Failed to load model: {err}")
-
-        return backend_pb2.Result(message="Model loaded successfully", success=True)
-
-    def Detect(self, request, context):
-        """
-        A gRPC method that performs object detection on an image.
-
-        Args:
-            request: A DetectOptions object that contains the image source.
-            context: A grpc.ServicerContext object that provides information about the RPC.
-
-        Returns:
-            A DetectResponse object that contains the detection results.
-        """
-        if self.model is None:
-            print(f"Model is None")
-            return backend_pb2.DetectResponse()
-        print(f"Model is not None")
-        try:
-            print(f"Decoding image")
-            # Decode the base64 image
-            print(f"Image data: {request.src}")
-
-            image_data = base64.b64decode(request.src)
-            image = Image.open(BytesIO(image_data))
-            
-            # Perform inference
-            predictions = self.model.infer(image, confidence=0.5)[0]
-          
-            # Convert to proto format
-            proto_detections = []
-            for i in range(len(predictions.predictions)):
-                pred = predictions.predictions[i]
-                print(f"Prediction: {pred}")
-                proto_detection = backend_pb2.Detection(
-                    x=float(pred.x),
-                    y=float(pred.y),
-                    width=float(pred.width),
-                    height=float(pred.height),
-                    confidence=float(pred.confidence),
-                    class_name=pred.class_name
-                )
-                proto_detections.append(proto_detection)
-            
-            return backend_pb2.DetectResponse(Detections=proto_detections)
-        except Exception as err:
-            print(f"Detection error: {err}")
-            return backend_pb2.DetectResponse()
-
-    def Status(self, request, context):
-        """
-        A gRPC method that returns the status of the backend service.
-
-        Args:
-            request: A HealthMessage object that contains the request parameters.
-            context: A grpc.ServicerContext object that provides information about the RPC.
-
-        Returns:
-            A StatusResponse object that contains the status information.
-        """
-        state = backend_pb2.StatusResponse.READY if self.model is not None else backend_pb2.StatusResponse.UNINITIALIZED
-        return backend_pb2.StatusResponse(state=state)
-
-def serve(address):
-    server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS),
-        options=[
-            ('grpc.max_message_length', 50 * 1024 * 1024),  # 50MB
-            ('grpc.max_send_message_length', 50 * 1024 * 1024),  # 50MB
-            ('grpc.max_receive_message_length', 50 * 1024 * 1024),  # 50MB
-        ])
-    backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
-    server.add_insecure_port(address)
-    server.start()
-    print("[RFDETR] Server started. Listening on: " + address, file=sys.stderr)
-
-    # Define the signal handler function
-    def signal_handler(sig, frame):
-        print("[RFDETR] Received termination signal. Shutting down...")
-        server.stop(0)
-        sys.exit(0)
-
-    # Set the signal handlers for SIGINT and SIGTERM
-    signal.signal(signal.SIGINT, signal_handler)
-    signal.signal(signal.SIGTERM, signal_handler)
-
-    try:
-        while True:
-            time.sleep(_ONE_DAY_IN_SECONDS)
-    except KeyboardInterrupt:
-        server.stop(0)
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description="Run the RFDETR gRPC server.")
-    parser.add_argument(
-        "--addr", default="localhost:50051", help="The address to bind the server to."
-    )
-    args = parser.parse_args()
-    print(f"[RFDETR] startup: {args}", file=sys.stderr)
-    serve(args.addr)
-
-
-

backend/python/rfdetr/install.sh

@@ -1,19 +0,0 @@
-#!/bin/bash
-set -e
-
-backend_dir=$(dirname $0)
-if [ -d $backend_dir/common ]; then
-    source $backend_dir/common/libbackend.sh
-else
-    source $backend_dir/../common/libbackend.sh
-fi
-
-# This is here because the Intel pip index is broken and returns 200 status codes for every package name, it just doesn't return any package links.
-# This makes uv think that the package exists in the Intel pip index, and by default it stops looking at other pip indexes once it finds a match.
-# We need uv to continue falling through to the pypi default index to find optimum[openvino] in the pypi index
-# the --upgrade actually allows us to *downgrade* torch to the version provided in the Intel pip index
-if [ "x${BUILD_PROFILE}" == "xintel" ]; then
-    EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
-fi
-
-installRequirements

backend/python/rfdetr/protogen.sh

@@ -1,13 +0,0 @@
-#!/bin/bash
-set -e
-
-backend_dir=$(dirname $0)
-if [ -d $backend_dir/common ]; then
-    source $backend_dir/common/libbackend.sh
-else
-    source $backend_dir/../common/libbackend.sh
-fi
-
-ensureVenv
-
-python3 -m grpc_tools.protoc -I../.. -I./ --python_out=. --grpc_python_out=. backend.proto

backend/python/rfdetr/requirements-cpu.txt

@@ -1,7 +0,0 @@
-rfdetr
-opencv-python
-accelerate
-peft
-inference
-torch==2.7.1
-optimum-quanto

backend/python/rfdetr/requirements-cublas11.txt

@@ -1,8 +0,0 @@
---extra-index-url https://download.pytorch.org/whl/cu118
-torch==2.7.1+cu118
-rfdetr
-opencv-python
-accelerate
-inference
-peft
-optimum-quanto

backend/python/rfdetr/requirements-cublas12.txt

@@ -1,7 +0,0 @@
-torch==2.7.1
-rfdetr
-opencv-python
-accelerate
-inference
-peft
-optimum-quanto

backend/python/rfdetr/requirements-hipblas.txt

@@ -1,9 +0,0 @@
---extra-index-url https://download.pytorch.org/whl/rocm6.3
-torch==2.7.1+rocm6.3
-torchvision==0.22.1+rocm6.3
-rfdetr
-opencv-python
-accelerate
-inference
-peft
-optimum-quanto

backend/python/rfdetr/requirements-intel.txt

@@ -1,13 +0,0 @@
---extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
-intel-extension-for-pytorch==2.3.110+xpu
-torch==2.3.1+cxx11.abi
-torchvision==0.18.1+cxx11.abi
-oneccl_bind_pt==2.3.100+xpu
-optimum[openvino]
-setuptools
-rfdetr
-inference
-opencv-python
-accelerate
-peft
-optimum-quanto

backend/python/rfdetr/requirements.txt

@@ -1,3 +0,0 @@
-grpcio==1.71.0
-protobuf
-grpcio-tools

backend/python/rfdetr/run.sh

@@ -1,9 +0,0 @@
-#!/bin/bash
-backend_dir=$(dirname $0)
-if [ -d $backend_dir/common ]; then
-    source $backend_dir/common/libbackend.sh
-else
-    source $backend_dir/../common/libbackend.sh
-fi
-
-startBackend $@

backend/python/rfdetr/test.sh

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

core/application/application.go

@@ -2,8 +2,8 @@ package application
 
 import (
 	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/core/templates"
 	"github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/templates"
 )
 
 type Application struct {

core/application/startup.go

@@ -10,8 +10,8 @@ import (
 	"github.com/mudler/LocalAI/core/services"
 	"github.com/mudler/LocalAI/internal"
 
-	coreStartup "github.com/mudler/LocalAI/core/startup"
 	"github.com/mudler/LocalAI/pkg/model"
+	pkgStartup "github.com/mudler/LocalAI/pkg/startup"
 	"github.com/mudler/LocalAI/pkg/xsysinfo"
 	"github.com/rs/zerolog/log"
 )
@@ -55,11 +55,11 @@ func New(opts ...config.AppOption) (*Application, error) {
 		}
 	}
 
-	if err := coreStartup.InstallModels(options.Galleries, options.BackendGalleries, options.ModelPath, options.BackendsPath, options.EnforcePredownloadScans, options.AutoloadBackendGalleries, nil, options.ModelsURL...); err != nil {
+	if err := pkgStartup.InstallModels(options.Galleries, options.BackendGalleries, options.ModelPath, options.BackendsPath, options.EnforcePredownloadScans, options.AutoloadBackendGalleries, nil, options.ModelsURL...); err != nil {
 		log.Error().Err(err).Msg("error installing models")
 	}
 
-	if err := coreStartup.InstallExternalBackends(options.BackendGalleries, options.BackendsPath, nil, options.ExternalBackends...); err != nil {
+	if err := pkgStartup.InstallExternalBackends(options.BackendGalleries, options.BackendsPath, nil, options.ExternalBackends...); err != nil {
 		log.Error().Err(err).Msg("error installing external backends")
 	}
 

core/backend/detection.go

@@ -1,34 +0,0 @@
-package backend
-
-import (
-	"context"
-	"fmt"
-
-	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/pkg/grpc/proto"
-	"github.com/mudler/LocalAI/pkg/model"
-)
-
-func Detection(
-	sourceFile string,
-	loader *model.ModelLoader,
-	appConfig *config.ApplicationConfig,
-	backendConfig config.BackendConfig,
-) (*proto.DetectResponse, error) {
-	opts := ModelOptions(backendConfig, appConfig)
-	detectionModel, err := loader.Load(opts...)
-	if err != nil {
-		return nil, err
-	}
-	defer loader.Close()
-
-	if detectionModel == nil {
-		return nil, fmt.Errorf("could not load detection model")
-	}
-
-	res, err := detectionModel.Detect(context.Background(), &proto.DetectOptions{
-		Src: sourceFile,
-	})
-
-	return res, err
-}

core/backend/image.go

@@ -7,7 +7,7 @@ import (
 	model "github.com/mudler/LocalAI/pkg/model"
 )
 
-func ImageGeneration(height, width, mode, step, seed int, positive_prompt, negative_prompt, src, dst string, loader *model.ModelLoader, backendConfig config.BackendConfig, appConfig *config.ApplicationConfig, refImages []string) (func() error, error) {
+func ImageGeneration(height, width, mode, step, seed int, positive_prompt, negative_prompt, src, dst string, loader *model.ModelLoader, backendConfig config.BackendConfig, appConfig *config.ApplicationConfig) (func() error, error) {
 
 	opts := ModelOptions(backendConfig, appConfig)
 	inferenceModel, err := loader.Load(
@@ -33,7 +33,6 @@ func ImageGeneration(height, width, mode, step, seed int, positive_prompt, negat
 				Dst:              dst,
 				Src:              src,
 				EnableParameters: backendConfig.Diffusers.EnableParameters,
-				RefImages:        refImages,
 			})
 		return err
 	}

core/cli/backends.go

@@ -8,7 +8,7 @@ import (
 	"github.com/mudler/LocalAI/core/config"
 
 	"github.com/mudler/LocalAI/core/gallery"
-	"github.com/mudler/LocalAI/core/startup"
+	"github.com/mudler/LocalAI/pkg/startup"
 	"github.com/rs/zerolog/log"
 	"github.com/schollz/progressbar/v3"
 )

core/cli/models.go

@@ -9,8 +9,8 @@ import (
 	"github.com/mudler/LocalAI/core/config"
 
 	"github.com/mudler/LocalAI/core/gallery"
-	"github.com/mudler/LocalAI/core/startup"
 	"github.com/mudler/LocalAI/pkg/downloader"
+	"github.com/mudler/LocalAI/pkg/startup"
 	"github.com/rs/zerolog/log"
 	"github.com/schollz/progressbar/v3"
 )

core/cli/run.go

@@ -25,6 +25,7 @@ type RunCMD struct {
 	ModelsPath                   string        `env:"LOCALAI_MODELS_PATH,MODELS_PATH" type:"path" default:"${basepath}/models" help:"Path containing models used for inferencing" group:"storage"`
 	GeneratedContentPath         string        `env:"LOCALAI_GENERATED_CONTENT_PATH,GENERATED_CONTENT_PATH" type:"path" default:"/tmp/generated/content" help:"Location for generated content (e.g. images, audio, videos)" group:"storage"`
 	UploadPath                   string        `env:"LOCALAI_UPLOAD_PATH,UPLOAD_PATH" type:"path" default:"/tmp/localai/upload" help:"Path to store uploads from files api" group:"storage"`
+	ConfigPath                   string        `env:"LOCALAI_CONFIG_PATH,CONFIG_PATH" default:"/tmp/localai/config" group:"storage"`
 	LocalaiConfigDir             string        `env:"LOCALAI_CONFIG_DIR" type:"path" default:"${basepath}/configuration" help:"Directory for dynamic loading of certain configuration files (currently api_keys.json and external_backends.json)" group:"storage"`
 	LocalaiConfigDirPollInterval time.Duration `env:"LOCALAI_CONFIG_DIR_POLL_INTERVAL" help:"Typically the config path picks up changes automatically, but if your system has broken fsnotify events, set this to an interval to poll the LocalAI Config Dir (example: 1m)" group:"storage"`
 	// The alias on this option is there to preserve functionality with the old `--config-file` parameter
@@ -87,6 +88,7 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 		config.WithDebug(zerolog.GlobalLevel() <= zerolog.DebugLevel),
 		config.WithGeneratedContentDir(r.GeneratedContentPath),
 		config.WithUploadDir(r.UploadPath),
+		config.WithConfigsDir(r.ConfigPath),
 		config.WithDynamicConfigDir(r.LocalaiConfigDir),
 		config.WithDynamicConfigDirPollInterval(r.LocalaiConfigDirPollInterval),
 		config.WithF16(r.F16),

core/cli/util.go

@@ -72,7 +72,7 @@ func (u *CreateOCIImageCMD) Run(ctx *cliContext.Context) error {
 }
 
 func (u *GGUFInfoCMD) Run(ctx *cliContext.Context) error {
-	if len(u.Args) == 0 {
+	if u.Args == nil || len(u.Args) == 0 {
 		return fmt.Errorf("no GGUF file provided")
 	}
 	// We try to guess only if we don't have a template defined already

core/config/application_config.go

@@ -21,7 +21,8 @@ type ApplicationConfig struct {
 	Debug                               bool
 	GeneratedContentDir                 string
 
-	UploadDir string
+	ConfigsDir string
+	UploadDir  string
 
 	DynamicConfigsDir             string
 	DynamicConfigsDirPollInterval time.Duration
@@ -301,6 +302,12 @@ func WithUploadDir(uploadDir string) AppOption {
 	}
 }
 
+func WithConfigsDir(configsDir string) AppOption {
+	return func(o *ApplicationConfig) {
+		o.ConfigsDir = configsDir
+	}
+}
+
 func WithDynamicConfigDir(dynamicConfigsDir string) AppOption {
 	return func(o *ApplicationConfig) {
 		o.DynamicConfigsDir = dynamicConfigsDir

core/config/backend_config.go

@@ -458,7 +458,6 @@ const (
 	FLAG_TOKENIZE         BackendConfigUsecases = 0b001000000000
 	FLAG_VAD              BackendConfigUsecases = 0b010000000000
 	FLAG_VIDEO            BackendConfigUsecases = 0b100000000000
-	FLAG_DETECTION        BackendConfigUsecases = 0b1000000000000
 
 	// Common Subsets
 	FLAG_LLM BackendConfigUsecases = FLAG_CHAT | FLAG_COMPLETION | FLAG_EDIT
@@ -480,7 +479,6 @@ func GetAllBackendConfigUsecases() map[string]BackendConfigUsecases {
 		"FLAG_VAD":              FLAG_VAD,
 		"FLAG_LLM":              FLAG_LLM,
 		"FLAG_VIDEO":            FLAG_VIDEO,
-		"FLAG_DETECTION":        FLAG_DETECTION,
 	}
 }
 
@@ -574,12 +572,6 @@ func (c *BackendConfig) GuessUsecases(u BackendConfigUsecases) bool {
 		}
 	}
 
-	if (u & FLAG_DETECTION) == FLAG_DETECTION {
-		if c.Backend != "rfdetr" {
-			return false
-		}
-	}
-
 	if (u & FLAG_SOUND_GENERATION) == FLAG_SOUND_GENERATION {
 		if c.Backend != "transformers-musicgen" {
 			return false

core/gallery/backend_types.go

@@ -2,8 +2,7 @@ package gallery
 
 import (
 	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/pkg/system"
+	"github.com/mudler/LocalAI/core/system"
-	"github.com/rs/zerolog/log"
 )
 
 // BackendMetadata represents the metadata stored in a JSON file for each installed backend
@@ -24,7 +23,6 @@ type GalleryBackend struct {
 	Metadata        `json:",inline" yaml:",inline"`
 	Alias           string            `json:"alias,omitempty" yaml:"alias,omitempty"`
 	URI             string            `json:"uri,omitempty" yaml:"uri,omitempty"`
-	Mirrors         []string          `json:"mirrors,omitempty" yaml:"mirrors,omitempty"`
 	CapabilitiesMap map[string]string `json:"capabilities,omitempty" yaml:"capabilities,omitempty"`
 }
 
@@ -35,11 +33,9 @@ func (backend *GalleryBackend) FindBestBackendFromMeta(systemState *system.Syste
 
 	realBackend := backend.CapabilitiesMap[systemState.Capability(backend.CapabilitiesMap)]
 	if realBackend == "" {
-		log.Debug().Str("backend", backend.Name).Str("reportedCapability", systemState.Capability(backend.CapabilitiesMap)).Msg("No backend found for reported capability")
 		return nil
 	}
 
-	log.Debug().Str("backend", backend.Name).Str("reportedCapability", systemState.Capability(backend.CapabilitiesMap)).Msg("Found backend for reported capability")
 	return backends.FindByName(realBackend)
 }
 

core/gallery/backends.go

@@ -8,9 +8,9 @@ import (
 	"time"
 
 	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/system"
 	"github.com/mudler/LocalAI/pkg/downloader"
 	"github.com/mudler/LocalAI/pkg/model"
-	"github.com/mudler/LocalAI/pkg/system"
 	"github.com/rs/zerolog/log"
 )
 
@@ -146,18 +146,7 @@ func InstallBackend(basePath string, config *GalleryBackend, downloadStatus func
 
 	uri := downloader.URI(config.URI)
 	if err := uri.DownloadFile(backendPath, "", 1, 1, downloadStatus); err != nil {
-		success := false
+		return fmt.Errorf("failed to download backend %q: %v", config.URI, err)
-		// Try to download from mirrors
-		for _, mirror := range config.Mirrors {
-			if err := downloader.URI(mirror).DownloadFile(backendPath, "", 1, 1, downloadStatus); err == nil {
-				success = true
-				break
-			}
-		}
-
-		if !success {
-			return fmt.Errorf("failed to download backend %q: %v", config.URI, err)
-		}
 	}
 
 	// Create metadata for the backend

core/gallery/backends_test.go

@@ -7,7 +7,7 @@ import (
 	"runtime"
 
 	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/pkg/system"
+	"github.com/mudler/LocalAI/core/system"
 	. "github.com/onsi/ginkgo/v2"
 	. "github.com/onsi/gomega"
 	"gopkg.in/yaml.v2"

core/gallery/gallery.go

@@ -95,7 +95,7 @@ func FindGalleryElement[T GalleryElement](models []T, name string, basePath stri
 
 	if !strings.Contains(name, "@") {
 		for _, m := range models {
-			if strings.EqualFold(strings.ToLower(m.GetName()), strings.ToLower(name)) {
+			if strings.EqualFold(m.GetName(), name) {
 				model = m
 				break
 			}
@@ -103,7 +103,7 @@ func FindGalleryElement[T GalleryElement](models []T, name string, basePath stri
 
 	} else {
 		for _, m := range models {
-			if strings.EqualFold(strings.ToLower(name), strings.ToLower(fmt.Sprintf("%s@%s", m.GetGallery().Name, m.GetName()))) {
+			if strings.EqualFold(name, fmt.Sprintf("%s@%s", m.GetGallery().Name, m.GetName())) {
 				model = m
 				break
 			}

core/gallery/models.go

@@ -10,8 +10,8 @@ import (
 	"dario.cat/mergo"
 	"github.com/mudler/LocalAI/core/config"
 	lconfig "github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/system"
 	"github.com/mudler/LocalAI/pkg/downloader"
-	"github.com/mudler/LocalAI/pkg/system"
 	"github.com/mudler/LocalAI/pkg/utils"
 
 	"github.com/rs/zerolog/log"

core/http/app.go

@@ -10,8 +10,10 @@ import (
 
 	"github.com/dave-gray101/v2keyauth"
 	"github.com/gofiber/websocket/v2"
+	"github.com/mudler/LocalAI/pkg/utils"
 
 	"github.com/mudler/LocalAI/core/http/endpoints/localai"
+	"github.com/mudler/LocalAI/core/http/endpoints/openai"
 	"github.com/mudler/LocalAI/core/http/middleware"
 	"github.com/mudler/LocalAI/core/http/routes"
 
@@ -197,6 +199,11 @@ func API(application *application.Application) (*fiber.App, error) {
 		router.Use(csrf.New())
 	}
 
+	// Load config jsons
+	utils.LoadConfig(application.ApplicationConfig().UploadDir, openai.UploadedFilesFile, &openai.UploadedFiles)
+	utils.LoadConfig(application.ApplicationConfig().ConfigsDir, openai.AssistantsConfigFile, &openai.Assistants)
+	utils.LoadConfig(application.ApplicationConfig().ConfigsDir, openai.AssistantsFileConfigFile, &openai.AssistantFiles)
+
 	galleryService := services.NewGalleryService(application.ApplicationConfig(), application.ModelLoader())
 	err = galleryService.Start(application.ApplicationConfig().Context, application.BackendLoader())
 	if err != nil {

core/http/endpoints/localai/backend.go

@@ -34,7 +34,7 @@ func CreateBackendEndpointService(galleries []config.Gallery, backendPath string
 
