ci(bump_deps): attempt to link also commit diff

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

.github/bump_deps.sh

@@ -6,4 +6,17 @@ VAR=$3
 
 LAST_COMMIT=$(curl -s -H "Accept: application/vnd.github.VERSION.sha" "https://api.github.com/repos/$REPO/commits/$BRANCH")
 
+# Read $VAR from Makefile (only first match)
+set +e
+CURRENT_COMMIT="$(grep -m1 "^$VAR?=" Makefile | cut -d'=' -f2)"
+set -e
+
 sed -i Makefile -e "s/$VAR?=.*/$VAR?=$LAST_COMMIT/"
+
+if [ -z "$CURRENT_COMMIT" ]; then
+    echo "Could not find $VAR in Makefile."
+    exit 0
+fi
+
+echo "Updated $VAR from $CURRENT_COMMIT to $LAST_COMMIT." > "$REPO_message.txt"
+echo "https://github.com/$REPO/compare/$CURRENT_COMMIT..$LAST_COMMIT" >> "$REPO_message.txt"

.github/workflows/bump_deps.yaml

@@ -40,8 +40,14 @@ jobs:
     steps:
       - uses: actions/checkout@v4
       - name: Bump dependencies 🔧
+        id: bump
         run: |
           bash .github/bump_deps.sh ${{ matrix.repository }} ${{ matrix.branch }} ${{ matrix.variable }}
+          {
+            echo 'message<<EOF'
+            cat "${{ matrix.repository }}_message.txt"
+            echo EOF
+          } >> "$GITHUB_OUTPUT"
       - name: Create Pull Request
         uses: peter-evans/create-pull-request@v6
         with:
@@ -50,7 +56,7 @@ jobs:
           commit-message: ':arrow_up: Update ${{ matrix.repository }}'
           title: 'chore: :arrow_up: Update ${{ matrix.repository }}'
           branch: "update/${{ matrix.variable }}"
-          body: Bump of ${{ matrix.repository }} version
+          body:  ${{ steps.bump.outputs.message }}
           signoff: true
 
 

.github/workflows/checksum_checker.yaml

@@ -41,7 +41,7 @@ jobs:
           token: ${{ secrets.UPDATE_BOT_TOKEN }}
           push-to-fork: ci-forks/LocalAI
           commit-message: ':arrow_up: Checksum updates in gallery/index.yaml'
-          title: 'models(gallery): :arrow_up: update checksum'
+          title: 'chore(model-gallery): :arrow_up: update checksum'
           branch: "update/checksum"
           body: Updating checksums in gallery/index.yaml
           signoff: true

.github/workflows/disabled/comment-pr.yaml

@@ -8,8 +8,10 @@ jobs:
         MODEL_NAME: hermes-2-theta-llama-3-8b
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v4
+    - name: Checkout code
+      uses: actions/checkout@v3
       with:
+        ref: "${{ github.event.pull_request.merge_commit_sha }}"
         fetch-depth: 0 # needed to checkout all branches for this Action to work
     - uses: mudler/localai-github-action@v1
       with:
@@ -21,6 +23,7 @@ jobs:
             json_diff_file_output: diff.json
             raw_diff_file_output: diff.txt
             file_output_only: "true"
+            base_branch: ${{ github.event.pull_request.base.sha }}
     - name: Show diff
       env:
         DIFF: ${{ steps.git-diff-action.outputs.raw-diff-path }}

.github/workflows/image-pr.yml

@@ -47,7 +47,7 @@ jobs:
           #   makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "4"
+            cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12-ffmpeg'
@@ -120,7 +120,7 @@ jobs:
           #   makeflags: "--jobs=3 --output-sync=target"
           # - build-type: 'cublas'
           #   cuda-major-version: "12"
-          #   cuda-minor-version: "4"
+          #   cuda-minor-version: "0"
           #   platforms: 'linux/amd64'
           #   tag-latest: 'false'
           #   tag-suffix: '-cublas-cuda12-ffmpeg-core'

.github/workflows/image.yml

@@ -75,7 +75,7 @@ jobs:
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "4"
+            cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12'
@@ -100,7 +100,7 @@ jobs:
             makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "4"
+            cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'auto'
             tag-suffix: '-cublas-cuda12-ffmpeg'
@@ -285,7 +285,7 @@ jobs:
             makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "4"
+            cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12-core'
@@ -307,7 +307,7 @@ jobs:
             makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
-            cuda-minor-version: "4"
+            cuda-minor-version: "0"
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-cublas-cuda12-ffmpeg-core'