 // GetOpStatusEndpoint returns the job status
 // @Summary Returns the job status
-// @Success 200 {object} services.GalleryOpStatus "Response"
+// @Success 200 {object} services.BackendOpStatus "Response"
 // @Router /backends/jobs/{uuid} [get]
 func (mgs *BackendEndpointService) GetOpStatusEndpoint() func(c *fiber.Ctx) error {
 	return func(c *fiber.Ctx) error {
@@ -48,7 +48,7 @@ func (mgs *BackendEndpointService) GetOpStatusEndpoint() func(c *fiber.Ctx) erro
 
 // GetAllStatusEndpoint returns all the jobs status progress
 // @Summary Returns all the jobs status progress
-// @Success 200 {object} map[string]services.GalleryOpStatus "Response"
+// @Success 200 {object} map[string]services.BackendOpStatus "Response"
 // @Router /backends/jobs [get]
 func (mgs *BackendEndpointService) GetAllStatusEndpoint() func(c *fiber.Ctx) error {
 	return func(c *fiber.Ctx) error {
@@ -58,7 +58,7 @@ func (mgs *BackendEndpointService) GetAllStatusEndpoint() func(c *fiber.Ctx) err
 
 // ApplyBackendEndpoint installs a new backend to a LocalAI instance
 // @Summary Install backends to LocalAI.
-// @Param request body GalleryBackend true "query params"
+// @Param request body BackendModel true "query params"
 // @Success 200 {object} schema.BackendResponse "Response"
 // @Router /backends/apply [post]
 func (mgs *BackendEndpointService) ApplyBackendEndpoint() func(c *fiber.Ctx) error {

core/http/endpoints/localai/detection.go

@@ -1,59 +0,0 @@
-package localai
-
-import (
-	"github.com/gofiber/fiber/v2"
-	"github.com/mudler/LocalAI/core/backend"
-	"github.com/mudler/LocalAI/core/config"
-	"github.com/mudler/LocalAI/core/http/middleware"
-	"github.com/mudler/LocalAI/core/schema"
-	"github.com/mudler/LocalAI/pkg/model"
-	"github.com/mudler/LocalAI/pkg/utils"
-	"github.com/rs/zerolog/log"
-)
-
-// DetectionEndpoint is the LocalAI Detection endpoint https://localai.io/docs/api-reference/detection
-// @Summary Detects objects in the input image.
-// @Param request body schema.DetectionRequest true "query params"
-// @Success 200 {object} schema.DetectionResponse "Response"
-// @Router /v1/detection [post]
-func DetectionEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
-	return func(c *fiber.Ctx) error {
-
-		input, ok := c.Locals(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST).(*schema.DetectionRequest)
-		if !ok || input.Model == "" {
-			return fiber.ErrBadRequest
-		}
-
-		cfg, ok := c.Locals(middleware.CONTEXT_LOCALS_KEY_MODEL_CONFIG).(*config.BackendConfig)
-		if !ok || cfg == nil {
-			return fiber.ErrBadRequest
-		}
-
-		log.Debug().Str("image", input.Image).Str("modelFile", "modelFile").Str("backend", cfg.Backend).Msg("Detection")
-
-		image, err := utils.GetContentURIAsBase64(input.Image)
-		if err != nil {
-			return err
-		}
-
-		res, err := backend.Detection(image, ml, appConfig, *cfg)
-		if err != nil {
-			return err
-		}
-
-		response := schema.DetectionResponse{
-			Detections: make([]schema.Detection, len(res.Detections)),
-		}
-		for i, detection := range res.Detections {
-			response.Detections[i] = schema.Detection{
-				X:         detection.X,
-				Y:         detection.Y,
-				Width:     detection.Width,
-				Height:    detection.Height,
-				ClassName: detection.ClassName,
-			}
-		}
-
-		return c.JSON(response)
-	}
-}

core/http/endpoints/localai/gallery.go

@@ -15,10 +15,9 @@ import (
 )
 
 type ModelGalleryEndpointService struct {
-	galleries        []config.Gallery
+	galleries      []config.Gallery
-	backendGalleries []config.Gallery
+	modelPath      string
-	modelPath        string
+	galleryApplier *services.GalleryService
-	galleryApplier   *services.GalleryService
 }
 
 type GalleryModel struct {
@@ -26,12 +25,11 @@ type GalleryModel struct {
 	gallery.GalleryModel
 }
 
-func CreateModelGalleryEndpointService(galleries []config.Gallery, backendGalleries []config.Gallery, modelPath string, galleryApplier *services.GalleryService) ModelGalleryEndpointService {
+func CreateModelGalleryEndpointService(galleries []config.Gallery, modelPath string, galleryApplier *services.GalleryService) ModelGalleryEndpointService {
 	return ModelGalleryEndpointService{
-		galleries:        galleries,
+		galleries:      galleries,
-		backendGalleries: backendGalleries,
+		modelPath:      modelPath,
-		modelPath:        modelPath,
+		galleryApplier: galleryApplier,
-		galleryApplier:   galleryApplier,
 	}
 }
 
@@ -81,7 +79,6 @@ func (mgs *ModelGalleryEndpointService) ApplyModelGalleryEndpoint() func(c *fibe
 			ID:                 uuid.String(),
 			GalleryElementName: input.ID,
 			Galleries:          mgs.galleries,
-			BackendGalleries:   mgs.backendGalleries,
 		}
 
 		return c.JSON(schema.GalleryResponse{ID: uuid.String(), StatusURL: fmt.Sprintf("%smodels/jobs/%s", utils.BaseURL(c), uuid.String())})

core/http/endpoints/openai/assistant.go

@@ -0,0 +1,522 @@
+package openai
+
+import (
+	"fmt"
+	"net/http"
+	"sort"
+	"strconv"
+	"strings"
+	"sync/atomic"
+	"time"
+
+	"github.com/gofiber/fiber/v2"
+	"github.com/microcosm-cc/bluemonday"
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/schema"
+	"github.com/mudler/LocalAI/core/services"
+	model "github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/utils"
+	"github.com/rs/zerolog/log"
+)
+
+// ToolType defines a type for tool options
+type ToolType string
+
+const (
+	CodeInterpreter ToolType = "code_interpreter"
+	Retrieval       ToolType = "retrieval"
+	Function        ToolType = "function"
+
+	MaxCharacterInstructions  = 32768
+	MaxCharacterDescription   = 512
+	MaxCharacterName          = 256
+	MaxToolsSize              = 128
+	MaxFileIdSize             = 20
+	MaxCharacterMetadataKey   = 64
+	MaxCharacterMetadataValue = 512
+)
+
+type Tool struct {
+	Type ToolType `json:"type"`
+}
+
+// Assistant represents the structure of an assistant object from the OpenAI API.
+type Assistant struct {
+	ID           string            `json:"id"`                     // The unique identifier of the assistant.
+	Object       string            `json:"object"`                 // Object type, which is "assistant".
+	Created      int64             `json:"created"`                // The time at which the assistant was created.
+	Model        string            `json:"model"`                  // The model ID used by the assistant.
+	Name         string            `json:"name,omitempty"`         // The name of the assistant.
+	Description  string            `json:"description,omitempty"`  // The description of the assistant.
+	Instructions string            `json:"instructions,omitempty"` // The system instructions that the assistant uses.
+	Tools        []Tool            `json:"tools,omitempty"`        // A list of tools enabled on the assistant.
+	FileIDs      []string          `json:"file_ids,omitempty"`     // A list of file IDs attached to this assistant.
+	Metadata     map[string]string `json:"metadata,omitempty"`     // Set of key-value pairs attached to the assistant.
+}
+
+var (
+	Assistants           = []Assistant{} // better to return empty array instead of "null"
+	AssistantsConfigFile = "assistants.json"
+)
+
+type AssistantRequest struct {
+	Model        string            `json:"model"`
+	Name         string            `json:"name,omitempty"`
+	Description  string            `json:"description,omitempty"`
+	Instructions string            `json:"instructions,omitempty"`
+	Tools        []Tool            `json:"tools,omitempty"`
+	FileIDs      []string          `json:"file_ids,omitempty"`
+	Metadata     map[string]string `json:"metadata,omitempty"`
+}
+
+// CreateAssistantEndpoint is the OpenAI Assistant API endpoint https://platform.openai.com/docs/api-reference/assistants/createAssistant
+// @Summary Create an assistant with a model and instructions.
+// @Param request body AssistantRequest true "query params"
+// @Success 200 {object} Assistant "Response"
+// @Router /v1/assistants [post]
+func CreateAssistantEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		request := new(AssistantRequest)
+		if err := c.BodyParser(request); err != nil {
+			log.Warn().AnErr("Unable to parse AssistantRequest", err)
+			return c.Status(fiber.StatusBadRequest).JSON(fiber.Map{"error": "Cannot parse JSON"})
+		}
+
+		if !modelExists(cl, ml, request.Model) {
+			log.Warn().Msgf("Model: %s was not found in list of models.", request.Model)
+			return c.Status(fiber.StatusBadRequest).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Model %q not found", request.Model)))
+		}
+
+		if request.Tools == nil {
+			request.Tools = []Tool{}
+		}
+
+		if request.FileIDs == nil {
+			request.FileIDs = []string{}
+		}
+
+		if request.Metadata == nil {
+			request.Metadata = make(map[string]string)
+		}
+
+		id := "asst_" + strconv.FormatInt(generateRandomID(), 10)
+
+		assistant := Assistant{
+			ID:           id,
+			Object:       "assistant",
+			Created:      time.Now().Unix(),
+			Model:        request.Model,
+			Name:         request.Name,
+			Description:  request.Description,
+			Instructions: request.Instructions,
+			Tools:        request.Tools,
+			FileIDs:      request.FileIDs,
+			Metadata:     request.Metadata,
+		}
+
+		Assistants = append(Assistants, assistant)
+		utils.SaveConfig(appConfig.ConfigsDir, AssistantsConfigFile, Assistants)
+		return c.Status(fiber.StatusOK).JSON(assistant)
+	}
+}
+
+var currentId int64 = 0
+
+func generateRandomID() int64 {
+	atomic.AddInt64(&currentId, 1)
+	return currentId
+}
+
+// ListAssistantsEndpoint is the OpenAI Assistant API endpoint to list assistents https://platform.openai.com/docs/api-reference/assistants/listAssistants
+// @Summary List available assistents
+// @Param limit query int false "Limit the number of assistants returned"
+// @Param order query string false "Order of assistants returned"
+// @Param after query string false "Return assistants created after the given ID"
+// @Param before query string false "Return assistants created before the given ID"
+// @Success 200 {object} []Assistant "Response"
+// @Router /v1/assistants [get]
+func ListAssistantsEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		// Because we're altering the existing assistants list we should just duplicate it for now.
+		returnAssistants := Assistants
+		// Parse query parameters
+		limitQuery := c.Query("limit", "20")
+		orderQuery := c.Query("order", "desc")
+		afterQuery := c.Query("after")
+		beforeQuery := c.Query("before")
+
+		// Convert string limit to integer
+		limit, err := strconv.Atoi(limitQuery)
+		if err != nil {
+			return c.Status(http.StatusBadRequest).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Invalid limit query value: %s", limitQuery)))
+		}
+
+		// Sort assistants
+		sort.SliceStable(returnAssistants, func(i, j int) bool {
+			if orderQuery == "asc" {
+				return returnAssistants[i].Created < returnAssistants[j].Created
+			}
+			return returnAssistants[i].Created > returnAssistants[j].Created
+		})
+
+		// After and before cursors
+		if afterQuery != "" {
+			returnAssistants = filterAssistantsAfterID(returnAssistants, afterQuery)
+		}
+		if beforeQuery != "" {
+			returnAssistants = filterAssistantsBeforeID(returnAssistants, beforeQuery)
+		}
+
+		// Apply limit
+		if limit < len(returnAssistants) {
+			returnAssistants = returnAssistants[:limit]
+		}
+
+		return c.JSON(returnAssistants)
+	}
+}
+
+// FilterAssistantsBeforeID filters out those assistants whose ID comes before the given ID
+// We assume that the assistants are already sorted
+func filterAssistantsBeforeID(assistants []Assistant, id string) []Assistant {
+	idInt, err := strconv.Atoi(id)
+	if err != nil {
+		return assistants // Return original slice if invalid id format is provided
+	}
+
+	var filteredAssistants []Assistant
+
+	for _, assistant := range assistants {
+		aid, err := strconv.Atoi(strings.TrimPrefix(assistant.ID, "asst_"))
+		if err != nil {
+			continue // Skip if invalid id in assistant
+		}
+
+		if aid < idInt {
+			filteredAssistants = append(filteredAssistants, assistant)
+		}
+	}
+
+	return filteredAssistants
+}
+
+// FilterAssistantsAfterID filters out those assistants whose ID comes after the given ID
+// We assume that the assistants are already sorted
+func filterAssistantsAfterID(assistants []Assistant, id string) []Assistant {
+	idInt, err := strconv.Atoi(id)
+	if err != nil {
+		return assistants // Return original slice if invalid id format is provided
+	}
+
+	var filteredAssistants []Assistant
+
+	for _, assistant := range assistants {
+		aid, err := strconv.Atoi(strings.TrimPrefix(assistant.ID, "asst_"))
+		if err != nil {
+			continue // Skip if invalid id in assistant
+		}
+
+		if aid > idInt {
+			filteredAssistants = append(filteredAssistants, assistant)
+		}
+	}
+
+	return filteredAssistants
+}
+
+func modelExists(cl *config.BackendConfigLoader, ml *model.ModelLoader, modelName string) (found bool) {
+	found = false
+	models, err := services.ListModels(cl, ml, config.NoFilterFn, services.SKIP_IF_CONFIGURED)
+	if err != nil {
+		return
+	}
+
+	for _, model := range models {
+		if model == modelName {
+			found = true
+			return
+		}
+	}
+	return
+}
+
+// DeleteAssistantEndpoint is the OpenAI Assistant API endpoint to delete assistents https://platform.openai.com/docs/api-reference/assistants/deleteAssistant
+// @Summary Delete assistents
+// @Success 200 {object} schema.DeleteAssistantResponse "Response"
+// @Router /v1/assistants/{assistant_id} [delete]
+func DeleteAssistantEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		assistantID := c.Params("assistant_id")
+		if assistantID == "" {
+			return c.Status(fiber.StatusBadRequest).SendString("parameter assistant_id is required")
+		}
+
+		for i, assistant := range Assistants {
+			if assistant.ID == assistantID {
+				Assistants = append(Assistants[:i], Assistants[i+1:]...)
+				utils.SaveConfig(appConfig.ConfigsDir, AssistantsConfigFile, Assistants)
+				return c.Status(fiber.StatusOK).JSON(schema.DeleteAssistantResponse{
+					ID:      assistantID,
+					Object:  "assistant.deleted",
+					Deleted: true,
+				})
+			}
+		}
+
+		log.Warn().Msgf("Unable to find assistant %s for deletion", assistantID)
+		return c.Status(fiber.StatusNotFound).JSON(schema.DeleteAssistantResponse{
+			ID:      assistantID,
+			Object:  "assistant.deleted",
+			Deleted: false,
+		})
+	}
+}
+
+// GetAssistantEndpoint is the OpenAI Assistant API endpoint to get assistents https://platform.openai.com/docs/api-reference/assistants/getAssistant
+// @Summary Get assistent data
+// @Success 200 {object} Assistant "Response"
+// @Router /v1/assistants/{assistant_id} [get]
+func GetAssistantEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		assistantID := c.Params("assistant_id")
+		if assistantID == "" {
+			return c.Status(fiber.StatusBadRequest).SendString("parameter assistant_id is required")
+		}
+
+		for _, assistant := range Assistants {
+			if assistant.ID == assistantID {
+				return c.Status(fiber.StatusOK).JSON(assistant)
+			}
+		}
+
+		return c.Status(fiber.StatusNotFound).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Unable to find assistant with id: %s", assistantID)))
+	}
+}
+
+type AssistantFile struct {
+	ID          string `json:"id"`
+	Object      string `json:"object"`
+	CreatedAt   int64  `json:"created_at"`
+	AssistantID string `json:"assistant_id"`
+}
+
+var (
+	AssistantFiles           []AssistantFile
+	AssistantsFileConfigFile = "assistantsFile.json"
+)
+
+func CreateAssistantFileEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		request := new(schema.AssistantFileRequest)
+		if err := c.BodyParser(request); err != nil {
+			return c.Status(fiber.StatusBadRequest).JSON(fiber.Map{"error": "Cannot parse JSON"})
+		}
+
+		assistantID := c.Params("assistant_id")
+		if assistantID == "" {
+			return c.Status(fiber.StatusBadRequest).SendString("parameter assistant_id is required")
+		}
+
+		for _, assistant := range Assistants {
+			if assistant.ID == assistantID {
+				if len(assistant.FileIDs) > MaxFileIdSize {
+					return c.Status(fiber.StatusBadRequest).SendString(fmt.Sprintf("Max files %d for assistant %s reached.", MaxFileIdSize, assistant.Name))
+				}
+
+				for _, file := range UploadedFiles {
+					if file.ID == request.FileID {
+						assistant.FileIDs = append(assistant.FileIDs, request.FileID)
+						assistantFile := AssistantFile{
+							ID:          file.ID,
+							Object:      "assistant.file",
+							CreatedAt:   time.Now().Unix(),
+							AssistantID: assistant.ID,
+						}
+						AssistantFiles = append(AssistantFiles, assistantFile)
+						utils.SaveConfig(appConfig.ConfigsDir, AssistantsFileConfigFile, AssistantFiles)
+						return c.Status(fiber.StatusOK).JSON(assistantFile)
+					}
+				}
+
+				return c.Status(fiber.StatusNotFound).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Unable to find file_id: %s", request.FileID)))
+			}
+		}
+
+		return c.Status(fiber.StatusNotFound).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Unable to find %q", assistantID)))
+	}
+}
+
+func ListAssistantFilesEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	type ListAssistantFiles struct {
+		Data   []schema.File
+		Object string
+	}
+
+	return func(c *fiber.Ctx) error {
+		assistantID := c.Params("assistant_id")
+		if assistantID == "" {
+			return c.Status(fiber.StatusBadRequest).SendString("parameter assistant_id is required")
+		}
+
+		limitQuery := c.Query("limit", "20")
+		order := c.Query("order", "desc")
+		limit, err := strconv.Atoi(limitQuery)
+		if err != nil || limit < 1 || limit > 100 {
+			limit = 20 // Default to 20 if there's an error or the limit is out of bounds
+		}
+
+		// Sort files by CreatedAt depending on the order query parameter
+		if order == "asc" {
+			sort.Slice(AssistantFiles, func(i, j int) bool {
+				return AssistantFiles[i].CreatedAt < AssistantFiles[j].CreatedAt
+			})
+		} else { // default to "desc"
+			sort.Slice(AssistantFiles, func(i, j int) bool {
+				return AssistantFiles[i].CreatedAt > AssistantFiles[j].CreatedAt
+			})
+		}
+
+		// Limit the number of files returned
+		var limitedFiles []AssistantFile
+		hasMore := false
+		if len(AssistantFiles) > limit {
+			hasMore = true
+			limitedFiles = AssistantFiles[:limit]
+		} else {
+			limitedFiles = AssistantFiles
+		}
+
+		response := map[string]interface{}{
+			"object": "list",
+			"data":   limitedFiles,
+			"first_id": func() string {
+				if len(limitedFiles) > 0 {
+					return limitedFiles[0].ID
+				}
+				return ""
+			}(),
+			"last_id": func() string {
+				if len(limitedFiles) > 0 {
+					return limitedFiles[len(limitedFiles)-1].ID
+				}
+				return ""
+			}(),
+			"has_more": hasMore,
+		}
+
+		return c.Status(fiber.StatusOK).JSON(response)
+	}
+}
+
+func ModifyAssistantEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		request := new(AssistantRequest)
+		if err := c.BodyParser(request); err != nil {
+			log.Warn().AnErr("Unable to parse AssistantRequest", err)
+			return c.Status(fiber.StatusBadRequest).JSON(fiber.Map{"error": "Cannot parse JSON"})
+		}
+
+		assistantID := c.Params("assistant_id")
+		if assistantID == "" {
+			return c.Status(fiber.StatusBadRequest).SendString("parameter assistant_id is required")
+		}
+
+		for i, assistant := range Assistants {
+			if assistant.ID == assistantID {
+				newAssistant := Assistant{
+					ID:           assistantID,
+					Object:       assistant.Object,
+					Created:      assistant.Created,
+					Model:        request.Model,
+					Name:         request.Name,
+					Description:  request.Description,
+					Instructions: request.Instructions,
+					Tools:        request.Tools,
+					FileIDs:      request.FileIDs, // todo: should probably verify fileids exist
+					Metadata:     request.Metadata,
+				}
+
+				// Remove old one and replace with new one
+				Assistants = append(Assistants[:i], Assistants[i+1:]...)
+				Assistants = append(Assistants, newAssistant)
+				utils.SaveConfig(appConfig.ConfigsDir, AssistantsConfigFile, Assistants)
+				return c.Status(fiber.StatusOK).JSON(newAssistant)
+			}
+		}
+		return c.Status(fiber.StatusNotFound).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Unable to find assistant with id: %s", assistantID)))
+	}
+}
+
+func DeleteAssistantFileEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		assistantID := c.Params("assistant_id")
+		fileId := c.Params("file_id")
+		if assistantID == "" {
+			return c.Status(fiber.StatusBadRequest).SendString("parameter assistant_id and file_id are required")
+		}
+		// First remove file from assistant
+		for i, assistant := range Assistants {
+			if assistant.ID == assistantID {
+				for j, fileId := range assistant.FileIDs {
+					Assistants[i].FileIDs = append(Assistants[i].FileIDs[:j], Assistants[i].FileIDs[j+1:]...)
+
+					// Check if the file exists in the assistantFiles slice
+					for i, assistantFile := range AssistantFiles {
+						if assistantFile.ID == fileId {
+							// Remove the file from the assistantFiles slice
+							AssistantFiles = append(AssistantFiles[:i], AssistantFiles[i+1:]...)
+							utils.SaveConfig(appConfig.ConfigsDir, AssistantsFileConfigFile, AssistantFiles)
+							return c.Status(fiber.StatusOK).JSON(schema.DeleteAssistantFileResponse{
+								ID:      fileId,
+								Object:  "assistant.file.deleted",
+								Deleted: true,
+							})
+						}
+					}
+				}
+
+				log.Warn().Msgf("Unable to locate file_id: %s in assistants: %s. Continuing to delete assistant file.", fileId, assistantID)
+				for i, assistantFile := range AssistantFiles {
+					if assistantFile.AssistantID == assistantID {
+
+						AssistantFiles = append(AssistantFiles[:i], AssistantFiles[i+1:]...)
+						utils.SaveConfig(appConfig.ConfigsDir, AssistantsFileConfigFile, AssistantFiles)
+
+						return c.Status(fiber.StatusNotFound).JSON(schema.DeleteAssistantFileResponse{
+							ID:      fileId,
+							Object:  "assistant.file.deleted",
+							Deleted: true,
+						})
+					}
+				}
+			}
+		}
+		log.Warn().Msgf("Unable to find assistant: %s", assistantID)
+
+		return c.Status(fiber.StatusNotFound).JSON(schema.DeleteAssistantFileResponse{
+			ID:      fileId,
+			Object:  "assistant.file.deleted",
+			Deleted: false,
+		})
+	}
+}
+
+func GetAssistantFileEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		assistantID := c.Params("assistant_id")
+		fileId := c.Params("file_id")
+		if assistantID == "" {
+			return c.Status(fiber.StatusBadRequest).SendString("parameter assistant_id and file_id are required")
+		}
+
+		for _, assistantFile := range AssistantFiles {
+			if assistantFile.AssistantID == assistantID {
+				if assistantFile.ID == fileId {
+					return c.Status(fiber.StatusOK).JSON(assistantFile)
+				}
+				return c.Status(fiber.StatusNotFound).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Unable to find assistant file with file_id: %s", fileId)))
+			}
+		}
+		return c.Status(fiber.StatusNotFound).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Unable to find assistant file with assistant_id: %s", assistantID)))
+	}
+}

core/http/endpoints/openai/assistant_test.go

@@ -0,0 +1,460 @@
+package openai
+
+import (
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/gofiber/fiber/v2"
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/schema"
+	"github.com/mudler/LocalAI/pkg/model"
+	"github.com/stretchr/testify/assert"
+)
+
+var configsDir string = "/tmp/localai/configs"
+
+type MockLoader struct {
+	models []string
+}
+
+func tearDown() func() {
+	return func() {
+		UploadedFiles = []schema.File{}
+		Assistants = []Assistant{}
+		AssistantFiles = []AssistantFile{}
+		_ = os.Remove(filepath.Join(configsDir, AssistantsConfigFile))
+		_ = os.Remove(filepath.Join(configsDir, AssistantsFileConfigFile))
+	}
+}
+
+func TestAssistantEndpoints(t *testing.T) {
+	// Preparing the mocked objects
+	cl := &config.BackendConfigLoader{}
+	//configsDir := "/tmp/localai/configs"
+	modelPath := "/tmp/localai/model"
+	var ml = model.NewModelLoader(modelPath, false)
+
+	appConfig := &config.ApplicationConfig{
+		ConfigsDir:    configsDir,
+		UploadLimitMB: 10,
+		UploadDir:     "test_dir",
+		ModelPath:     modelPath,
+	}
+
+	_ = os.RemoveAll(appConfig.ConfigsDir)
+	_ = os.MkdirAll(appConfig.ConfigsDir, 0750)
+	_ = os.MkdirAll(modelPath, 0750)
+	os.Create(filepath.Join(modelPath, "ggml-gpt4all-j"))
+
+	app := fiber.New(fiber.Config{
+		BodyLimit: 20 * 1024 * 1024, // sets the limit to 20MB.
+	})
+
+	// Create a Test Server
+	app.Get("/assistants", ListAssistantsEndpoint(cl, ml, appConfig))
+	app.Post("/assistants", CreateAssistantEndpoint(cl, ml, appConfig))
+	app.Delete("/assistants/:assistant_id", DeleteAssistantEndpoint(cl, ml, appConfig))
+	app.Get("/assistants/:assistant_id", GetAssistantEndpoint(cl, ml, appConfig))
+	app.Post("/assistants/:assistant_id", ModifyAssistantEndpoint(cl, ml, appConfig))
+
+	app.Post("/files", UploadFilesEndpoint(cl, appConfig))
+	app.Get("/assistants/:assistant_id/files", ListAssistantFilesEndpoint(cl, ml, appConfig))
+	app.Post("/assistants/:assistant_id/files", CreateAssistantFileEndpoint(cl, ml, appConfig))
+	app.Delete("/assistants/:assistant_id/files/:file_id", DeleteAssistantFileEndpoint(cl, ml, appConfig))
+	app.Get("/assistants/:assistant_id/files/:file_id", GetAssistantFileEndpoint(cl, ml, appConfig))
+
+	t.Run("CreateAssistantEndpoint", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		ar := &AssistantRequest{
+			Model:        "ggml-gpt4all-j",
+			Name:         "3.5-turbo",
+			Description:  "Test Assistant",
+			Instructions: "You are computer science teacher answering student questions",
+			Tools:        []Tool{{Type: Function}},
+			FileIDs:      nil,
+			Metadata:     nil,
+		}
+
+		resultAssistant, resp, err := createAssistant(app, *ar)
+		assert.NoError(t, err)
+		assert.Equal(t, fiber.StatusOK, resp.StatusCode)
+
+		assert.Equal(t, 1, len(Assistants))
+		//t.Cleanup(cleanupAllAssistants(t, app, []string{resultAssistant.ID}))
+
+		assert.Equal(t, ar.Name, resultAssistant.Name)
+		assert.Equal(t, ar.Model, resultAssistant.Model)
+		assert.Equal(t, ar.Tools, resultAssistant.Tools)
+		assert.Equal(t, ar.Description, resultAssistant.Description)
+		assert.Equal(t, ar.Instructions, resultAssistant.Instructions)
+		assert.Equal(t, ar.FileIDs, resultAssistant.FileIDs)
+		assert.Equal(t, ar.Metadata, resultAssistant.Metadata)
+	})
+
+	t.Run("ListAssistantsEndpoint", func(t *testing.T) {
+		var ids []string
+		var resultAssistant []Assistant
+		for i := 0; i < 4; i++ {
+			ar := &AssistantRequest{
+				Model:        "ggml-gpt4all-j",
+				Name:         fmt.Sprintf("3.5-turbo-%d", i),
+				Description:  fmt.Sprintf("Test Assistant - %d", i),
+				Instructions: fmt.Sprintf("You are computer science teacher answering student questions - %d", i),
+				Tools:        []Tool{{Type: Function}},
+				FileIDs:      []string{"fid-1234"},
+				Metadata:     map[string]string{"meta": "data"},
+			}
+
+			//var err error
+			ra, _, err := createAssistant(app, *ar)
+			// Because we create the assistants so fast all end up with the same created time.
+			time.Sleep(time.Second)
+			resultAssistant = append(resultAssistant, ra)
+			assert.NoError(t, err)
+			ids = append(ids, resultAssistant[i].ID)
+		}
+
+		t.Cleanup(cleanupAllAssistants(t, app, ids))
+
+		tests := []struct {
+			name                 string
+			reqURL               string
+			expectedStatus       int
+			expectedResult       []Assistant
+			expectedStringResult string
+		}{
+			{
+				name:           "Valid Usage - limit only",
+				reqURL:         "/assistants?limit=2",
+				expectedStatus: http.StatusOK,
+				expectedResult: Assistants[:2], // Expecting the first two assistants
+			},
+			{
+				name:           "Valid Usage - order asc",
+				reqURL:         "/assistants?order=asc",
+				expectedStatus: http.StatusOK,
+				expectedResult: Assistants, // Expecting all assistants in ascending order
+			},
+			{
+				name:           "Valid Usage - order desc",
+				reqURL:         "/assistants?order=desc",
+				expectedStatus: http.StatusOK,
+				expectedResult: []Assistant{Assistants[3], Assistants[2], Assistants[1], Assistants[0]}, // Expecting all assistants in descending order
+			},
+			{
+				name:           "Valid Usage - after specific ID",
+				reqURL:         "/assistants?after=2",
+				expectedStatus: http.StatusOK,
+				// Note this is correct because it's put in descending order already
+				expectedResult: Assistants[:3], // Expecting assistants after (excluding) ID 2
+			},
+			{
+				name:           "Valid Usage - before specific ID",
+				reqURL:         "/assistants?before=4",
+				expectedStatus: http.StatusOK,
+				expectedResult: Assistants[2:], // Expecting assistants before (excluding) ID 3.
+			},
+			{
+				name:                 "Invalid Usage - non-integer limit",
+				reqURL:               "/assistants?limit=two",
+				expectedStatus:       http.StatusBadRequest,
+				expectedStringResult: "Invalid limit query value: two",
+			},
+			{
+				name:           "Invalid Usage - non-existing id in after",
+				reqURL:         "/assistants?after=100",
+				expectedStatus: http.StatusOK,
+				expectedResult: []Assistant(nil), // Expecting empty list as there are no IDs above 100
+			},
+		}
+
+		for _, tt := range tests {
+			t.Run(tt.name, func(t *testing.T) {
+				request := httptest.NewRequest(http.MethodGet, tt.reqURL, nil)
+				response, err := app.Test(request)
+				assert.NoError(t, err)
+				assert.Equal(t, tt.expectedStatus, response.StatusCode)
+				if tt.expectedStatus != fiber.StatusOK {
+					all, _ := io.ReadAll(response.Body)
+					assert.Equal(t, tt.expectedStringResult, string(all))
+				} else {
+					var result []Assistant
+					err = json.NewDecoder(response.Body).Decode(&result)
+					assert.NoError(t, err)
+
+					assert.Equal(t, tt.expectedResult, result)
+				}
+			})
+		}
+	})
+
+	t.Run("DeleteAssistantEndpoint", func(t *testing.T) {
+		ar := &AssistantRequest{
+			Model:        "ggml-gpt4all-j",
+			Name:         "3.5-turbo",
+			Description:  "Test Assistant",
+			Instructions: "You are computer science teacher answering student questions",
+			Tools:        []Tool{{Type: Function}},
+			FileIDs:      nil,
+			Metadata:     nil,
+		}
+
+		resultAssistant, _, err := createAssistant(app, *ar)
+		assert.NoError(t, err)
+
+		target := fmt.Sprintf("/assistants/%s", resultAssistant.ID)
+		deleteReq := httptest.NewRequest(http.MethodDelete, target, nil)
+		_, err = app.Test(deleteReq)
+		assert.NoError(t, err)
+		assert.Equal(t, 0, len(Assistants))
+	})
+
+	t.Run("GetAssistantEndpoint", func(t *testing.T) {
+		ar := &AssistantRequest{
+			Model:        "ggml-gpt4all-j",
+			Name:         "3.5-turbo",
+			Description:  "Test Assistant",
+			Instructions: "You are computer science teacher answering student questions",
+			Tools:        []Tool{{Type: Function}},
+			FileIDs:      nil,
+			Metadata:     nil,
+		}
+
+		resultAssistant, _, err := createAssistant(app, *ar)
+		assert.NoError(t, err)
+		t.Cleanup(cleanupAllAssistants(t, app, []string{resultAssistant.ID}))
+
+		target := fmt.Sprintf("/assistants/%s", resultAssistant.ID)
+		request := httptest.NewRequest(http.MethodGet, target, nil)
+		response, err := app.Test(request)
+		assert.NoError(t, err)
+
+		var getAssistant Assistant
+		err = json.NewDecoder(response.Body).Decode(&getAssistant)
+		assert.NoError(t, err)
+
+		assert.Equal(t, resultAssistant.ID, getAssistant.ID)
+	})
+
+	t.Run("ModifyAssistantEndpoint", func(t *testing.T) {
+		ar := &AssistantRequest{
+			Model:        "ggml-gpt4all-j",
+			Name:         "3.5-turbo",
+			Description:  "Test Assistant",
+			Instructions: "You are computer science teacher answering student questions",
+			Tools:        []Tool{{Type: Function}},
+			FileIDs:      nil,
+			Metadata:     nil,
+		}
+
+		resultAssistant, _, err := createAssistant(app, *ar)
+		assert.NoError(t, err)
+
+		modifiedAr := &AssistantRequest{
+			Model:        "ggml-gpt4all-j",
+			Name:         "4.0-turbo",
+			Description:  "Modified Test Assistant",
+			Instructions: "You are math teacher answering student questions",
+			Tools:        []Tool{{Type: CodeInterpreter}},
+			FileIDs:      nil,
+			Metadata:     nil,
+		}
+
+		modifiedArJson, err := json.Marshal(modifiedAr)
+		assert.NoError(t, err)
+
+		target := fmt.Sprintf("/assistants/%s", resultAssistant.ID)
+		request := httptest.NewRequest(http.MethodPost, target, strings.NewReader(string(modifiedArJson)))
+		request.Header.Set(fiber.HeaderContentType, "application/json")
+
+		modifyResponse, err := app.Test(request)
+		assert.NoError(t, err)
+		var getAssistant Assistant
+		err = json.NewDecoder(modifyResponse.Body).Decode(&getAssistant)
+		assert.NoError(t, err)
+
+		t.Cleanup(cleanupAllAssistants(t, app, []string{getAssistant.ID}))
+
+		assert.Equal(t, resultAssistant.ID, getAssistant.ID) // IDs should match even if contents change
+		assert.Equal(t, modifiedAr.Tools, getAssistant.Tools)
+		assert.Equal(t, modifiedAr.Name, getAssistant.Name)
+		assert.Equal(t, modifiedAr.Instructions, getAssistant.Instructions)
+		assert.Equal(t, modifiedAr.Description, getAssistant.Description)
+	})
+
+	t.Run("CreateAssistantFileEndpoint", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		file, assistant, err := createFileAndAssistant(t, app, appConfig)
+		assert.NoError(t, err)
+
+		afr := schema.AssistantFileRequest{FileID: file.ID}
+		af, _, err := createAssistantFile(app, afr, assistant.ID)
+
+		assert.NoError(t, err)
+		assert.Equal(t, assistant.ID, af.AssistantID)
+	})
+	t.Run("ListAssistantFilesEndpoint", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		file, assistant, err := createFileAndAssistant(t, app, appConfig)
+		assert.NoError(t, err)
+
+		afr := schema.AssistantFileRequest{FileID: file.ID}
+		af, _, err := createAssistantFile(app, afr, assistant.ID)
+		assert.NoError(t, err)
+
+		assert.Equal(t, assistant.ID, af.AssistantID)
+	})
+	t.Run("GetAssistantFileEndpoint", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		file, assistant, err := createFileAndAssistant(t, app, appConfig)
+		assert.NoError(t, err)
+
+		afr := schema.AssistantFileRequest{FileID: file.ID}
+		af, _, err := createAssistantFile(app, afr, assistant.ID)
+		assert.NoError(t, err)
+		t.Cleanup(cleanupAssistantFile(t, app, af.ID, af.AssistantID))
+
+		target := fmt.Sprintf("/assistants/%s/files/%s", assistant.ID, file.ID)
+		request := httptest.NewRequest(http.MethodGet, target, nil)
+		response, err := app.Test(request)
+		assert.NoError(t, err)
+
+		var assistantFile AssistantFile
+		err = json.NewDecoder(response.Body).Decode(&assistantFile)
+		assert.NoError(t, err)
+
+		assert.Equal(t, af.ID, assistantFile.ID)
+		assert.Equal(t, af.AssistantID, assistantFile.AssistantID)
+	})
+	t.Run("DeleteAssistantFileEndpoint", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		file, assistant, err := createFileAndAssistant(t, app, appConfig)
+		assert.NoError(t, err)
+
+		afr := schema.AssistantFileRequest{FileID: file.ID}
+		af, _, err := createAssistantFile(app, afr, assistant.ID)
+		assert.NoError(t, err)
+
+		cleanupAssistantFile(t, app, af.ID, af.AssistantID)()
+
+		assert.Empty(t, AssistantFiles)
+	})
+
+}
+
+func createFileAndAssistant(t *testing.T, app *fiber.App, o *config.ApplicationConfig) (schema.File, Assistant, error) {
+	ar := &AssistantRequest{
+		Model:        "ggml-gpt4all-j",
+		Name:         "3.5-turbo",
+		Description:  "Test Assistant",
+		Instructions: "You are computer science teacher answering student questions",
+		Tools:        []Tool{{Type: Function}},
+		FileIDs:      nil,
+		Metadata:     nil,
+	}
+
+	assistant, _, err := createAssistant(app, *ar)
+	if err != nil {
+		return schema.File{}, Assistant{}, err
+	}
+	t.Cleanup(cleanupAllAssistants(t, app, []string{assistant.ID}))
+
+	file := CallFilesUploadEndpointWithCleanup(t, app, "test.txt", "file", "fine-tune", 5, o)
+	t.Cleanup(func() {
+		_, err := CallFilesDeleteEndpoint(t, app, file.ID)
+		assert.NoError(t, err)
+	})
+	return file, assistant, nil
+}
+
+func createAssistantFile(app *fiber.App, afr schema.AssistantFileRequest, assistantId string) (AssistantFile, *http.Response, error) {
+	afrJson, err := json.Marshal(afr)
+	if err != nil {
+		return AssistantFile{}, nil, err
+	}
+
+	target := fmt.Sprintf("/assistants/%s/files", assistantId)
+	request := httptest.NewRequest(http.MethodPost, target, strings.NewReader(string(afrJson)))
+	request.Header.Set(fiber.HeaderContentType, "application/json")
+	request.Header.Set("OpenAi-Beta", "assistants=v1")
+
+	resp, err := app.Test(request)
+	if err != nil {
+		return AssistantFile{}, resp, err
+	}
+
+	var assistantFile AssistantFile
+	all, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return AssistantFile{}, resp, err
+	}
+	err = json.NewDecoder(strings.NewReader(string(all))).Decode(&assistantFile)
+	if err != nil {
+		return AssistantFile{}, resp, err
+	}
+
+	return assistantFile, resp, nil
+}
+
+func createAssistant(app *fiber.App, ar AssistantRequest) (Assistant, *http.Response, error) {
+	assistant, err := json.Marshal(ar)
+	if err != nil {
+		return Assistant{}, nil, err
+	}
+
+	request := httptest.NewRequest(http.MethodPost, "/assistants", strings.NewReader(string(assistant)))
+	request.Header.Set(fiber.HeaderContentType, "application/json")
+	request.Header.Set("OpenAi-Beta", "assistants=v1")
+
+	resp, err := app.Test(request)
+	if err != nil {
+		return Assistant{}, resp, err
+	}
+
+	bodyString, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return Assistant{}, resp, err
+	}
+
+	var resultAssistant Assistant
+	err = json.NewDecoder(strings.NewReader(string(bodyString))).Decode(&resultAssistant)
+	return resultAssistant, resp, err
+}
+
+func cleanupAllAssistants(t *testing.T, app *fiber.App, ids []string) func() {
+	return func() {
+		for _, assistant := range ids {
+			target := fmt.Sprintf("/assistants/%s", assistant)
+			deleteReq := httptest.NewRequest(http.MethodDelete, target, nil)
+			_, err := app.Test(deleteReq)
+			if err != nil {
+				t.Fatalf("Failed to delete assistant %s: %v", assistant, err)
+			}
+		}
+	}
+}
+
+func cleanupAssistantFile(t *testing.T, app *fiber.App, fileId, assistantId string) func() {
+	return func() {
+		target := fmt.Sprintf("/assistants/%s/files/%s", assistantId, fileId)
+		request := httptest.NewRequest(http.MethodDelete, target, nil)
+		request.Header.Set(fiber.HeaderContentType, "application/json")
+		request.Header.Set("OpenAi-Beta", "assistants=v1")
+
+		resp, err := app.Test(request)
+		assert.NoError(t, err)
+
+		var dafr schema.DeleteAssistantFileResponse
+		err = json.NewDecoder(resp.Body).Decode(&dafr)
+		assert.NoError(t, err)
+		assert.True(t, dafr.Deleted)
+	}
+}

core/http/endpoints/openai/chat.go

@@ -15,8 +15,8 @@ import (
 	"github.com/mudler/LocalAI/core/schema"
 	"github.com/mudler/LocalAI/pkg/functions"
 