.github/workflows/release.yaml

@@ -4,6 +4,8 @@ on:
   push:
     branches:
       - master
+    tags:
+      - 'v*'
   pull_request:
 
 env:
@@ -29,11 +31,10 @@ jobs:
         with:
           go-version: '1.21.x'
           cache: false
-
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential ffmpeg protobuf-compiler ccache gawk
+          sudo apt-get install build-essential ffmpeg protobuf-compiler ccache upx-ucl gawk
           sudo apt-get install -qy binutils-aarch64-linux-gnu gcc-aarch64-linux-gnu g++-aarch64-linux-gnu libgmock-dev
       - name: Install CUDA Dependencies
         run: |
@@ -149,7 +150,7 @@ jobs:
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install -y wget curl build-essential ffmpeg protobuf-compiler ccache gawk cmake libgmock-dev
+          sudo apt-get install -y wget curl build-essential ffmpeg protobuf-compiler ccache upx-ucl gawk cmake libgmock-dev
       - name: Intel Dependencies
         run: |
           wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB | gpg --dearmor | sudo tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null
@@ -250,7 +251,7 @@ jobs:
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install -y --no-install-recommends libopencv-dev protobuf-compiler ccache
+          sudo apt-get install -y --no-install-recommends libopencv-dev protobuf-compiler ccache upx-ucl
           go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
           go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.2
       - name: Build stablediffusion

.github/workflows/test.yml

@@ -70,7 +70,7 @@ jobs:
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential curl ffmpeg
+          sudo apt-get install build-essential ccache upx-ucl curl ffmpeg
           sudo apt-get install -y libgmock-dev
           curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
              sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \

Dockerfile

@@ -24,7 +24,7 @@ RUN apt-get update && \
         cmake \
         curl \
         git \
-        unzip && \
+        unzip upx-ucl && \
     apt-get clean && \
     rm -rf /var/lib/apt/lists/*
 
@@ -99,7 +99,7 @@ FROM requirements-${IMAGE_TYPE} AS requirements-drivers
 
 ARG BUILD_TYPE
 ARG CUDA_MAJOR_VERSION=12
-ARG CUDA_MINOR_VERSION=4
+ARG CUDA_MINOR_VERSION=0
 
 ENV BUILD_TYPE=${BUILD_TYPE}
 

Makefile

@@ -8,7 +8,7 @@ DETECT_LIBS?=true
 # llama.cpp versions
 GOLLAMA_REPO?=https://github.com/go-skynet/go-llama.cpp
 GOLLAMA_VERSION?=2b57a8ae43e4699d3dc5d1496a1ccd42922993be
-CPPLLAMA_VERSION?=b3283448ce9a5098226afe1d8648ccc578511fe4
+CPPLLAMA_VERSION?=0d6fb52be0c1b7e77eb855f3adc4952771c8ce4c
 
 # gpt4all version
 GPT4ALL_REPO?=https://github.com/nomic-ai/gpt4all
@@ -20,7 +20,7 @@ RWKV_VERSION?=661e7ae26d442f5cfebd2a0881b44e8c55949ec6
 
 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggerganov/whisper.cpp
-WHISPER_CPP_VERSION?=f68298ce06ca3edd6e6f3f21c3d0bb5f073942c3
+WHISPER_CPP_VERSION?=fe36c909715e6751277ddb020e7892c7670b61d4
 
 # bert.cpp version
 BERT_REPO?=https://github.com/go-skynet/go-bert.cpp
@@ -58,7 +58,7 @@ RANDOM := $(shell bash -c 'echo $$RANDOM')
 
 VERSION?=$(shell git describe --always --tags || echo "dev" )
 # go tool nm ./local-ai | grep Commit
-LD_FLAGS?=
+LD_FLAGS?=-s -w
 override LD_FLAGS += -X "github.com/mudler/LocalAI/internal.Version=$(VERSION)"
 override LD_FLAGS += -X "github.com/mudler/LocalAI/internal.Commit=$(shell git rev-parse HEAD)"
 