-	"github.com/mudler/LocalAI/core/templates"
 	"github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/templates"
 
 	"github.com/rs/zerolog/log"
 	"github.com/valyala/fasthttp"
@@ -175,7 +175,7 @@ func ChatEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, evaluat
 		textContentToReturn = ""
 		id = uuid.New().String()
 		created = int(time.Now().Unix())
-
+		
 		input, ok := c.Locals(middleware.CONTEXT_LOCALS_KEY_LOCALAI_REQUEST).(*schema.OpenAIRequest)
 		if !ok || input.Model == "" {
 			return fiber.ErrBadRequest

core/http/endpoints/openai/completion.go

@@ -15,9 +15,9 @@ import (
 	"github.com/gofiber/fiber/v2"
 	"github.com/google/uuid"
 	"github.com/mudler/LocalAI/core/schema"
-	"github.com/mudler/LocalAI/core/templates"
 	"github.com/mudler/LocalAI/pkg/functions"
 	"github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/templates"
 	"github.com/rs/zerolog/log"
 	"github.com/valyala/fasthttp"
 )

core/http/endpoints/openai/edit.go

@@ -12,8 +12,8 @@ import (
 	"github.com/google/uuid"
 	"github.com/mudler/LocalAI/core/schema"
 
-	"github.com/mudler/LocalAI/core/templates"
 	"github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/templates"
 
 	"github.com/rs/zerolog/log"
 )

core/http/endpoints/openai/files.go

@@ -0,0 +1,194 @@
+package openai
+
+import (
+	"errors"
+	"fmt"
+	"os"
+	"path/filepath"
+	"sync/atomic"
+	"time"
+
+	"github.com/microcosm-cc/bluemonday"
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/schema"
+
+	"github.com/gofiber/fiber/v2"
+	"github.com/mudler/LocalAI/pkg/utils"
+)
+
+var UploadedFiles []schema.File
+
+const UploadedFilesFile = "uploadedFiles.json"
+
+// UploadFilesEndpoint https://platform.openai.com/docs/api-reference/files/create
+func UploadFilesEndpoint(cm *config.BackendConfigLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		file, err := c.FormFile("file")
+		if err != nil {
+			return err
+		}
+
+		// Check the file size
+		if file.Size > int64(appConfig.UploadLimitMB*1024*1024) {
+			return c.Status(fiber.StatusBadRequest).SendString(fmt.Sprintf("File size %d exceeds upload limit %d", file.Size, appConfig.UploadLimitMB))
+		}
+
+		purpose := c.FormValue("purpose", "") //TODO put in purpose dirs
+		if purpose == "" {
+			return c.Status(fiber.StatusBadRequest).SendString("Purpose is not defined")
+		}
+
+		// Sanitize the filename to prevent directory traversal
+		filename := utils.SanitizeFileName(file.Filename)
+
+		savePath := filepath.Join(appConfig.UploadDir, filename)
+
+		// Check if file already exists
+		if _, err := os.Stat(savePath); !os.IsNotExist(err) {
+			return c.Status(fiber.StatusBadRequest).SendString("File already exists")
+		}
+
+		err = c.SaveFile(file, savePath)
+		if err != nil {
+			return c.Status(fiber.StatusInternalServerError).SendString("Failed to save file: " + bluemonday.StrictPolicy().Sanitize(err.Error()))
+		}
+
+		f := schema.File{
+			ID:        fmt.Sprintf("file-%d", getNextFileId()),
+			Object:    "file",
+			Bytes:     int(file.Size),
+			CreatedAt: time.Now(),
+			Filename:  file.Filename,
+			Purpose:   purpose,
+		}
+
+		UploadedFiles = append(UploadedFiles, f)
+		utils.SaveConfig(appConfig.UploadDir, UploadedFilesFile, UploadedFiles)
+		return c.Status(fiber.StatusOK).JSON(f)
+	}
+}
+
+var currentFileId int64 = 0
+
+func getNextFileId() int64 {
+	atomic.AddInt64(&currentId, 1)
+	return currentId
+}
+
+// ListFilesEndpoint https://platform.openai.com/docs/api-reference/files/list
+// @Summary List files.
+// @Success 200 {object} schema.ListFiles "Response"
+// @Router /v1/files [get]
+func ListFilesEndpoint(cm *config.BackendConfigLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+
+	return func(c *fiber.Ctx) error {
+		var listFiles schema.ListFiles
+
+		purpose := c.Query("purpose")
+		if purpose == "" {
+			listFiles.Data = UploadedFiles
+		} else {
+			for _, f := range UploadedFiles {
+				if purpose == f.Purpose {
+					listFiles.Data = append(listFiles.Data, f)
+				}
+			}
+		}
+		listFiles.Object = "list"
+		return c.Status(fiber.StatusOK).JSON(listFiles)
+	}
+}
+
+func getFileFromRequest(c *fiber.Ctx) (*schema.File, error) {
+	id := c.Params("file_id")
+	if id == "" {
+		return nil, fmt.Errorf("file_id parameter is required")
+	}
+
+	for _, f := range UploadedFiles {
+		if id == f.ID {
+			return &f, nil
+		}
+	}
+
+	return nil, fmt.Errorf("unable to find file id %s", id)
+}
+
+// GetFilesEndpoint is the OpenAI API endpoint to get files https://platform.openai.com/docs/api-reference/files/retrieve
+// @Summary Returns information about a specific file.
+// @Success 200 {object} schema.File "Response"
+// @Router /v1/files/{file_id} [get]
+func GetFilesEndpoint(cm *config.BackendConfigLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		file, err := getFileFromRequest(c)
+		if err != nil {
+			return c.Status(fiber.StatusInternalServerError).SendString(bluemonday.StrictPolicy().Sanitize(err.Error()))
+		}
+
+		return c.JSON(file)
+	}
+}
+
+type DeleteStatus struct {
+	Id      string
+	Object  string
+	Deleted bool
+}
+
+// DeleteFilesEndpoint is the OpenAI API endpoint to delete files https://platform.openai.com/docs/api-reference/files/delete
+// @Summary Delete a file.
+// @Success 200 {object} DeleteStatus "Response"
+// @Router /v1/files/{file_id} [delete]
+func DeleteFilesEndpoint(cm *config.BackendConfigLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+
+	return func(c *fiber.Ctx) error {
+		file, err := getFileFromRequest(c)
+		if err != nil {
+			return c.Status(fiber.StatusInternalServerError).SendString(bluemonday.StrictPolicy().Sanitize(err.Error()))
+		}
+
+		err = os.Remove(filepath.Join(appConfig.UploadDir, file.Filename))
+		if err != nil {
+			// If the file doesn't exist then we should just continue to remove it
+			if !errors.Is(err, os.ErrNotExist) {
+				return c.Status(fiber.StatusInternalServerError).SendString(bluemonday.StrictPolicy().Sanitize(fmt.Sprintf("Unable to delete file: %s, %v", file.Filename, err)))
+			}
+		}
+
+		// Remove upload from list
+		for i, f := range UploadedFiles {
+			if f.ID == file.ID {
+				UploadedFiles = append(UploadedFiles[:i], UploadedFiles[i+1:]...)
+				break
+			}
+		}
+
+		utils.SaveConfig(appConfig.UploadDir, UploadedFilesFile, UploadedFiles)
+		return c.JSON(DeleteStatus{
+			Id:      file.ID,
+			Object:  "file",
+			Deleted: true,
+		})
+	}
+}
+
+// GetFilesContentsEndpoint is the OpenAI API endpoint to get files content https://platform.openai.com/docs/api-reference/files/retrieve-contents
+// @Summary Returns information about a specific file.
+// @Success	200		{string}	binary				"file"
+// @Router /v1/files/{file_id}/content [get]
+// GetFilesContentsEndpoint
+func GetFilesContentsEndpoint(cm *config.BackendConfigLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+		file, err := getFileFromRequest(c)
+		if err != nil {
+			return c.Status(fiber.StatusInternalServerError).SendString(bluemonday.StrictPolicy().Sanitize(err.Error()))
+		}
+
+		fileContents, err := os.ReadFile(filepath.Join(appConfig.UploadDir, file.Filename))
+		if err != nil {
+			return c.Status(fiber.StatusInternalServerError).SendString(bluemonday.StrictPolicy().Sanitize(err.Error()))
+		}
+
+		return c.Send(fileContents)
+	}
+}

core/http/endpoints/openai/files_test.go

@@ -0,0 +1,301 @@
+package openai
+
+import (
+	"encoding/json"
+	"fmt"
+	"io"
+	"mime/multipart"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"path/filepath"
+	"strings"
+
+	"github.com/rs/zerolog/log"
+
+	"github.com/mudler/LocalAI/core/config"
+	"github.com/mudler/LocalAI/core/schema"
+
+	"github.com/gofiber/fiber/v2"
+	utils2 "github.com/mudler/LocalAI/pkg/utils"
+	"github.com/stretchr/testify/assert"
+
+	"testing"
+)
+
+func startUpApp() (app *fiber.App, option *config.ApplicationConfig, loader *config.BackendConfigLoader) {
+	// Preparing the mocked objects
+	loader = &config.BackendConfigLoader{}
+
+	option = &config.ApplicationConfig{
+		UploadLimitMB: 10,
+		UploadDir:     "test_dir",
+	}
+
+	_ = os.RemoveAll(option.UploadDir)
+
+	app = fiber.New(fiber.Config{
+		BodyLimit: 20 * 1024 * 1024, // sets the limit to 20MB.
+	})
+
+	// Create a Test Server
+	app.Post("/files", UploadFilesEndpoint(loader, option))
+	app.Get("/files", ListFilesEndpoint(loader, option))
+	app.Get("/files/:file_id", GetFilesEndpoint(loader, option))
+	app.Delete("/files/:file_id", DeleteFilesEndpoint(loader, option))
+	app.Get("/files/:file_id/content", GetFilesContentsEndpoint(loader, option))
+
+	return
+}
+
+func TestUploadFileExceedSizeLimit(t *testing.T) {
+	// Preparing the mocked objects
+	loader := &config.BackendConfigLoader{}
+
+	option := &config.ApplicationConfig{
+		UploadLimitMB: 10,
+		UploadDir:     "test_dir",
+	}
+
+	_ = os.RemoveAll(option.UploadDir)
+
+	app := fiber.New(fiber.Config{
+		BodyLimit: 20 * 1024 * 1024, // sets the limit to 20MB.
+	})
+
+	// Create a Test Server
+	app.Post("/files", UploadFilesEndpoint(loader, option))
+	app.Get("/files", ListFilesEndpoint(loader, option))
+	app.Get("/files/:file_id", GetFilesEndpoint(loader, option))
+	app.Delete("/files/:file_id", DeleteFilesEndpoint(loader, option))
+	app.Get("/files/:file_id/content", GetFilesContentsEndpoint(loader, option))
+
+	t.Run("UploadFilesEndpoint file size exceeds limit", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		resp, err := CallFilesUploadEndpoint(t, app, "foo.txt", "file", "fine-tune", 11, option)
+		assert.NoError(t, err)
+
+		assert.Equal(t, fiber.StatusBadRequest, resp.StatusCode)
+		assert.Contains(t, bodyToString(resp, t), "exceeds upload limit")
+	})
+	t.Run("UploadFilesEndpoint purpose not defined", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		resp, _ := CallFilesUploadEndpoint(t, app, "foo.txt", "file", "", 5, option)
+
+		assert.Equal(t, fiber.StatusBadRequest, resp.StatusCode)
+		assert.Contains(t, bodyToString(resp, t), "Purpose is not defined")
+	})
+	t.Run("UploadFilesEndpoint file already exists", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		f1 := CallFilesUploadEndpointWithCleanup(t, app, "foo.txt", "file", "fine-tune", 5, option)
+
+		resp, err := CallFilesUploadEndpoint(t, app, "foo.txt", "file", "fine-tune", 5, option)
+		fmt.Println(f1)
+		fmt.Printf("ERror: %v\n", err)
+		fmt.Printf("resp: %+v\n", resp)
+
+		assert.Equal(t, fiber.StatusBadRequest, resp.StatusCode)
+		assert.Contains(t, bodyToString(resp, t), "File already exists")
+	})
+	t.Run("UploadFilesEndpoint file uploaded successfully", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		file := CallFilesUploadEndpointWithCleanup(t, app, "test.txt", "file", "fine-tune", 5, option)
+
+		// Check if file exists in the disk
+		testName := strings.Split(t.Name(), "/")[1]
+		fileName := testName + "-test.txt"
+		filePath := filepath.Join(option.UploadDir, utils2.SanitizeFileName(fileName))
+		_, err := os.Stat(filePath)
+
+		assert.False(t, os.IsNotExist(err))
+		assert.Equal(t, file.Bytes, 5242880)
+		assert.NotEmpty(t, file.CreatedAt)
+		assert.Equal(t, file.Filename, fileName)
+		assert.Equal(t, file.Purpose, "fine-tune")
+	})
+	t.Run("ListFilesEndpoint without purpose parameter", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		resp, err := CallListFilesEndpoint(t, app, "")
+		assert.NoError(t, err)
+
+		assert.Equal(t, 200, resp.StatusCode)
+
+		listFiles := responseToListFile(t, resp)
+		if len(listFiles.Data) != len(UploadedFiles) {
+			t.Errorf("Expected %v files, got %v files", len(UploadedFiles), len(listFiles.Data))
+		}
+	})
+	t.Run("ListFilesEndpoint with valid purpose parameter", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		_ = CallFilesUploadEndpointWithCleanup(t, app, "test.txt", "file", "fine-tune", 5, option)
+
+		resp, err := CallListFilesEndpoint(t, app, "fine-tune")
+		assert.NoError(t, err)
+
+		listFiles := responseToListFile(t, resp)
+		if len(listFiles.Data) != 1 {
+			t.Errorf("Expected 1 file, got %v files", len(listFiles.Data))
+		}
+	})
+	t.Run("ListFilesEndpoint with invalid query parameter", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		resp, err := CallListFilesEndpoint(t, app, "not-so-fine-tune")
+		assert.NoError(t, err)
+		assert.Equal(t, 200, resp.StatusCode)
+
+		listFiles := responseToListFile(t, resp)
+
+		if len(listFiles.Data) != 0 {
+			t.Errorf("Expected 0 file, got %v files", len(listFiles.Data))
+		}
+	})
+	t.Run("GetFilesContentsEndpoint get file content", func(t *testing.T) {
+		t.Cleanup(tearDown())
+		req := httptest.NewRequest("GET", "/files", nil)
+		resp, _ := app.Test(req)
+		assert.Equal(t, 200, resp.StatusCode)
+
+		var listFiles schema.ListFiles
+		if err := json.Unmarshal(bodyToByteArray(resp, t), &listFiles); err != nil {
+			t.Errorf("Failed to decode response: %v", err)
+			return
+		}
+
+		if len(listFiles.Data) != 0 {
+			t.Errorf("Expected 0 file, got %v files", len(listFiles.Data))
+		}
+	})
+}
+
+func CallListFilesEndpoint(t *testing.T, app *fiber.App, purpose string) (*http.Response, error) {
+	var target string
+	if purpose != "" {
+		target = fmt.Sprintf("/files?purpose=%s", purpose)
+	} else {
+		target = "/files"
+	}
+	req := httptest.NewRequest("GET", target, nil)
+	return app.Test(req)
+}
+
+func CallFilesContentEndpoint(t *testing.T, app *fiber.App, fileId string) (*http.Response, error) {
+	request := httptest.NewRequest("GET", "/files?file_id="+fileId, nil)
+	return app.Test(request)
+}
+
+func CallFilesUploadEndpoint(t *testing.T, app *fiber.App, fileName, tag, purpose string, fileSize int, appConfig *config.ApplicationConfig) (*http.Response, error) {
+	testName := strings.Split(t.Name(), "/")[1]
+
+	// Create a file that exceeds the limit
+	file := createTestFile(t, testName+"-"+fileName, fileSize, appConfig)
+
+	// Creating a new HTTP Request
+	body, writer := newMultipartFile(file.Name(), tag, purpose)
+
+	req := httptest.NewRequest(http.MethodPost, "/files", body)
+	req.Header.Set(fiber.HeaderContentType, writer.FormDataContentType())
+	return app.Test(req)
+}
+
+func CallFilesUploadEndpointWithCleanup(t *testing.T, app *fiber.App, fileName, tag, purpose string, fileSize int, appConfig *config.ApplicationConfig) schema.File {
+	// Create a file that exceeds the limit
+	testName := strings.Split(t.Name(), "/")[1]
+	file := createTestFile(t, testName+"-"+fileName, fileSize, appConfig)
+
+	// Creating a new HTTP Request
+	body, writer := newMultipartFile(file.Name(), tag, purpose)
+
+	req := httptest.NewRequest(http.MethodPost, "/files", body)
+	req.Header.Set(fiber.HeaderContentType, writer.FormDataContentType())
+	resp, err := app.Test(req)
+	assert.NoError(t, err)
+	f := responseToFile(t, resp)
+
+	//id := f.ID
+	//t.Cleanup(func() {
+	//	_, err := CallFilesDeleteEndpoint(t, app, id)
+	//	assert.NoError(t, err)
+	//	assert.Empty(t, UploadedFiles)
+	//})
+
+	return f
+
+}
+
+func CallFilesDeleteEndpoint(t *testing.T, app *fiber.App, fileId string) (*http.Response, error) {
+	target := fmt.Sprintf("/files/%s", fileId)
+	req := httptest.NewRequest(http.MethodDelete, target, nil)
+	return app.Test(req)
+}
+
+// Helper to create multi-part file
+func newMultipartFile(filePath, tag, purpose string) (*strings.Reader, *multipart.Writer) {
+	body := new(strings.Builder)
+	writer := multipart.NewWriter(body)
+	file, _ := os.Open(filePath)
+	defer file.Close()
+	part, _ := writer.CreateFormFile(tag, filepath.Base(filePath))
+	io.Copy(part, file)
+
+	if purpose != "" {
+		_ = writer.WriteField("purpose", purpose)
+	}
+
+	writer.Close()
+	return strings.NewReader(body.String()), writer
+}
+
+// Helper to create test files
+func createTestFile(t *testing.T, name string, sizeMB int, option *config.ApplicationConfig) *os.File {
+	err := os.MkdirAll(option.UploadDir, 0750)
+	if err != nil {
+
+		t.Fatalf("Error MKDIR: %v", err)
+	}
+
+	file, err := os.Create(name)
+	assert.NoError(t, err)
+	file.WriteString(strings.Repeat("a", sizeMB*1024*1024)) // sizeMB MB File
+
+	t.Cleanup(func() {
+		os.Remove(name)
+		os.RemoveAll(option.UploadDir)
+	})
+	return file
+}
+
+func bodyToString(resp *http.Response, t *testing.T) string {
+	return string(bodyToByteArray(resp, t))
+}
+
+func bodyToByteArray(resp *http.Response, t *testing.T) []byte {
+	bodyBytes, err := io.ReadAll(resp.Body)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return bodyBytes
+}
+
+func responseToFile(t *testing.T, resp *http.Response) schema.File {
+	var file schema.File
+	responseToString := bodyToString(resp, t)
+
+	err := json.NewDecoder(strings.NewReader(responseToString)).Decode(&file)
+	if err != nil {
+		t.Errorf("Failed to decode response: %s", err)
+	}
+
+	return file
+}
+
+func responseToListFile(t *testing.T, resp *http.Response) schema.ListFiles {
+	var listFiles schema.ListFiles
+	responseToString := bodyToString(resp, t)
+
+	err := json.NewDecoder(strings.NewReader(responseToString)).Decode(&listFiles)
+	if err != nil {
+		log.Error().Err(err).Msg("failed to decode response")
+	}
+
+	return listFiles
+}

core/http/endpoints/openai/image.go

@@ -79,37 +79,49 @@ func ImageEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appCon
 			return fiber.ErrBadRequest
 		}
 
-		// Process input images (for img2img/inpainting)
 		src := ""
 		if input.File != "" {
-			src = processImageFile(input.File, appConfig.GeneratedContentDir)
-			if src != "" {
-				defer os.RemoveAll(src)
-			}
-		}
 
-		// Process multiple input images
+			fileData := []byte{}
-		var inputImages []string
+			var err error
-		if len(input.Files) > 0 {
+			// check if input.File is an URL, if so download it and save it
-			for _, file := range input.Files {
+			// to a temporary file
-				processedFile := processImageFile(file, appConfig.GeneratedContentDir)
+			if strings.HasPrefix(input.File, "http://") || strings.HasPrefix(input.File, "https://") {
-				if processedFile != "" {
+				out, err := downloadFile(input.File)
-					inputImages = append(inputImages, processedFile)
+				if err != nil {
-					defer os.RemoveAll(processedFile)
+					return fmt.Errorf("failed downloading file:%w", err)
+				}
+				defer os.RemoveAll(out)
+
+				fileData, err = os.ReadFile(out)
+				if err != nil {
+					return fmt.Errorf("failed reading file:%w", err)
+				}
+
+			} else {
+				// base 64 decode the file and write it somewhere
+				// that we will cleanup
+				fileData, err = base64.StdEncoding.DecodeString(input.File)
+				if err != nil {
+					return err
 				}
 			}
-		}
 
-		// Process reference images
+			// Create a temporary file
-		var refImages []string
+			outputFile, err := os.CreateTemp(appConfig.GeneratedContentDir, "b64")
-		if len(input.RefImages) > 0 {
+			if err != nil {
-			for _, file := range input.RefImages {
+				return err
-				processedFile := processImageFile(file, appConfig.GeneratedContentDir)
-				if processedFile != "" {
-					refImages = append(refImages, processedFile)
-					defer os.RemoveAll(processedFile)
-				}
 			}
+			// write the base64 result
+			writer := bufio.NewWriter(outputFile)
+			_, err = writer.Write(fileData)
+			if err != nil {
+				outputFile.Close()
+				return err
+			}
+			outputFile.Close()
+			src = outputFile.Name()
+			defer os.RemoveAll(src)
 		}
 
 		log.Debug().Msgf("Parameter Config: %+v", config)
@@ -190,13 +202,7 @@ func ImageEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appCon
 
 				baseURL := c.BaseURL()
 
-				// Use the first input image as src if available, otherwise use the original src
+				fn, err := backend.ImageGeneration(height, width, mode, step, *config.Seed, positive_prompt, negative_prompt, src, output, ml, *config, appConfig)
-				inputSrc := src
-				if len(inputImages) > 0 {
-					inputSrc = inputImages[0]
-				}
-
-				fn, err := backend.ImageGeneration(height, width, mode, step, *config.Seed, positive_prompt, negative_prompt, inputSrc, output, ml, *config, appConfig, refImages)
 				if err != nil {
 					return err
 				}
@@ -237,51 +243,3 @@ func ImageEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appCon
 		return c.JSON(resp)
 	}
 }
-
-// processImageFile handles a single image file (URL or base64) and returns the path to the temporary file
-func processImageFile(file string, generatedContentDir string) string {
-	fileData := []byte{}
-	var err error
-
-	// check if file is an URL, if so download it and save it to a temporary file
-	if strings.HasPrefix(file, "http://") || strings.HasPrefix(file, "https://") {
-		out, err := downloadFile(file)
-		if err != nil {
-			log.Error().Err(err).Msgf("Failed downloading file: %s", file)
-			return ""
-		}
-		defer os.RemoveAll(out)
-
-		fileData, err = os.ReadFile(out)
-		if err != nil {
-			log.Error().Err(err).Msgf("Failed reading downloaded file: %s", out)
-			return ""
-		}
-	} else {
-		// base 64 decode the file and write it somewhere that we will cleanup
-		fileData, err = base64.StdEncoding.DecodeString(file)
-		if err != nil {
-			log.Error().Err(err).Msgf("Failed decoding base64 file")
-			return ""
-		}
-	}
-
-	// Create a temporary file
-	outputFile, err := os.CreateTemp(generatedContentDir, "b64")
-	if err != nil {
-		log.Error().Err(err).Msg("Failed creating temporary file")
-		return ""
-	}
-
-	// write the base64 result
-	writer := bufio.NewWriter(outputFile)
-	_, err = writer.Write(fileData)
-	if err != nil {
-		outputFile.Close()
-		log.Error().Err(err).Msg("Failed writing to temporary file")
-		return ""
-	}
-	outputFile.Close()
-
-	return outputFile.Name()
-}

core/http/endpoints/openai/realtime.go

@@ -16,12 +16,12 @@ import (
 	"github.com/mudler/LocalAI/core/application"
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/http/endpoints/openai/types"
-	"github.com/mudler/LocalAI/core/templates"
 	laudio "github.com/mudler/LocalAI/pkg/audio"
 	"github.com/mudler/LocalAI/pkg/functions"
 	"github.com/mudler/LocalAI/pkg/grpc/proto"
 	model "github.com/mudler/LocalAI/pkg/model"
 	"github.com/mudler/LocalAI/pkg/sound"
+	"github.com/mudler/LocalAI/pkg/templates"
 
 	"google.golang.org/grpc"
 
@@ -29,8 +29,8 @@ import (
 )
 
 const (
-	localSampleRate  = 16000
+	localSampleRate       = 16000
-	remoteSampleRate = 24000
+	remoteSampleRate      = 24000
 )
 
 // A model can be "emulated" that is: transcribe audio to text -> feed text to the LLM -> generate audio as result
@@ -210,9 +210,9 @@ func registerRealtime(application *application.Application) func(c *websocket.Co
 					// TODO: Need some way to pass this to the backend
 					Threshold: 0.5,
 					// TODO: This is ignored and the amount of padding is random at present
-					PrefixPaddingMs:   30,
+					PrefixPaddingMs: 30,
 					SilenceDurationMs: 500,
-					CreateResponse:    func() *bool { t := true; return &t }(),
+					CreateResponse: func() *bool { t := true; return &t }(),
 				},
 			},
 			InputAudioTranscription: &types.InputAudioTranscription{
@@ -233,7 +233,7 @@ func registerRealtime(application *application.Application) func(c *websocket.Co
 		// TODO: The API has no way to configure the VAD model or other models that make up a pipeline to fake any-to-any
 		//       So possibly we could have a way to configure a composite model that can be used in situations where any-to-any is expected
 		pipeline := config.Pipeline{
-			VAD:           "silero-vad",
+			VAD: "silero-vad",
 			Transcription: session.InputAudioTranscription.Model,
 		}
 
@@ -567,8 +567,8 @@ func updateTransSession(session *Session, update *types.ClientSession, cl *confi
 	trCur := session.InputAudioTranscription
 
 	if trUpd != nil && trUpd.Model != "" && trUpd.Model != trCur.Model {
-		pipeline := config.Pipeline{
+		pipeline := config.Pipeline {
-			VAD:           "silero-vad",
+			VAD: "silero-vad",
 			Transcription: trUpd.Model,
 		}
 
@@ -684,7 +684,7 @@ func handleVAD(cfg *config.BackendConfig, evaluator *templates.Evaluator, sessio
 				sendEvent(c, types.InputAudioBufferClearedEvent{
 					ServerEventBase: types.ServerEventBase{
 						EventID: "event_TODO",
-						Type:    types.ServerEventTypeInputAudioBufferCleared,
+						Type: types.ServerEventTypeInputAudioBufferCleared,
 					},
 				})
 
@@ -697,7 +697,7 @@ func handleVAD(cfg *config.BackendConfig, evaluator *templates.Evaluator, sessio
 				sendEvent(c, types.InputAudioBufferSpeechStartedEvent{
 					ServerEventBase: types.ServerEventBase{
 						EventID: "event_TODO",
-						Type:    types.ServerEventTypeInputAudioBufferSpeechStarted,
+						Type: types.ServerEventTypeInputAudioBufferSpeechStarted,
 					},
 					AudioStartMs: time.Now().Sub(startTime).Milliseconds(),
 				})
@@ -719,7 +719,7 @@ func handleVAD(cfg *config.BackendConfig, evaluator *templates.Evaluator, sessio
 				sendEvent(c, types.InputAudioBufferSpeechStoppedEvent{
 					ServerEventBase: types.ServerEventBase{
 						EventID: "event_TODO",
-						Type:    types.ServerEventTypeInputAudioBufferSpeechStopped,
+						Type: types.ServerEventTypeInputAudioBufferSpeechStopped,
 					},
 					AudioEndMs: time.Now().Sub(startTime).Milliseconds(),
 				})
@@ -728,9 +728,9 @@ func handleVAD(cfg *config.BackendConfig, evaluator *templates.Evaluator, sessio
 				sendEvent(c, types.InputAudioBufferCommittedEvent{
 					ServerEventBase: types.ServerEventBase{
 						EventID: "event_TODO",
-						Type:    types.ServerEventTypeInputAudioBufferCommitted,
+						Type: types.ServerEventTypeInputAudioBufferCommitted,
 					},
-					ItemID:         generateItemID(),
+					ItemID: generateItemID(),
 					PreviousItemID: "TODO",
 				})
 
@@ -833,9 +833,9 @@ func commitUtterance(ctx context.Context, utt []byte, cfg *config.BackendConfig,
 
 func runVAD(ctx context.Context, session *Session, adata []int16) ([]*proto.VADSegment, error) {
 	soundIntBuffer := &audio.IntBuffer{
-		Format:         &audio.Format{SampleRate: localSampleRate, NumChannels: 1},
+		Format: &audio.Format{SampleRate: localSampleRate, NumChannels: 1},
 		SourceBitDepth: 16,
-		Data:           sound.ConvertInt16ToInt(adata),
+		Data: sound.ConvertInt16ToInt(adata),
 	}
 
 	float32Data := soundIntBuffer.AsFloat32Buffer().Data

core/http/middleware/request.go

@@ -11,9 +11,9 @@ import (
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/schema"
 	"github.com/mudler/LocalAI/core/services"
-	"github.com/mudler/LocalAI/core/templates"
 	"github.com/mudler/LocalAI/pkg/functions"
 	"github.com/mudler/LocalAI/pkg/model"
+	"github.com/mudler/LocalAI/pkg/templates"
 	"github.com/mudler/LocalAI/pkg/utils"
 
 	"github.com/gofiber/fiber/v2"

core/http/routes/localai.go

@@ -23,7 +23,7 @@ func RegisterLocalAIRoutes(router *fiber.App,
 
 	// LocalAI API endpoints
 	if !appConfig.DisableGalleryEndpoint {
-		modelGalleryEndpointService := localai.CreateModelGalleryEndpointService(appConfig.Galleries, appConfig.BackendGalleries, appConfig.ModelPath, galleryService)
+		modelGalleryEndpointService := localai.CreateModelGalleryEndpointService(appConfig.Galleries, appConfig.ModelPath, galleryService)
 		router.Post("/models/apply", modelGalleryEndpointService.ApplyModelGalleryEndpoint())
 		router.Post("/models/delete/:name", modelGalleryEndpointService.DeleteModelGalleryEndpoint())
 
@@ -41,11 +41,6 @@ func RegisterLocalAIRoutes(router *fiber.App,
 		router.Get("/backends/jobs/:uuid", backendGalleryEndpointService.GetOpStatusEndpoint())
 	}
 
-	router.Post("/v1/detection",
-		requestExtractor.BuildFilteredFirstAvailableDefaultModel(config.BuildUsecaseFilterFn(config.FLAG_DETECTION)),
-		requestExtractor.SetModelAndConfig(func() schema.LocalAIRequest { return new(schema.DetectionRequest) }),
-		localai.DetectionEndpoint(cl, ml, appConfig))
-
 	router.Post("/tts",
 		requestExtractor.BuildFilteredFirstAvailableDefaultModel(config.BuildUsecaseFilterFn(config.FLAG_TTS)),
 		requestExtractor.SetModelAndConfig(func() schema.LocalAIRequest { return new(schema.TTSRequest) }),

core/http/routes/openai.go

@@ -54,6 +54,38 @@ func RegisterOpenAIRoutes(app *fiber.App,
 	app.Post("/completions", completionChain...)
 	app.Post("/v1/engines/:model/completions", completionChain...)
 
+	// assistant
+	app.Get("/v1/assistants", openai.ListAssistantsEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Get("/assistants", openai.ListAssistantsEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Post("/v1/assistants", openai.CreateAssistantEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Post("/assistants", openai.CreateAssistantEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Delete("/v1/assistants/:assistant_id", openai.DeleteAssistantEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Delete("/assistants/:assistant_id", openai.DeleteAssistantEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Get("/v1/assistants/:assistant_id", openai.GetAssistantEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Get("/assistants/:assistant_id", openai.GetAssistantEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Post("/v1/assistants/:assistant_id", openai.ModifyAssistantEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Post("/assistants/:assistant_id", openai.ModifyAssistantEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Get("/v1/assistants/:assistant_id/files", openai.ListAssistantFilesEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Get("/assistants/:assistant_id/files", openai.ListAssistantFilesEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Post("/v1/assistants/:assistant_id/files", openai.CreateAssistantFileEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Post("/assistants/:assistant_id/files", openai.CreateAssistantFileEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Delete("/v1/assistants/:assistant_id/files/:file_id", openai.DeleteAssistantFileEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Delete("/assistants/:assistant_id/files/:file_id", openai.DeleteAssistantFileEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Get("/v1/assistants/:assistant_id/files/:file_id", openai.GetAssistantFileEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+	app.Get("/assistants/:assistant_id/files/:file_id", openai.GetAssistantFileEndpoint(application.BackendLoader(), application.ModelLoader(), application.ApplicationConfig()))
+
+	// files
+	app.Post("/v1/files", openai.UploadFilesEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Post("/files", openai.UploadFilesEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Get("/v1/files", openai.ListFilesEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Get("/files", openai.ListFilesEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Get("/v1/files/:file_id", openai.GetFilesEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Get("/files/:file_id", openai.GetFilesEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Delete("/v1/files/:file_id", openai.DeleteFilesEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Delete("/files/:file_id", openai.DeleteFilesEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Get("/v1/files/:file_id/content", openai.GetFilesContentsEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+	app.Get("/files/:file_id/content", openai.GetFilesContentsEndpoint(application.BackendLoader(), application.ApplicationConfig()))
+
 	// embeddings
 	embeddingChain := []fiber.Handler{
 		re.BuildFilteredFirstAvailableDefaultModel(config.BuildUsecaseFilterFn(config.FLAG_EMBEDDINGS)),

core/http/routes/ui_gallery.go

@@ -180,7 +180,6 @@ func registerGalleryRoutes(app *fiber.App, cl *config.BackendConfigLoader, appCo
 			ID:                 uid,
 			GalleryElementName: galleryID,
 			Galleries:          appConfig.Galleries,
-			BackendGalleries:   appConfig.BackendGalleries,
 		}
 		go func() {
 			galleryService.ModelGalleryChannel <- op
@@ -220,7 +219,6 @@ func registerGalleryRoutes(app *fiber.App, cl *config.BackendConfigLoader, appCo
 			Delete:             true,
 			GalleryElementName: galleryName,
 			Galleries:          appConfig.Galleries,
-			BackendGalleries:   appConfig.BackendGalleries,
 		}
 		go func() {
 			galleryService.ModelGalleryChannel <- op

core/http/views/backends.html

@@ -90,14 +90,6 @@
                         hx-indicator=".htmx-indicator">
                         <i class="fas fa-headphones mr-2"></i>Whisper
                     </button>
-                    <button hx-post="browse/search/backends" 
-                        class="inline-flex items-center rounded-full px-4 py-2 text-sm font-medium bg-red-900/60 text-red-200 border border-red-700/50 hover:bg-red-800 transition duration-200 ease-in-out"
-                        hx-target="#search-results" 
-                        hx-vals='{"search": "object-detection"}'
-                        onclick="hidePagination()"
-                        hx-indicator=".htmx-indicator">
-                        <i class="fas fa-eye mr-2"></i>Object detection
-                    </button>
                 </div>
             </div>
         </div>

core/http/views/models.html

@@ -115,14 +115,6 @@
                         hx-indicator=".htmx-indicator">
                         <i class="fas fa-headphones mr-2"></i>Audio transcription
                     </button>
-                    <button hx-post="browse/search/models"
-                        class="inline-flex items-center rounded-full px-4 py-2 text-sm font-medium bg-red-900/60 text-red-200 border border-red-700/50 hover:bg-red-800 transition duration-200 ease-in-out"
-                        hx-target="#search-results" 
-                        hx-vals='{"search": "object-detection"}'
-                        onclick="hidePagination()"
-                        hx-indicator=".htmx-indicator">
-                        <i class="fas fa-eye mr-2"></i>Object detection
-                    </button>
                 </div>
             </div>
             

core/p2p/p2p.go

@@ -278,7 +278,6 @@ func ensureService(ctx context.Context, n *node.Node, nd *NodeData, sserv string
 			port, err := freeport.GetFreePort()
 			if err != nil {
 				zlog.Error().Err(err).Msgf("Could not allocate a free port for %s", nd.ID)
-				cancel()
 				return
 			}
 

core/schema/localai.go

@@ -120,20 +120,3 @@ type SystemInformationResponse struct {
 	Backends []string       `json:"backends"`
 	Models   []SysInfoModel `json:"loaded_models"`
 }
-
-type DetectionRequest struct {
-	BasicModelRequest
-	Image string `json:"image"`
-}
-
-type DetectionResponse struct {
-	Detections []Detection `json:"detections"`
-}
-
-type Detection struct {
-	X         float32 `json:"x"`
-	Y         float32 `json:"y"`
-	Width     float32 `json:"width"`
-	Height    float32 `json:"height"`
-	ClassName string  `json:"class_name"`
-}

core/schema/openai.go

@@ -2,6 +2,7 @@ package schema
 
 import (
 	"context"
+	"time"
 
 	functions "github.com/mudler/LocalAI/pkg/functions"
 )
@@ -114,6 +115,37 @@ type OpenAIModel struct {
 	Object string `json:"object"`
 }
 
+type DeleteAssistantResponse struct {
+	ID      string `json:"id"`
+	Object  string `json:"object"`
+	Deleted bool   `json:"deleted"`
+}
+
+// File represents the structure of a file object from the OpenAI API.
+type File struct {
+	ID        string    `json:"id"`         // Unique identifier for the file
+	Object    string    `json:"object"`     // Type of the object (e.g., "file")
+	Bytes     int       `json:"bytes"`      // Size of the file in bytes
+	CreatedAt time.Time `json:"created_at"` // The time at which the file was created
+	Filename  string    `json:"filename"`   // The name of the file
+	Purpose   string    `json:"purpose"`    // The purpose of the file (e.g., "fine-tune", "classifications", etc.)
+}
+
+type ListFiles struct {
+	Data   []File
+	Object string
+}
+
+type AssistantFileRequest struct {
+	FileID string `json:"file_id"`
+}
+
+type DeleteAssistantFileResponse struct {
+	ID      string `json:"id"`
+	Object  string `json:"object"`
+	Deleted bool   `json:"deleted"`
+}
+
 type ImageGenerationResponseFormat string
 
 type ChatCompletionResponseFormatType string
@@ -141,10 +173,6 @@ type OpenAIRequest struct {
 
 	// whisper
 	File string `json:"file" validate:"required"`
-	// Multiple input images for img2img or inpainting
-	Files []string `json:"files,omitempty"`
-	// Reference images for models that support them (e.g., Flux Kontext)
-	RefImages []string `json:"ref_images,omitempty"`
 	//whisper/image
 	ResponseFormat interface{} `json:"response_format,omitempty"`
 	// image

core/services/backends.go

@@ -2,7 +2,7 @@ package services
 
 import (
 	"github.com/mudler/LocalAI/core/gallery"
-	"github.com/mudler/LocalAI/pkg/system"
+	"github.com/mudler/LocalAI/core/system"
 
 	"github.com/mudler/LocalAI/pkg/utils"
 	"github.com/rs/zerolog/log"
@@ -24,7 +24,6 @@ func (g *GalleryService) backendHandler(op *GalleryOp[gallery.GalleryBackend], s
 		g.modelLoader.DeleteExternalBackend(op.GalleryElementName)
 	} else {
 		log.Warn().Msgf("installing backend %s", op.GalleryElementName)
-		log.Debug().Msgf("backend galleries: %v", g.appConfig.BackendGalleries)
 		err = gallery.InstallBackendFromGallery(g.appConfig.BackendGalleries, systemState, op.GalleryElementName, g.appConfig.BackendsPath, progressCallback, true)
 		if err == nil {
 			err = gallery.RegisterBackends(g.appConfig.BackendsPath, g.modelLoader)

core/services/gallery.go

@@ -7,8 +7,8 @@ import (
 
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/gallery"
+	"github.com/mudler/LocalAI/core/system"
 	"github.com/mudler/LocalAI/pkg/model"
-	"github.com/mudler/LocalAI/pkg/system"
 	"github.com/rs/zerolog/log"
 )
 

core/services/models.go

@@ -7,7 +7,7 @@ import (
 
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/gallery"
-	"github.com/mudler/LocalAI/pkg/system"
+	"github.com/mudler/LocalAI/core/system"
 	"github.com/mudler/LocalAI/pkg/utils"
 	"gopkg.in/yaml.v2"
 )

core/system/capabilities.go

@@ -18,13 +18,6 @@ const (
 	nvidiaL4T         = "nvidia-l4t"
 	darwinX86         = "darwin-x86"
 	metal             = "metal"
-	nvidia            = "nvidia"
-	amd               = "amd"
-	intel             = "intel"
-
-	capabilityEnv        = "LOCALAI_FORCE_META_BACKEND_CAPABILITY"
-	capabilityRunFileEnv = "LOCALAI_FORCE_META_BACKEND_CAPABILITY_RUN_FILE"
-	defaultRunFile       = "/run/localai/capability"
 )
 
 func (s *SystemState) Capability(capMap map[string]string) string {
@@ -32,26 +25,21 @@ func (s *SystemState) Capability(capMap map[string]string) string {
 