@@ -72,6 +72,14 @@ WHITE  := $(shell tput -Txterm setaf 7)
 CYAN   := $(shell tput -Txterm setaf 6)
 RESET  := $(shell tput -Txterm sgr0)
 
+UPX?=
+# check if upx exists
+ifeq (, $(shell which upx))
+	UPX=
+else
+	UPX=$(shell which upx)
+endif
+
 # Default Docker bridge IP
 E2E_BRIDGE_IP?=172.17.0.1
 
@@ -377,6 +385,7 @@ build: prepare backend-assets grpcs ## Build the project
 	$(info ${GREEN}I BUILD_TYPE: ${YELLOW}$(BUILD_TYPE)${RESET})
 	$(info ${GREEN}I GO_TAGS: ${YELLOW}$(GO_TAGS)${RESET})
 	$(info ${GREEN}I LD_FLAGS: ${YELLOW}$(LD_FLAGS)${RESET})
+	$(info ${GREEN}I UPX: ${YELLOW}$(UPX)${RESET})
 ifneq ($(BACKEND_LIBS),)
 	$(MAKE) backend-assets/lib
 	cp -f $(BACKEND_LIBS) backend-assets/lib/
@@ -421,7 +430,7 @@ else
 endif
 
 dist-cross-linux-arm64:
-	CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_NATIVE=off" GRPC_BACKENDS="backend-assets/grpc/llama-cpp-fallback backend-assets/grpc/llama-cpp-grpc backend-assets/util/llama-cpp-rpc-server" \
+	CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_NATIVE=off" GRPC_BACKENDS="backend-assets/grpc/llama-cpp-fallback backend-assets/grpc/llama-cpp-grpc backend-assets/util/llama-cpp-rpc-server" GO_TAGS="p2p" \
 	STATIC=true $(MAKE) build
 	mkdir -p release
 # if BUILD_ID is empty, then we don't append it to the binary name
@@ -471,7 +480,7 @@ prepare-e2e:
 	mkdir -p $(TEST_DIR)
 	cp -rfv $(abspath ./tests/e2e-fixtures)/gpu.yaml $(TEST_DIR)/gpu.yaml
 	test -e $(TEST_DIR)/ggllm-test-model.bin || wget -q https://huggingface.co/TheBloke/CodeLlama-7B-Instruct-GGUF/resolve/main/codellama-7b-instruct.Q2_K.gguf -O $(TEST_DIR)/ggllm-test-model.bin
-	docker build --build-arg GRPC_BACKENDS="$(GRPC_BACKENDS)" --build-arg IMAGE_TYPE=core --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg CUDA_MAJOR_VERSION=12 --build-arg CUDA_MINOR_VERSION=4 --build-arg FFMPEG=true -t localai-tests .
+	docker build --build-arg GRPC_BACKENDS="$(GRPC_BACKENDS)" --build-arg IMAGE_TYPE=core --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg CUDA_MAJOR_VERSION=12 --build-arg CUDA_MINOR_VERSION=0 --build-arg FFMPEG=true -t localai-tests .
 
 run-e2e-image:
 	ls -liah $(abspath ./tests/e2e-fixtures)
@@ -733,13 +742,22 @@ backend-assets/grpc: protogen-go replace
 backend-assets/grpc/bert-embeddings: sources/go-bert.cpp sources/go-bert.cpp/libgobert.a backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-bert.cpp LIBRARY_PATH=$(CURDIR)/sources/go-bert.cpp \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/bert-embeddings ./backend/go/llm/bert/
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/bert-embeddings
+endif
 
 backend-assets/grpc/gpt4all: sources/gpt4all sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a backend-assets/gpt4all backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang/ LIBRARY_PATH=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang/ \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gpt4all ./backend/go/llm/gpt4all/
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/gpt4all
+endif
 
 backend-assets/grpc/huggingface: backend-assets/grpc
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/huggingface ./backend/go/llm/langchain/
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/huggingface
+endif
 
 backend/cpp/llama/llama.cpp:
 	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama llama.cpp
@@ -841,29 +859,50 @@ backend-assets/util/llama-cpp-rpc-server: backend-assets/grpc/llama-cpp-grpc
 backend-assets/grpc/llama-ggml: sources/go-llama.cpp sources/go-llama.cpp/libbinding.a backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-llama.cpp LIBRARY_PATH=$(CURDIR)/sources/go-llama.cpp \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama-ggml ./backend/go/llm/llama-ggml/
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/llama-ggml
+endif
 