 	// Check if the reported capability is in the map
 	if _, exists := capMap[reportedCapability]; exists {
-		log.Debug().Str("reportedCapability", reportedCapability).Any("capMap", capMap).Msg("Using reported capability")
 		return reportedCapability
 	}
 
-	log.Debug().Str("reportedCapability", reportedCapability).Any("capMap", capMap).Msg("The requested capability was not found, using default capability")
 	// Otherwise, return the default capability (catch-all)
 	return defaultCapability
 }
 
 func (s *SystemState) getSystemCapabilities() string {
-	capability := os.Getenv(capabilityEnv)
+	if os.Getenv("LOCALAI_FORCE_META_BACKEND_CAPABILITY") != "" {
-	if capability != "" {
+		return os.Getenv("LOCALAI_FORCE_META_BACKEND_CAPABILITY")
-		log.Info().Str("capability", capability).Msgf("Using forced capability from environment variable (%s)", capabilityEnv)
-		return capability
 	}
 
-	capabilityRunFile := defaultRunFile
+	capabilityRunFile := "/run/localai/capability"
-	capabilityRunFileEnv := os.Getenv(capabilityRunFileEnv)
+	if os.Getenv("LOCALAI_FORCE_META_BACKEND_CAPABILITY_RUN_FILE") != "" {
-	if capabilityRunFileEnv != "" {
+		capabilityRunFile = os.Getenv("LOCALAI_FORCE_META_BACKEND_CAPABILITY_RUN_FILE")
-		capabilityRunFile = capabilityRunFileEnv
 	}
 
 	// Check if /run/localai/capability exists and use it
@@ -60,37 +48,31 @@ func (s *SystemState) getSystemCapabilities() string {
 	if _, err := os.Stat(capabilityRunFile); err == nil {
 		capability, err := os.ReadFile(capabilityRunFile)
 		if err == nil {
-			log.Info().Str("capabilityRunFile", capabilityRunFile).Str("capability", string(capability)).Msgf("Using forced capability run file (%s)", capabilityRunFileEnv)
+			return string(capability)
-			return strings.Trim(strings.TrimSpace(string(capability)), "\n")
 		}
 	}
 
 	// If we are on mac and arm64, we will return metal
 	if runtime.GOOS == "darwin" && runtime.GOARCH == "arm64" {
-		log.Info().Msgf("Using metal capability (arm64 on mac), set %s to override", capabilityEnv)
 		return metal
 	}
 
 	// If we are on mac and x86, we will return darwin-x86
 	if runtime.GOOS == "darwin" && runtime.GOARCH == "amd64" {
-		log.Info().Msgf("Using darwin-x86 capability (amd64 on mac), set %s to override", capabilityEnv)
 		return darwinX86
 	}
 
 	// If arm64 on linux and a nvidia gpu is detected, we will return nvidia-l4t
 	if runtime.GOOS == "linux" && runtime.GOARCH == "arm64" {
 		if s.GPUVendor == "nvidia" {
-			log.Info().Msgf("Using nvidia-l4t capability (arm64 on linux), set %s to override", capabilityEnv)
 			return nvidiaL4T
 		}
 	}
 
 	if s.GPUVendor == "" {
-		log.Info().Msgf("Default capability (no GPU detected), set %s to override", capabilityEnv)
 		return defaultCapability
 	}
 
-	log.Info().Str("Capability", s.GPUVendor).Msgf("Capability automatically detected, set %s to override", capabilityEnv)
 	return s.GPUVendor
 }
 
@@ -114,16 +96,18 @@ func detectGPUVendor() (string, error) {
 			if gpu.DeviceInfo.Vendor != nil {
 				gpuVendorName := strings.ToUpper(gpu.DeviceInfo.Vendor.Name)
 				if strings.Contains(gpuVendorName, "NVIDIA") {
-					return nvidia, nil
+					return "nvidia", nil
 				}
 				if strings.Contains(gpuVendorName, "AMD") {
-					return amd, nil
+					return "amd", nil
 				}
 				if strings.Contains(gpuVendorName, "INTEL") {
-					return intel, nil
+					return "intel", nil
 				}
+				return "nvidia", nil
 			}
 		}
+
 	}
 
 	return "", nil

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

@@ -448,7 +448,7 @@ there are additional environment variables available that modify the behavior of
 | Environment variable       | Default | Description                                                                                                |
 |----------------------------|---------|------------------------------------------------------------------------------------------------------------|
 | `REBUILD`                  | `false` | Rebuild LocalAI on startup                                                                                 |
-| `BUILD_TYPE`               |         | Build type. Available: `cublas`, `openblas`, `clblas`, `intel` (intel core), `sycl_f16`, `sycl_f32` (intel backends)                                                      |
+| `BUILD_TYPE`               |         | Build type. Available: `cublas`, `openblas`, `clblas`                                                      |
 | `GO_TAGS`                  |         | Go tags. Available: `stablediffusion`                                                                      |
 | `HUGGINGFACEHUB_API_TOKEN` |         | Special token for interacting with HuggingFace Inference API, required only when using the `langchain-huggingface` backend |
 | `EXTRA_BACKENDS`          |         | A space separated list of backends to prepare. For example `EXTRA_BACKENDS="backend/python/diffusers backend/python/transformers"` prepares the python environment on start |

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

@@ -15,16 +15,6 @@ This section contains instruction on how to use LocalAI with GPU acceleration.
 For acceleration for AMD or Metal HW is still in development, for additional details see the [build]({{%relref "docs/getting-started/build#Acceleration" %}})
 {{% /alert %}}
 
-## Automatic Backend Detection
-
-When you install a model from the gallery (or a YAML file), LocalAI intelligently detects the required backend and your system's capabilities, then downloads the correct version for you. Whether you're running on a standard CPU, an NVIDIA GPU, an AMD GPU, or an Intel GPU, LocalAI handles it automatically.
-
-For advanced use cases or to override auto-detection, you can use the `LOCALAI_FORCE_META_BACKEND_CAPABILITY` environment variable. Here are the available options:
-
-- `default`: Forces CPU-only backend. This is the fallback if no specific hardware is detected.
-- `nvidia`: Forces backends compiled with CUDA support for NVIDIA GPUs.
-- `amd`: Forces backends compiled with ROCm support for AMD GPUs.
-- `intel`: Forces backends compiled with SYCL/oneAPI support for Intel GPUs.
 
 ## Model configuration
 
@@ -81,8 +71,8 @@ To use CUDA, use the images with the `cublas` tag, for example.
 
 The image list is on [quay](https://quay.io/repository/go-skynet/local-ai?tab=tags):
 
-- CUDA `11` tags: `master-gpu-nvidia-cuda-11`, `v1.40.0-gpu-nvidia-cuda-11`, ...
+- CUDA `11` tags: `master-gpu-nvidia-cuda11`, `v1.40.0-gpu-nvidia-cuda11`, ...
-- CUDA `12` tags: `master-gpu-nvidia-cuda-12`, `v1.40.0-gpu-nvidia-cuda-12`, ...
+- CUDA `12` tags: `master-gpu-nvidia-cuda12`, `v1.40.0-gpu-nvidia-cuda12`, ...
 
 In addition to the commands to run LocalAI normally, you need to specify `--gpus all` to docker, for example:
 
@@ -267,7 +257,7 @@ If building from source, you need to install [Intel oneAPI Base Toolkit](https:/
 
 ### Container images
 
-To use SYCL, use the images with `gpu-intel` in the tag, for example `{{< version >}}-gpu-intel`, ...
+To use SYCL, use the images with the `gpu-intel-f16` or `gpu-intel-f32` tag, for example `{{< version >}}-gpu-intel-f32-core`, `{{< version >}}-gpu-intel-f16`, ...
 
 The image list is on [quay](https://quay.io/repository/go-skynet/local-ai?tab=tags).
 
@@ -276,7 +266,7 @@ The image list is on [quay](https://quay.io/repository/go-skynet/local-ai?tab=ta
 To run LocalAI with Docker and sycl starting `phi-2`, you can use the following command as an example:
 
 ```bash
-docker run -e DEBUG=true --privileged -ti -v $PWD/models:/models -p 8080:8080  -v /dev/dri:/dev/dri --rm quay.io/go-skynet/local-ai:master-gpu-intel phi-2
+docker run -e DEBUG=true --privileged -ti -v $PWD/models:/models -p 8080:8080  -v /dev/dri:/dev/dri --rm quay.io/go-skynet/local-ai:master-gpu-intel-f32 phi-2
 ```
 
 ### Notes
@@ -284,7 +274,7 @@ docker run -e DEBUG=true --privileged -ti -v $PWD/models:/models -p 8080:8080  -
 In addition to the commands to run LocalAI normally, you need to specify `--device /dev/dri` to docker, for example:
 
 ```bash
-docker run --rm -ti --device /dev/dri -p 8080:8080 -e DEBUG=true -e MODELS_PATH=/models -e THREADS=1 -v $PWD/models:/models quay.io/go-skynet/local-ai:{{< version >}}-gpu-intel
+docker run --rm -ti --device /dev/dri -p 8080:8080 -e DEBUG=true -e MODELS_PATH=/models -e THREADS=1 -v $PWD/models:/models quay.io/go-skynet/local-ai:{{< version >}}-gpu-intel-f16
 ```
 
 Note also that sycl does have a known issue to hang with `mmap: true`. You have to disable it in the model configuration if explicitly enabled.

docs/content/docs/features/backends.md

@@ -96,8 +96,8 @@ Your backend container should:
 For getting started, see the available backends in LocalAI here: https://github.com/mudler/LocalAI/tree/master/backend . 
 
 - For Python based backends there is a template that can be used as starting point: https://github.com/mudler/LocalAI/tree/master/backend/python/common/template . 
-- For Golang based backends, you can see the `bark-cpp` backend as an example: https://github.com/mudler/LocalAI/tree/master/backend/go/bark-cpp
+- For Golang based backends, you can see the `bark-cpp` backend as an example: https://github.com/mudler/LocalAI/tree/master/backend/go/bark
-- For C++ based backends, you can see the `llama-cpp` backend as an example: https://github.com/mudler/LocalAI/tree/master/backend/cpp/llama-cpp
+- For C++ based backends, you can see the `llama-cpp` backend as an example: https://github.com/mudler/LocalAI/tree/master/backend/cpp/llama
 
 ### Publishing Your Backend
 

docs/content/docs/features/object-detection.md

@@ -1,193 +0,0 @@
-+++
-disableToc = false
-title = "🔍 Object detection"
-weight = 13
-url = "/features/object-detection/"
-+++
-
-LocalAI supports object detection through various backends. This feature allows you to identify and locate objects within images with high accuracy and real-time performance. Currently, [RF-DETR](https://github.com/roboflow/rf-detr) is available as an implementation.
-
-## Overview
-
-Object detection in LocalAI is implemented through dedicated backends that can identify and locate objects within images. Each backend provides different capabilities and model architectures.
-
-**Key Features:**
-- Real-time object detection
-- High accuracy detection with bounding boxes
-- Support for multiple hardware accelerators (CPU, NVIDIA GPU, Intel GPU, AMD GPU)
-- Structured detection results with confidence scores
-- Easy integration through the `/v1/detection` endpoint
-
-## Usage
-
-### Detection Endpoint
-
-LocalAI provides a dedicated `/v1/detection` endpoint for object detection tasks. This endpoint is specifically designed for object detection and returns structured detection results with bounding boxes and confidence scores.
-
-### API Reference
-
-To perform object detection, send a POST request to the `/v1/detection` endpoint:
-
-```bash
-curl -X POST http://localhost:8080/v1/detection \
-  -H "Content-Type: application/json" \
-  -d '{
-    "model": "rfdetr-base",
-    "image": "https://media.roboflow.com/dog.jpeg"
-  }'
-```
-
-### Request Format
-
-The request body should contain:
-
-- `model`: The name of the object detection model (e.g., "rfdetr-base")
-- `image`: The image to analyze, which can be:
-  - A URL to an image
-  - A base64-encoded image
-
-### Response Format
-
-The API returns a JSON response with detected objects:
-
-```json
-{
-  "detections": [
-    {
-      "x": 100.5,
-      "y": 150.2,
-      "width": 200.0,
-      "height": 300.0,
-      "confidence": 0.95,
-      "class_name": "dog"
-    },
-    {
-      "x": 400.0,
-      "y": 200.0,
-      "width": 150.0,
-      "height": 250.0,
-      "confidence": 0.87,
-      "class_name": "person"
-    }
-  ]
-}
-```
-
-Each detection includes:
-- `x`, `y`: Coordinates of the bounding box top-left corner
-- `width`, `height`: Dimensions of the bounding box
-- `confidence`: Detection confidence score (0.0 to 1.0)
-- `class_name`: The detected object class
-
-## Backends
-
-### RF-DETR Backend
-
-The RF-DETR backend is implemented as a Python-based gRPC service that integrates seamlessly with LocalAI. It provides object detection capabilities using the RF-DETR model architecture and supports multiple hardware configurations:
-
-- **CPU**: Optimized for CPU inference
-- **NVIDIA GPU**: CUDA acceleration for NVIDIA GPUs
-- **Intel GPU**: Intel oneAPI optimization
-- **AMD GPU**: ROCm acceleration for AMD GPUs
-- **NVIDIA Jetson**: Optimized for ARM64 NVIDIA Jetson devices
-
-#### Setup
-
-1. **Using the Model Gallery (Recommended)**
-
-   The easiest way to get started is using the model gallery. The `rfdetr-base` model is available in the official LocalAI gallery:
-
-   ```bash
-   # Install and run the rfdetr-base model
-   local-ai run rfdetr-base
-   ```
-
-   You can also install it through the web interface by navigating to the Models section and searching for "rfdetr-base".
-
-2. **Manual Configuration**
-
-   Create a model configuration file in your `models` directory:
-
-   ```yaml
-   name: rfdetr
-   backend: rfdetr
-   parameters:
-     model: rfdetr-base
-   ```
-
-#### Available Models
-
-Currently, the following model is available in the [Model Gallery]({{%relref "docs/features/model-gallery" %}}):
-
-- **rfdetr-base**: Base model with balanced performance and accuracy
-
-You can browse and install this model through the LocalAI web interface or using the command line.
-
-## Examples
-
-### Basic Object Detection
-
-```bash
-# Detect objects in an image from URL
-curl -X POST http://localhost:8080/v1/detection \
-  -H "Content-Type: application/json" \
-  -d '{
-    "model": "rfdetr-base",
-    "image": "https://example.com/image.jpg"
-  }'
-```
-
-### Base64 Image Detection
-
-```bash
-# Convert image to base64 and send
-base64_image=$(base64 -w 0 image.jpg)
-curl -X POST http://localhost:8080/v1/detection \
-  -H "Content-Type: application/json" \
-  -d "{
-    \"model\": \"rfdetr-base\",
-    \"image\": \"data:image/jpeg;base64,$base64_image\"
-  }"
-```
-
-## Troubleshooting
-
-### Common Issues
-
-1. **Model Loading Errors**
-   - Ensure the model file is properly downloaded
-   - Check available disk space
-   - Verify model compatibility with your backend version
-
-2. **Low Detection Accuracy**
-   - Ensure good image quality and lighting
-   - Check if objects are clearly visible
-   - Consider using a larger model for better accuracy
-
-3. **Slow Performance**
-   - Enable GPU acceleration if available
-   - Use a smaller model for faster inference
-   - Optimize image resolution
-
-### Debug Mode
-
-Enable debug logging for troubleshooting:
-
-```bash
-local-ai run --debug rfdetr-base
-```
-
-## Object Detection Category
-
-LocalAI includes a dedicated **object-detection** category for models and backends that specialize in identifying and locating objects within images. This category currently includes:
-
-- **RF-DETR**: Real-time transformer-based object detection
-
-Additional object detection models and backends will be added to this category in the future. You can filter models by the `object-detection` tag in the model gallery to find all available object detection models.
-
-## Related Features
-
-- [🎨 Image generation]({{%relref "docs/features/image-generation" %}}): Generate images with AI
-- [📖 Text generation]({{%relref "docs/features/text-generation" %}}): Generate text with language models
-- [🔍 GPT Vision]({{%relref "docs/features/gpt-vision" %}}): Analyze images with language models
-- [🚀 GPU acceleration]({{%relref "docs/features/GPU-acceleration" %}}): Optimize performance with GPU acceleration 

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

@@ -9,11 +9,13 @@ ico = "rocket_launch"
 
 ### Build
 
-LocalAI can be built as a container image or as a single, portable binary. Note that some model architectures might require Python libraries, which are not included in the binary.
+LocalAI can be built as a container image or as a single, portable binary. Note that some model architectures might require Python libraries, which are not included in the binary. The binary contains only the core backends written in Go and C++. 
 
 LocalAI's extensible architecture allows you to add your own backends, which can be written in any language, and as such the container images contains also the Python dependencies to run all the available backends (for example, in order to run backends like __Diffusers__ that allows to generate images and videos from text).
 
-This section contains instructions on how to build LocalAI from source.
+In some cases you might want to re-build LocalAI from source (for instance to leverage Apple Silicon acceleration), or to build a custom container image with your own backends. This section contains instructions on how to build LocalAI from source.
+
+
 
 #### Build LocalAI locally
 
@@ -22,6 +24,7 @@ This section contains instructions on how to build LocalAI from source.
 In order to build LocalAI locally, you need the following requirements:
 
 - Golang >= 1.21
+- Cmake/make
 - GCC
 - GRPC
 
@@ -33,14 +36,20 @@ To install the dependencies follow the instructions below:
 Install `xcode` from the App Store
 
 ```bash
-brew install go protobuf protoc-gen-go protoc-gen-go-grpc wget
+brew install abseil cmake go grpc protobuf protoc-gen-go protoc-gen-go-grpc python wget
+```
+
+After installing the above dependencies, you need to install grpcio-tools from PyPI. You could do this via a pip --user install or a virtualenv.
+
+```bash
+pip install --user grpcio-tools
 ```
 
 {{% /tab %}}
 {{% tab tabName="Debian" %}}
 
 ```bash
-apt install golang make protobuf-compiler-grpc
+apt install cmake golang libgrpc-dev make protobuf-compiler-grpc python3-grpc-tools
 ```
 
 After you have golang installed and working, you can install the required binaries for compiling the golang protobuf components via the following commands
@@ -54,8 +63,10 @@ go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f1
 {{% /tab %}}
 {{% tab tabName="From source" %}}
 
+Specify `BUILD_GRPC_FOR_BACKEND_LLAMA=true` to build automatically the gRPC dependencies
+
 ```bash
-make build
+make ... BUILD_GRPC_FOR_BACKEND_LLAMA=true build
 ```
 
 {{% /tab %}}
@@ -72,6 +83,36 @@ make build
 
 This should produce the binary `local-ai`
 
+Here is the list of the variables available that can be used to customize the build:
+
+| Variable | Default | Description |
+| ---------------------| ------- | ----------- |
+| `BUILD_TYPE`         |   None      | Build type. Available: `cublas`, `openblas`, `clblas`, `metal`,`hipblas`, `sycl_f16`, `sycl_f32` |
+| `GO_TAGS`            |   `tts stablediffusion`      | Go tags. Available: `stablediffusion`, `tts` |
+| `CLBLAST_DIR`        |         | Specify a CLBlast directory |
+| `CUDA_LIBPATH`       |         | Specify a CUDA library path |
+| `BUILD_API_ONLY` | false | Set to true to build only the API (no backends will be built) |
+
+{{% alert note %}}
+
+#### CPU flagset compatibility
+
+
+LocalAI uses different backends based on ggml and llama.cpp to run models. If your CPU doesn't support common instruction sets, you can disable them during build:
+
+```
+CMAKE_ARGS="-DGGML_F16C=OFF -DGGML_AVX512=OFF -DGGML_AVX2=OFF -DGGML_AVX=OFF -DGGML_FMA=OFF" make build
+```
+
+To have effect on the container image, you need to set `REBUILD=true`:
+
+```
+docker run  quay.io/go-skynet/localai
+docker run --rm -ti -p 8080:8080 -e DEBUG=true -e MODELS_PATH=/models -e THREADS=1 -e REBUILD=true -e CMAKE_ARGS="-DGGML_F16C=OFF -DGGML_AVX512=OFF -DGGML_AVX2=OFF -DGGML_AVX=OFF -DGGML_FMA=OFF" -v $PWD/models:/models quay.io/go-skynet/local-ai:latest
+```
+
+{{% /alert %}}
+
 #### Container image
 
 Requirements:
@@ -112,9 +153,6 @@ wget https://huggingface.co/TheBloke/phi-2-GGUF/resolve/main/phi-2.Q2_K.gguf -O
 # Use a template from the examples
 cp -rf prompt-templates/ggml-gpt4all-j.tmpl models/phi-2.Q2_K.tmpl
 
-# Install the llama-cpp backend
-./local-ai backends install llama-cpp
-
 # Run LocalAI
 ./local-ai --models-path=./models/ --debug=true
 
@@ -148,53 +186,131 @@ sudo xcode-select --switch /Applications/Xcode.app/Contents/Developer
 
 ```
 # reinstall build dependencies
-brew reinstall go grpc protobuf wget
+brew reinstall abseil cmake go grpc protobuf wget
 
 make clean
 
 make build
 ```
 
-## Build backends
+**Requirements**: OpenCV, Gomp
 
-LocalAI have several backends available for installation in the backend gallery. The backends can be also built by source. As backends might vary from language and dependencies that they require, the documentation will provide generic guidance for few of the backends, which can be applied with some slight modifications also to the others.
+Image generation requires `GO_TAGS=stablediffusion` to be set during build:
-
-### Manually
-
-Typically each backend include a Makefile which allow to package the backend.
-
-In the LocalAI repository, for instance you can build `bark-cpp` by doing:
 
 ```
-git clone https://github.com/go-skynet/LocalAI.git
+make GO_TAGS=stablediffusion build
-
-# Build the bark-cpp backend (requires cmake)
-make -C LocalAI/backend/go/bark-cpp build package
-
-# Build vllm backend (requires python)
-make -C LocalAI/backend/python/vllm
 ```
 
-### With Docker
+### Build with Text to audio support
 
-Building with docker is simpler as abstracts away all the requirement, and focuses on building the final OCI images that are available in the gallery. This allows for instance also to build locally a backend and install it with LocalAI. You can refer to [Backends](https://localai.io/backends/) for general guidance on how to install and develop backends.
+**Requirements**: piper-phonemize
 
-In the LocalAI repository, you can build `bark-cpp` by doing:
+Text to audio support is experimental and requires `GO_TAGS=tts` to be set during build:
 
 ```
-git clone https://github.com/go-skynet/LocalAI.git
+make GO_TAGS=tts build
-
-# Build the bark-cpp backend (requires docker)
-make docker-build-bark-cpp
 ```
 
-Note that `make` is only by convenience, in reality it just runs a simple `docker` command as:
+### Acceleration
+
+#### OpenBLAS
+
+Software acceleration.
+
+Requirements: OpenBLAS
+
+```
+make BUILD_TYPE=openblas build
+```
+
+#### CuBLAS
+
+Nvidia Acceleration.
+
+Requirement: Nvidia CUDA toolkit
+
+Note: CuBLAS support is experimental, and has not been tested on real HW. please report any issues you find!
+
+```
+make BUILD_TYPE=cublas build
+```
+
+More informations available in the upstream PR: https://github.com/ggerganov/llama.cpp/pull/1412
+
+
+#### Hipblas (AMD GPU with ROCm on Arch Linux)
+
+Packages:
+```
+pacman -S base-devel git rocm-hip-sdk rocm-opencl-sdk opencv clblast grpc
+```
+
+Library links:
+```
+export CGO_CFLAGS="-I/usr/include/opencv4"
+export CGO_CXXFLAGS="-I/usr/include/opencv4"
+export CGO_LDFLAGS="-L/opt/rocm/hip/lib -lamdhip64 -L/opt/rocm/lib -lOpenCL -L/usr/lib -lclblast -lrocblas -lhipblas -lrocrand -lomp -O3 --rtlib=compiler-rt -unwindlib=libgcc -lhipblas -lrocblas --hip-link"
+```
+
+Build:
+```
+make BUILD_TYPE=hipblas GPU_TARGETS=gfx1030
+```
+
+#### ClBLAS
+
+AMD/Intel GPU acceleration.
+
+Requirement: OpenCL, CLBlast
+
+```
+make BUILD_TYPE=clblas build
+```
+
+To specify a clblast dir set: `CLBLAST_DIR`
+
+#### Intel GPU acceleration
+
+Intel GPU acceleration is supported via SYCL.
+
+Requirements: [Intel oneAPI Base Toolkit](https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit-download.html) (see also [llama.cpp setup installations instructions](https://github.com/ggerganov/llama.cpp/blob/d71ac90985854b0905e1abba778e407e17f9f887/README-sycl.md?plain=1#L56))
+
+```
+make BUILD_TYPE=sycl_f16 build # for float16
+make BUILD_TYPE=sycl_f32 build # for float32
+```
+
+#### Metal (Apple Silicon)
+
+```
+make build
+
+# correct build type is automatically used on mac (BUILD_TYPE=metal)
+# Set `gpu_layers: 256` (or equal to the number of model layers) to your YAML model config file and `f16: true`
+```
+
+### Windows compatibility
+
+Make sure to give enough resources to the running container. See https://github.com/go-skynet/LocalAI/issues/2
+
+### Examples
+
+More advanced build options are available, for instance to build only a single backend.
+
+#### Build only a single backend
+
+You can control the backends that are built by setting the `GRPC_BACKENDS` environment variable. For instance, to build only the `llama-cpp` backend only:
 
 ```bash
-docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:bark-cpp -f LocalAI/backend/Dockerfile.golang --build-arg BACKEND=bark-cpp .               
+make GRPC_BACKENDS=backend-assets/grpc/llama-cpp build
 ```
 
-Note:
+By default, all the backends are built.
 
-- BUILD_TYPE can be either: `cublas`, `hipblas`, `sycl_f16`, `sycl_f32`, `metal`.
+#### Specific llama.cpp version
-- BASE_IMAGE is tested on `ubuntu:22.04` (and defaults to it) and `quay.io/go-skynet/intel-oneapi-base:latest` for intel/sycl
+
+To build with a specific version of llama.cpp, set `CPPLLAMA_VERSION` to the tag or wanted sha:
+
+```
+CPPLLAMA_VERSION=<sha> make build
+```

docs/content/docs/getting-started/container-images.md

@@ -131,7 +131,8 @@ docker run -p 8080:8080 --name local-ai -ti -v localai-models:/models localai/lo
 | Latest images for Nvidia GPU (CUDA11) | `quay.io/go-skynet/local-ai:latest-aio-gpu-nvidia-cuda-11` | `localai/localai:latest-aio-gpu-nvidia-cuda-11`                      |
 | Latest images for Nvidia GPU (CUDA12) | `quay.io/go-skynet/local-ai:latest-aio-gpu-nvidia-cuda-12` | `localai/localai:latest-aio-gpu-nvidia-cuda-12`                      |
 | Latest images for AMD GPU | `quay.io/go-skynet/local-ai:latest-aio-gpu-hipblas` | `localai/localai:latest-aio-gpu-hipblas`                      |
-| Latest images for Intel GPU | `quay.io/go-skynet/local-ai:latest-aio-gpu-intel` | `localai/localai:latest-aio-gpu-intel`                      |
+| Latest images for Intel GPU (sycl f16) | `quay.io/go-skynet/local-ai:latest-aio-gpu-intel-f16` | `localai/localai:latest-aio-gpu-intel-f16`                      |
+| Latest images for Intel GPU (sycl f32) | `quay.io/go-skynet/local-ai:latest-aio-gpu-intel-f32` | `localai/localai:latest-aio-gpu-intel-f32`                      |
 
 ### Available environment variables
 
@@ -162,9 +163,9 @@ Standard container images do not have pre-installed models.
 
 | Description | Quay | Docker Hub                                                  |
 | --- | --- |-------------------------------------------------------------|
-| Latest images from the branch (development) | `quay.io/go-skynet/local-ai:master-gpu-nvidia-cuda-11` | `localai/localai:master-gpu-nvidia-cuda-11`                      |
+| Latest images from the branch (development) | `quay.io/go-skynet/local-ai:master-gpu-nvidia-cuda11` | `localai/localai:master-gpu-nvidia-cuda11`                      |
 | Latest tag | `quay.io/go-skynet/local-ai:latest-gpu-nvidia-cuda-11` | `localai/localai:latest-gpu-nvidia-cuda-11`                      |
-| Versioned image | `quay.io/go-skynet/local-ai:{{< version >}}-gpu-nvidia-cuda-11` | `localai/localai:{{< version >}}-gpu-nvidia-cuda-11`             |
+| Versioned image | `quay.io/go-skynet/local-ai:{{< version >}}-gpu-nvidia-cuda11` | `localai/localai:{{< version >}}-gpu-nvidia-cuda11`             |
 
 {{% /tab %}}
 
@@ -172,19 +173,29 @@ Standard container images do not have pre-installed models.
 
 | Description | Quay | Docker Hub                                                  |
 | --- | --- |-------------------------------------------------------------|
-| Latest images from the branch (development) | `quay.io/go-skynet/local-ai:master-gpu-nvidia-cuda-12` | `localai/localai:master-gpu-nvidia-cuda-12`                      |
+| Latest images from the branch (development) | `quay.io/go-skynet/local-ai:master-gpu-nvidia-cuda12` | `localai/localai:master-gpu-nvidia-cuda12`                      |
 | Latest tag | `quay.io/go-skynet/local-ai:latest-gpu-nvidia-cuda-12` | `localai/localai:latest-gpu-nvidia-cuda-12`                 |
-| Versioned image | `quay.io/go-skynet/local-ai:{{< version >}}-gpu-nvidia-cuda-12` | `localai/localai:{{< version >}}-gpu-nvidia-cuda-12`             |
+| Versioned image | `quay.io/go-skynet/local-ai:{{< version >}}-gpu-nvidia-cuda12` | `localai/localai:{{< version >}}-gpu-nvidia-cuda12`             |
 
 {{% /tab %}}
 
-{{% tab tabName="Intel GPU" %}}
+{{% tab tabName="Intel GPU (sycl f16)" %}}
 
 | Description | Quay | Docker Hub                                                  |
 | --- | --- |-------------------------------------------------------------|
-| Latest images from the branch (development) | `quay.io/go-skynet/local-ai:master-gpu-intel` | `localai/localai:master-gpu-intel`                      |
+| Latest images from the branch (development) | `quay.io/go-skynet/local-ai:master-gpu-intel-f16` | `localai/localai:master-gpu-intel-f16`                      |
-| Latest tag | `quay.io/go-skynet/local-ai:latest-gpu-intel` | `localai/localai:latest-gpu-intel`                      |
+| Latest tag | `quay.io/go-skynet/local-ai:latest-gpu-intel-f16` | `localai/localai:latest-gpu-intel-f16`                      |
-| Versioned image | `quay.io/go-skynet/local-ai:{{< version >}}-gpu-intel` | `localai/localai:{{< version >}}-gpu-intel`             |
+| Versioned image | `quay.io/go-skynet/local-ai:{{< version >}}-gpu-intel-f16` | `localai/localai:{{< version >}}-gpu-intel-f16`             |
+
+{{% /tab %}}
+
+{{% tab tabName="Intel GPU (sycl f32)" %}}
+
+| Description | Quay | Docker Hub                                                  |
+| --- | --- |-------------------------------------------------------------|
+| Latest images from the branch (development) | `quay.io/go-skynet/local-ai:master-gpu-intel-f32` | `localai/localai:master-gpu-intel-f32`                      |
+| Latest tag | `quay.io/go-skynet/local-ai:latest-gpu-intel-f32` | `localai/localai:latest-gpu-intel-f32`                      |
+| Versioned image | `quay.io/go-skynet/local-ai:{{< version >}}-gpu-intel-f32` | `localai/localai:{{< version >}}-gpu-intel-f32`             |
 
 {{% /tab %}}
 

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

@@ -59,7 +59,11 @@ docker run -ti --name local-ai -p 8080:8080 --device=/dev/kfd --device=/dev/dri
 #### Intel GPU Images (oneAPI):
 
 ```bash
-docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-gpu-intel
+# Intel GPU with FP16 support
+docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-gpu-intel-f16
+
+# Intel GPU with FP32 support
+docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-gpu-intel-f32
 ```
 
 #### Vulkan GPU Images:
@@ -81,7 +85,7 @@ docker run -ti --name local-ai -p 8080:8080 --gpus all localai/localai:latest-ai
 docker run -ti --name local-ai -p 8080:8080 --gpus all localai/localai:latest-aio-gpu-nvidia-cuda-11
 
 # Intel GPU version
-docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-aio-gpu-intel
+docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-aio-gpu-intel-f16
 
 # AMD GPU version
 docker run -ti --name local-ai -p 8080:8080 --device=/dev/kfd --device=/dev/dri --group-add=video localai/localai:latest-aio-gpu-hipblas
@@ -102,9 +106,6 @@ local-ai run https://gist.githubusercontent.com/.../phi-2.yaml
 local-ai run oci://localai/phi-2:latest
 ```
 
-{{% alert icon="⚡" %}}
-**Automatic Backend Detection**: When you install models from the gallery or YAML files, LocalAI automatically detects your system's GPU capabilities (NVIDIA, AMD, Intel) and downloads the appropriate backend. For advanced configuration options, see [GPU Acceleration]({{% relref "docs/features/gpu-acceleration#automatic-backend-detection" %}}).
-{{% /alert %}}
 
 For a full list of options, refer to the [Installer Options]({{% relref "docs/advanced/installer" %}}) documentation.
 
@@ -153,7 +154,7 @@ For instructions on using AIO images, see [Using container images]({{% relref "d
 
 LocalAI is part of the Local family stack, along with LocalAGI and LocalRecall.
 
-[LocalAGI](https://github.com/mudler/LocalAGI) is a powerful, self-hostable AI Agent platform designed for maximum privacy and flexibility which encompassess and uses all the software stack. It provides a complete drop-in replacement for OpenAI's Responses APIs with advanced agentic capabilities, working entirely locally on consumer-grade hardware (CPU and GPU).
+[LocalAGI](https://github.com/mudler/LocalAGI) is a powerful, self-hostable AI Agent platform designed for maximum privacy and flexibility which encompassess and uses all the softwre stack. It provides a complete drop-in replacement for OpenAI's Responses APIs with advanced agentic capabilities, working entirely locally on consumer-grade hardware (CPU and GPU).
 
 ### Quick Start
 

docs/data/version.json

@@ -1,3 +1,3 @@
 {
-  "version": "v3.3.0"
+  "version": "v3.1.1"
 }

docs/static/install.sh

@@ -660,7 +660,7 @@ install_docker() {
 
     IMAGE_TAG=
     if [ "$USE_VULKAN" = true ]; then
-        IMAGE_TAG=${LOCALAI_VERSION}-gpu-vulkan
+        IMAGE_TAG=${LOCALAI_VERSION}-vulkan
 
         info "Starting LocalAI Docker container..."
         $SUDO docker run -v local-ai-data:/models \
@@ -672,7 +672,7 @@ install_docker() {
             -d -p $PORT:8080 --name local-ai localai/localai:$IMAGE_TAG $STARTCOMMAND
     elif [ "$HAS_CUDA" ]; then
         # Default to CUDA 12
-        IMAGE_TAG=${LOCALAI_VERSION}-gpu-nvidia-cuda-12
+        IMAGE_TAG=${LOCALAI_VERSION}-gpu-nvidia-cuda12
         # AIO
         if [ "$USE_AIO" = true ]; then
             IMAGE_TAG=${LOCALAI_VERSION}-aio-gpu-nvidia-cuda-12
@@ -715,10 +715,11 @@ install_docker() {
             $envs \
             -d -p $PORT:8080 --name local-ai localai/localai:$IMAGE_TAG $STARTCOMMAND
     elif [ "$HAS_INTEL" ]; then
-        IMAGE_TAG=${LOCALAI_VERSION}-gpu-intel
+        # Default to FP32 for better compatibility
+        IMAGE_TAG=${LOCALAI_VERSION}-gpu-intel-f32
         # AIO
         if [ "$USE_AIO" = true ]; then
-            IMAGE_TAG=${LOCALAI_VERSION}-aio-gpu-intel
+            IMAGE_TAG=${LOCALAI_VERSION}-aio-gpu-intel-f32
         fi
 
         info "Starting LocalAI Docker container..."
@@ -756,7 +757,7 @@ install_binary_darwin() {
     [ "$(uname -s)" = "Darwin" ] || fatal 'This script is intended to run on macOS only.'
 
     info "Downloading LocalAI ${LOCALAI_VERSION}..."
-    curl --fail --show-error --location --progress-bar -o $TEMP_DIR/local-ai "https://github.com/mudler/LocalAI/releases/download/${LOCALAI_VERSION}/local-ai-${LOCALAI_VERSION}-darwin-${ARCH}"
+    curl --fail --show-error --location --progress-bar -o $TEMP_DIR/local-ai "https://github.com/mudler/LocalAI/releases/download/${LOCALAI_VERSION}/local-ai-Darwin-${ARCH}"
 
     info "Installing to /usr/local/bin/local-ai"
     install -o0 -g0 -m755 $TEMP_DIR/local-ai /usr/local/bin/local-ai
@@ -788,7 +789,7 @@ install_binary() {
     fi
 
     info "Downloading LocalAI ${LOCALAI_VERSION}..."
-    curl --fail --location --progress-bar -o $TEMP_DIR/local-ai "https://github.com/mudler/LocalAI/releases/download/${LOCALAI_VERSION}/local-ai-${LOCALAI_VERSION}-linux-${ARCH}"
+    curl --fail --location --progress-bar -o $TEMP_DIR/local-ai "https://github.com/mudler/LocalAI/releases/download/${LOCALAI_VERSION}/local-ai-Linux-${ARCH}"
 
     for BINDIR in /usr/local/bin /usr/bin /bin; do
         echo $PATH | grep -q $BINDIR && break || continue
@@ -867,7 +868,7 @@ OS="$(uname -s)"
 
 ARCH=$(uname -m)
 case "$ARCH" in
-    x86_64) ARCH="amd64" ;;
+    x86_64) ARCH="x86_64" ;;
     aarch64|arm64) ARCH="arm64" ;;
     *) fatal "Unsupported architecture: $ARCH" ;;
 esac

gallery/index.yaml

@@ -1,54 +1,4 @@
 ---
-- &afm
-  name: "arcee-ai_afm-4.5b"
-  url: "github:mudler/LocalAI/gallery/chatml.yaml@master"
-  icon: https://cdn-uploads.huggingface.co/production/uploads/6435718aaaef013d1aec3b8b/Lj9YVLIKKdImV_jID0A1g.png
-  license: aml
-  urls:
-    - https://huggingface.co/arcee-ai/AFM-4.5B
-    - https://huggingface.co/bartowski/arcee-ai_AFM-4.5B-GGUF
-  tags:
-    - gguf
-    - gpu
-    - gpu
-    - text-generation
-  description: |
-    AFM-4.5B is a 4.5 billion parameter instruction-tuned model developed by Arcee.ai, designed for enterprise-grade performance across diverse deployment environments from cloud to edge. The base model was trained on a dataset of 8 trillion tokens, comprising 6.5 trillion tokens of general pretraining data followed by 1.5 trillion tokens of midtraining data with enhanced focus on mathematical reasoning and code generation. Following pretraining, the model underwent supervised fine-tuning on high-quality instruction datasets. The instruction-tuned model was further refined through reinforcement learning on verifiable rewards as well as for human preference. We use a modified version of TorchTitan for pretraining, Axolotl for supervised fine-tuning, and a modified version of Verifiers for reinforcement learning.
-
-    The development of AFM-4.5B prioritized data quality as a fundamental requirement for achieving robust model performance. We collaborated with DatologyAI, a company specializing in large-scale data curation. DatologyAI's curation pipeline integrates a suite of proprietary algorithms—model-based quality filtering, embedding-based curation, target distribution-matching, source mixing, and synthetic data. Their expertise enabled the creation of a curated dataset tailored to support strong real-world performance.
-
-    The model architecture follows a standard transformer decoder-only design based on Vaswani et al., incorporating several key modifications for enhanced performance and efficiency. Notable architectural features include grouped query attention for improved inference efficiency and ReLU^2 activation functions instead of SwiGLU to enable sparsification while maintaining or exceeding performance benchmarks.
-
-    The model available in this repo is the instruct model following supervised fine-tuning and reinforcement learning.
-  overrides:
-    parameters:
-      model: arcee-ai_AFM-4.5B-Q4_K_M.gguf
-  files:
-    - filename: arcee-ai_AFM-4.5B-Q4_K_M.gguf
-      sha256: f05516b323f581bebae1af2cbf900d83a2569b0a60c54366daf4a9c15ae30d4f
-      uri: huggingface://bartowski/arcee-ai_AFM-4.5B-GGUF/arcee-ai_AFM-4.5B-Q4_K_M.gguf
-- &rfdetr
-  name: "rfdetr-base"
-  url: "github:mudler/LocalAI/gallery/virtual.yaml@master"
-  icon: https://avatars.githubusercontent.com/u/53104118?s=200&v=4
-  license: apache-2.0
-  description: |
-    RF-DETR is a real-time, transformer-based object detection model architecture developed by Roboflow and released under the Apache 2.0 license.
-    RF-DETR is the first real-time model to exceed 60 AP on the Microsoft COCO benchmark alongside competitive performance at base sizes. It also achieves state-of-the-art performance on RF100-VL, an object detection benchmark that measures model domain adaptability to real world problems. RF-DETR is fastest and most accurate for its size when compared current real-time objection models.
-    RF-DETR is small enough to run on the edge using Inference, making it an ideal model for deployments that need both strong accuracy and real-time performance.
-  tags:
-    - object-detection
-    - rfdetr
-    - gpu
-    - cpu
-  urls:
-    - https://github.com/roboflow/rf-detr
-  overrides:
-    backend: rfdetr
-    parameters:
-      model: rfdetr-base
-    known_usecases:
-      - detection
 - name: "dream-org_dream-v0-instruct-7b"
   # chatml
   url: "github:mudler/LocalAI/gallery/chatml.yaml@master"
@@ -1906,43 +1856,6 @@
     - filename: Menlo_Lucy-128k-Q4_K_M.gguf
       sha256: fb3e591cccc5d2821f3c615fd6dc2ca86d409f56fbc124275510a9612a90e61f
       uri: huggingface://bartowski/Menlo_Lucy-128k-GGUF/Menlo_Lucy-128k-Q4_K_M.gguf
-- !!merge <<: *qwen3
-  name: "qwen_qwen3-30b-a3b-instruct-2507"
-  urls:
-    - https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507
-    - https://huggingface.co/bartowski/Qwen_Qwen3-30B-A3B-Instruct-2507-GGUF
-  description: |
-    We introduce the updated version of the Qwen3-30B-A3B non-thinking mode, named Qwen3-30B-A3B-Instruct-2507, featuring the following key enhancements:
-
-        Significant improvements in general capabilities, including instruction following, logical reasoning, text comprehension, mathematics, science, coding and tool usage.
-        Substantial gains in long-tail knowledge coverage across multiple languages.
-        Markedly better alignment with user preferences in subjective and open-ended tasks, enabling more helpful responses and higher-quality text generation.
-        Enhanced capabilities in 256K long-context understanding.
-  overrides:
-    parameters:
-      model: Qwen_Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf
-  files:
-    - filename: Qwen_Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf
-      sha256: 382b4f5a164d200f93790ee0e339fae12852896d23485cfb203ce868fea33a95
-      uri: huggingface://bartowski/Qwen_Qwen3-30B-A3B-Instruct-2507-GGUF/Qwen_Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf
-- !!merge <<: *qwen3
-  name: "qwen_qwen3-30b-a3b-thinking-2507"
-  urls:
-    - https://huggingface.co/Qwen/Qwen3-30B-A3B-Thinking-2507
-    - https://huggingface.co/bartowski/Qwen_Qwen3-30B-A3B-Thinking-2507-GGUF
-  description: |
-    Over the past three months, we have continued to scale the thinking capability of Qwen3-30B-A3B, improving both the quality and depth of reasoning. We are pleased to introduce Qwen3-30B-A3B-Thinking-2507, featuring the following key enhancements:
-    Significantly improved performance on reasoning tasks, including logical reasoning, mathematics, science, coding, and academic benchmarks that typically require human expertise.
-    Markedly better general capabilities, such as instruction following, tool usage, text generation, and alignment with human preferences.
-    Enhanced 256K long-context understanding capabilities.
-    NOTE: This version has an increased thinking length. We strongly recommend its use in highly complex reasoning tasks.
-  overrides:
-    parameters:
-      model: Qwen_Qwen3-30B-A3B-Thinking-2507-Q4_K_M.gguf
-  files:
-    - filename: Qwen_Qwen3-30B-A3B-Thinking-2507-Q4_K_M.gguf
-      sha256: 1359aa08e2f2dfe7ce4b5ff88c4c996e6494c9d916b1ebacd214bb74bbd5a9db
-      uri: huggingface://bartowski/Qwen_Qwen3-30B-A3B-Thinking-2507-GGUF/Qwen_Qwen3-30B-A3B-Thinking-2507-Q4_K_M.gguf
 - &gemma3
   url: "github:mudler/LocalAI/gallery/gemma.yaml@master"
   name: "gemma-3-27b-it"
@@ -19144,148 +19057,6 @@
   overrides:
     parameters:
       model: SicariusSicariiStuff/flux.1dev-abliteratedv2
-- name: flux.1-kontext-dev
-  license: flux-1-dev-non-commercial-license
-  url: "github:mudler/LocalAI/gallery/flux-ggml.yaml@master"
-  icon: https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev/media/main/teaser.png
-  description: |
-      FLUX.1 Kontext [dev] is a 12 billion parameter rectified flow transformer capable of editing images based on text instructions. For more information, please read our blog post and our technical report. You can find information about the [pro] version in here.
-      Key Features
-          Change existing images based on an edit instruction.
-          Have character, style and object reference without any finetuning.
-          Robust consistency allows users to refine an image through multiple successive edits with minimal visual drift.
-          Trained using guidance distillation, making FLUX.1 Kontext [dev] more efficient.
-          Open weights to drive new scientific research, and empower artists to develop innovative workflows.
-          Generated outputs can be used for personal, scientific, and commercial purposes, as described in the FLUX.1 [dev] Non-Commercial License.
-  urls:
-    - https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev
-    - https://huggingface.co/QuantStack/FLUX.1-Kontext-dev-GGUF
-  tags:
-    - image-to-image
-    - flux
-    - gpu
-    - cpu
-  overrides:
-    parameters:
-      model: flux1-kontext-dev-Q8_0.gguf
-  files:
-    - filename: "flux1-kontext-dev-Q8_0.gguf"
-      sha256: "ff2ff71c3755c8ab394398a412252c23382a83138b65190b16e736d457b80f73"
-      uri: "huggingface://QuantStack/FLUX.1-Kontext-dev-GGUF/flux1-kontext-dev-Q8_0.gguf"
-    - filename: ae.safetensors
-      sha256: afc8e28272cd15db3919bacdb6918ce9c1ed22e96cb12c4d5ed0fba823529e38
-      uri: https://huggingface.co/ChuckMcSneed/FLUX.1-dev/resolve/main/ae.safetensors
-    - filename: clip_l.safetensors
-      sha256: 660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors
-    - filename: t5xxl_fp16.safetensors
-      sha256: 6e480b09fae049a72d2a8c5fbccb8d3e92febeb233bbe9dfe7256958a9167635
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors
-- !!merge <<: *flux
-  name: flux.1-dev-ggml-q8_0
-  license: flux-1-dev-non-commercial-license
-  url: "github:mudler/LocalAI/gallery/flux-ggml.yaml@master"
-  urls:
-    - https://huggingface.co/black-forest-labs/FLUX.1-dev
-    - https://huggingface.co/city96/FLUX.1-dev-gguf
-  overrides:
-    parameters:
-      model: flux1-dev-Q8_0.gguf
-  files:
-    - filename: "flux1-dev-Q8_0.gguf"
-      sha256: "129032f32224bf7138f16e18673d8008ba5f84c1ec74063bf4511a8bb4cf553d"
-      uri: "huggingface://city96/FLUX.1-dev-gguf/flux1-dev-Q8_0.gguf"
-    - filename: ae.safetensors
-      sha256: afc8e28272cd15db3919bacdb6918ce9c1ed22e96cb12c4d5ed0fba823529e38
-      uri: https://huggingface.co/ChuckMcSneed/FLUX.1-dev/resolve/main/ae.safetensors
-    - filename: clip_l.safetensors
-      sha256: 660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors
-    - filename: t5xxl_fp16.safetensors
-      sha256: 6e480b09fae049a72d2a8c5fbccb8d3e92febeb233bbe9dfe7256958a9167635
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors
-- !!merge <<: *flux
-  name: flux.1-dev-ggml-abliterated-v2-q8_0
-  url: "github:mudler/LocalAI/gallery/flux-ggml.yaml@master"
-  description: |
-    FLUX.1 [dev] is an abliterated version of FLUX.1 [dev]
-  urls:
-    - https://huggingface.co/black-forest-labs/FLUX.1-dev
-    - https://huggingface.co/t8star/flux.1-dev-abliterated-V2-GGUF
-  overrides:
-    parameters:
-      model: T8-flux.1-dev-abliterated-V2-GGUF-Q8_0.gguf
-  files:
-    - filename: "T8-flux.1-dev-abliterated-V2-GGUF-Q8_0.gguf"
-      sha256: "aba8163ff644018da195212a1c33aeddbf802a0c2bba96abc584a2d0b6b42272"
-      uri: "huggingface://t8star/flux.1-dev-abliterated-V2-GGUF/T8-flux.1-dev-abliterated-V2-GGUF-Q8_0.gguf"
-    - filename: ae.safetensors
-      sha256: afc8e28272cd15db3919bacdb6918ce9c1ed22e96cb12c4d5ed0fba823529e38
-      uri: https://huggingface.co/ChuckMcSneed/FLUX.1-dev/resolve/main/ae.safetensors
-    - filename: clip_l.safetensors
-      sha256: 660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors
-    - filename: t5xxl_fp16.safetensors
-      sha256: 6e480b09fae049a72d2a8c5fbccb8d3e92febeb233bbe9dfe7256958a9167635
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors
-- !!merge <<: *flux
-  name: flux.1-krea-dev-ggml
-  url: "github:mudler/LocalAI/gallery/flux-ggml.yaml@master"
-  description: |
-    FLUX.1 Krea [dev] is a 12 billion parameter rectified flow transformer capable of generating images from text descriptions. For more information, please read our blog post and Krea's blog post.
-    Cutting-edge output quality, with a focus on aesthetic photography.
-    Competitive prompt following, matching the performance of closed source alternatives.
-    Trained using guidance distillation, making FLUX.1 Krea [dev] more efficient.
-    Open weights to drive new scientific research, and empower artists to develop innovative workflows.
-    Generated outputs can be used for personal, scientific, and commercial purposes, as described in the flux-1-dev-non-commercial-license.
-  urls:
-    - https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev
-    - https://huggingface.co/QuantStack/FLUX.1-Krea-dev-GGUF
-  overrides:
-    parameters:
-      model: flux1-krea-dev-Q4_K_M.gguf
-  files:
-    - filename: "flux1-krea-dev-Q4_K_M.gguf"
-      sha256: "cf199b88509be2b3476a3372ff03eaaa662cb2b5d3710abf939ebb4838dbdcaf"
-      uri: "huggingface://QuantStack/FLUX.1-Krea-dev-GGUF/flux1-krea-dev-Q4_K_M.gguf"
-    - filename: ae.safetensors
-      sha256: afc8e28272cd15db3919bacdb6918ce9c1ed22e96cb12c4d5ed0fba823529e38
-      uri: https://huggingface.co/ChuckMcSneed/FLUX.1-dev/resolve/main/ae.safetensors
-    - filename: clip_l.safetensors
-      sha256: 660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors
-    - filename: t5xxl_fp16.safetensors
-      sha256: 6e480b09fae049a72d2a8c5fbccb8d3e92febeb233bbe9dfe7256958a9167635
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors
-- !!merge <<: *flux
-  name: flux.1-krea-dev-ggml-q8_0
-  url: "github:mudler/LocalAI/gallery/flux-ggml.yaml@master"
-  description: |
-    FLUX.1 Krea [dev] is a 12 billion parameter rectified flow transformer capable of generating images from text descriptions. For more information, please read our blog post and Krea's blog post.
-    Cutting-edge output quality, with a focus on aesthetic photography.
-    Competitive prompt following, matching the performance of closed source alternatives.
-    Trained using guidance distillation, making FLUX.1 Krea [dev] more efficient.
-    Open weights to drive new scientific research, and empower artists to develop innovative workflows.
-    Generated outputs can be used for personal, scientific, and commercial purposes, as described in the flux-1-dev-non-commercial-license.
-  urls:
-    - https://huggingface.co/black-forest-labs/FLUX.1-Krea-dev
-    - https://huggingface.co/markury/FLUX.1-Krea-dev-gguf
-  overrides:
-    parameters:
-      model: flux1-krea-dev-Q8_0.gguf
-  files:
-    - filename: "flux1-krea-dev-Q8_0.gguf"
-      sha256: "0d085b1e3ae0b90e5dbf74da049a80a565617de622a147d28ee37a07761fbd90"
-      uri: "huggingface://markury/FLUX.1-Krea-dev-gguf/flux1-krea-dev-Q8_0.gguf"
-    - filename: ae.safetensors
-      sha256: afc8e28272cd15db3919bacdb6918ce9c1ed22e96cb12c4d5ed0fba823529e38
-      uri: https://huggingface.co/ChuckMcSneed/FLUX.1-dev/resolve/main/ae.safetensors
-    - filename: clip_l.safetensors
-      sha256: 660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors
-    - filename: t5xxl_fp16.safetensors
-      sha256: 6e480b09fae049a72d2a8c5fbccb8d3e92febeb233bbe9dfe7256958a9167635
-      uri: https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors
 - &whisper
   url: "github:mudler/LocalAI/gallery/whisper-base.yaml@master" ## Whisper
   name: "whisper-1"

pkg/grpc/backend.go

@@ -9,7 +9,7 @@ import (
 
 var embeds = map[string]*embedBackend{}
 
-func Provide(addr string, llm AIModel) {
+func Provide(addr string, llm LLM) {
 	embeds[addr] = &embedBackend{s: &server{llm: llm}}
 }
 
@@ -42,7 +42,6 @@ type Backend interface {
 	GenerateVideo(ctx context.Context, in *pb.GenerateVideoRequest, opts ...grpc.CallOption) (*pb.Result, error)
 	TTS(ctx context.Context, in *pb.TTSRequest, opts ...grpc.CallOption) (*pb.Result, error)
 	SoundGeneration(ctx context.Context, in *pb.SoundGenerationRequest, opts ...grpc.CallOption) (*pb.Result, error)
-	Detect(ctx context.Context, in *pb.DetectOptions, opts ...grpc.CallOption) (*pb.DetectResponse, error)
 	AudioTranscription(ctx context.Context, in *pb.TranscriptRequest, opts ...grpc.CallOption) (*pb.TranscriptResult, error)
 	TokenizeString(ctx context.Context, in *pb.PredictOptions, opts ...grpc.CallOption) (*pb.TokenizationResponse, error)
 	Status(ctx context.Context) (*pb.StatusResponse, error)

pkg/grpc/base/base.go

@@ -69,10 +69,6 @@ func (llm *Base) SoundGeneration(*pb.SoundGenerationRequest) error {
 	return fmt.Errorf("unimplemented")
 }
 
-func (llm *Base) Detect(*pb.DetectOptions) (pb.DetectResponse, error) {
-	return pb.DetectResponse{}, fmt.Errorf("unimplemented")
-}
-
 func (llm *Base) TokenizeString(opts *pb.PredictOptions) (pb.TokenizationResponse, error) {
 	return pb.TokenizationResponse{}, fmt.Errorf("unimplemented")
 }

pkg/grpc/client.go

@@ -504,25 +504,3 @@ func (c *Client) VAD(ctx context.Context, in *pb.VADRequest, opts ...grpc.CallOp
 	client := pb.NewBackendClient(conn)
 	return client.VAD(ctx, in, opts...)
 }
-
-func (c *Client) Detect(ctx context.Context, in *pb.DetectOptions, opts ...grpc.CallOption) (*pb.DetectResponse, error) {
-	if !c.parallel {
-		c.opMutex.Lock()
-		defer c.opMutex.Unlock()
-	}
-	c.setBusy(true)
-	defer c.setBusy(false)
-	c.wdMark()
-	defer c.wdUnMark()
-	conn, err := grpc.Dial(c.address, grpc.WithTransportCredentials(insecure.NewCredentials()),
-		grpc.WithDefaultCallOptions(
-			grpc.MaxCallRecvMsgSize(50*1024*1024), // 50MB
-			grpc.MaxCallSendMsgSize(50*1024*1024), // 50MB
-		))
-	if err != nil {
-		return nil, err
-	}
-	defer conn.Close()
-	client := pb.NewBackendClient(conn)
-	return client.Detect(ctx, in, opts...)
-}

pkg/grpc/embed.go

@@ -59,10 +59,6 @@ func (e *embedBackend) SoundGeneration(ctx context.Context, in *pb.SoundGenerati
 	return e.s.SoundGeneration(ctx, in)
 }
 
-func (e *embedBackend) Detect(ctx context.Context, in *pb.DetectOptions, opts ...grpc.CallOption) (*pb.DetectResponse, error) {
-	return e.s.Detect(ctx, in)
-}
-
 func (e *embedBackend) AudioTranscription(ctx context.Context, in *pb.TranscriptRequest, opts ...grpc.CallOption) (*pb.TranscriptResult, error) {
 	return e.s.AudioTranscription(ctx, in)
 }

pkg/grpc/interface.go

@@ -4,7 +4,7 @@ import (
 	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
 )
 
-type AIModel interface {
+type LLM interface {
 	Busy() bool
 	Lock()
 	Unlock()
@@ -15,7 +15,6 @@ type AIModel interface {
 	Embeddings(*pb.PredictOptions) ([]float32, error)
 	GenerateImage(*pb.GenerateImageRequest) error
 	GenerateVideo(*pb.GenerateVideoRequest) error
-	Detect(*pb.DetectOptions) (pb.DetectResponse, error)
 	AudioTranscription(*pb.TranscriptRequest) (pb.TranscriptResult, error)
 	TTS(*pb.TTSRequest) error
 	SoundGeneration(*pb.SoundGenerationRequest) error

pkg/grpc/server.go

@@ -22,7 +22,7 @@ import (
 // server is used to implement helloworld.GreeterServer.
 type server struct {
 	pb.UnimplementedBackendServer
-	llm AIModel
+	llm LLM
 }
 
 func (s *server) Health(ctx context.Context, in *pb.HealthMessage) (*pb.Reply, error) {
@@ -111,18 +111,6 @@ func (s *server) SoundGeneration(ctx context.Context, in *pb.SoundGenerationRequ
 	return &pb.Result{Message: "Sound Generation audio generated", Success: true}, nil
 }
 
-func (s *server) Detect(ctx context.Context, in *pb.DetectOptions) (*pb.DetectResponse, error) {
-	if s.llm.Locking() {
-		s.llm.Lock()
-		defer s.llm.Unlock()
-	}
-	res, err := s.llm.Detect(in)
-	if err != nil {
-		return nil, err
-	}
-	return &res, nil
-}
-
 func (s *server) AudioTranscription(ctx context.Context, in *pb.TranscriptRequest) (*pb.TranscriptResult, error) {
 	if s.llm.Locking() {
 		s.llm.Lock()
@@ -263,7 +251,7 @@ func (s *server) VAD(ctx context.Context, in *pb.VADRequest) (*pb.VADResponse, e
 	return &res, nil
 }
 
-func StartServer(address string, model AIModel) error {
+func StartServer(address string, model LLM) error {
 	lis, err := net.Listen("tcp", address)
 	if err != nil {
 		return err
@@ -281,7 +269,7 @@ func StartServer(address string, model AIModel) error {
 	return nil
 }
 
-func RunServer(address string, model AIModel) (func() error, error) {
+func RunServer(address string, model LLM) (func() error, error) {
 	lis, err := net.Listen("tcp", address)
 	if err != nil {
 		return nil, err

pkg/startup/backend_preload.go

@@ -8,8 +8,8 @@ import (
 
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/gallery"
+	"github.com/mudler/LocalAI/core/system"
 	"github.com/mudler/LocalAI/pkg/downloader"
-	"github.com/mudler/LocalAI/pkg/system"
 	"github.com/rs/zerolog/log"
 )
 

pkg/startup/model_preload.go

@@ -10,8 +10,8 @@ import (
 
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/gallery"
+	"github.com/mudler/LocalAI/core/system"
 	"github.com/mudler/LocalAI/pkg/downloader"
-	"github.com/mudler/LocalAI/pkg/system"
 	"github.com/mudler/LocalAI/pkg/utils"
 	"github.com/rs/zerolog/log"
 	"gopkg.in/yaml.v2"

pkg/startup/model_preload_test.go

@@ -6,7 +6,7 @@ import (
 	"path/filepath"
 
 	"github.com/mudler/LocalAI/core/config"
-	. "github.com/mudler/LocalAI/core/startup"
+	. "github.com/mudler/LocalAI/pkg/startup"
 
 	. "github.com/onsi/ginkgo/v2"
 	. "github.com/onsi/gomega"

pkg/templates/evaluator_test.go

@@ -3,8 +3,8 @@ package templates_test
 import (
 	"github.com/mudler/LocalAI/core/config"
 	"github.com/mudler/LocalAI/core/schema"
-	. "github.com/mudler/LocalAI/core/templates"
 	"github.com/mudler/LocalAI/pkg/functions"
+	. "github.com/mudler/LocalAI/pkg/templates"
 
 	. "github.com/onsi/ginkgo/v2"
 	. "github.com/onsi/gomega"

pkg/templates/multimodal_test.go

@@ -1,7 +1,7 @@
 package templates_test
 
 import (
-	. "github.com/mudler/LocalAI/core/templates" // Update with your module path
+	. "github.com/mudler/LocalAI/pkg/templates" // Update with your module path
 
 	// Update with your module path
 	. "github.com/onsi/ginkgo/v2"