 backend-assets/grpc/piper: sources/go-piper sources/go-piper/libpiper_binding.a backend-assets/grpc backend-assets/espeak-ng-data
 	CGO_CXXFLAGS="$(PIPER_CGO_CXXFLAGS)" CGO_LDFLAGS="$(PIPER_CGO_LDFLAGS)" LIBRARY_PATH=$(CURDIR)/sources/go-piper \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/piper ./backend/go/tts/
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/piper
+endif
 
 backend-assets/grpc/rwkv: sources/go-rwkv.cpp sources/go-rwkv.cpp/librwkv.a backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-rwkv.cpp LIBRARY_PATH=$(CURDIR)/sources/go-rwkv.cpp \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/rwkv ./backend/go/llm/rwkv
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/rwkv
+endif
 
 backend-assets/grpc/stablediffusion: sources/go-stable-diffusion sources/go-stable-diffusion/libstablediffusion.a backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" CPATH="$(CPATH):$(CURDIR)/sources/go-stable-diffusion/:/usr/include/opencv4" LIBRARY_PATH=$(CURDIR)/sources/go-stable-diffusion/ \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/stablediffusion ./backend/go/image/stablediffusion
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/stablediffusion
+endif
 
 backend-assets/grpc/tinydream: sources/go-tiny-dream sources/go-tiny-dream/libtinydream.a backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" LIBRARY_PATH=$(CURDIR)/go-tiny-dream \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/tinydream ./backend/go/image/tinydream
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/tinydream
+endif
 
 backend-assets/grpc/whisper: sources/whisper.cpp sources/whisper.cpp/libwhisper.a backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS) $(CGO_LDFLAGS_WHISPER)" C_INCLUDE_PATH="$(CURDIR)/sources/whisper.cpp/include:$(CURDIR)/sources/whisper.cpp/ggml/include" LIBRARY_PATH=$(CURDIR)/sources/whisper.cpp \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/whisper ./backend/go/transcribe/
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/whisper
+endif
 
 backend-assets/grpc/local-store: backend-assets/grpc
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/local-store ./backend/go/stores/
+ifneq ($(UPX),)
+	$(UPX) backend-assets/grpc/local-store
+endif
 
 grpcs: prepare $(GRPC_BACKENDS)
 

README.md

@@ -84,6 +84,7 @@ docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-aio-cpu
 
 Hot topics (looking for contributors):
 
+- 🔥🔥 Distributed, P2P Global community pools: https://github.com/mudler/LocalAI/issues/3113
 - WebUI improvements: https://github.com/mudler/LocalAI/issues/2156
 - Backends v2: https://github.com/mudler/LocalAI/issues/1126
 - Improving UX v2: https://github.com/mudler/LocalAI/issues/1373
@@ -150,6 +151,7 @@ Other:
 
 ## :book: 🎥 [Media, Blogs, Social](https://localai.io/basics/news/#media-blogs-social)
 
+- [Run Visual studio code with LocalAI (SUSE)](https://www.suse.com/c/running-ai-locally/)
 - 🆕 [Run LocalAI on Jetson Nano Devkit](https://mudler.pm/posts/local-ai-jetson-nano-devkit/)
 - [Run LocalAI on AWS EKS with Pulumi](https://www.pulumi.com/blog/low-code-llm-apps-with-local-ai-flowise-and-pulumi/)
 - [Run LocalAI on AWS](https://staleks.hashnode.dev/installing-localai-on-aws-ec2-instance)

backend/cpp/llama/grpc-server.cpp

@@ -2259,7 +2259,6 @@ static void params_parse(const backend::ModelOptions* request,
      // get the directory of modelfile
      std::string model_dir = params.model.substr(0, params.model.find_last_of("/\\"));
      params.lora_adapter.push_back(std::make_tuple(model_dir + "/"+request->loraadapter(), scale_factor));
-     params.lora_base  =  model_dir + "/"+request->lorabase();
     }
     params.use_mlock = request->mlock();
     params.use_mmap = request->mmap();

backend/python/autogptq/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/autogptq/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/autogptq/requirements-intel.txt

@@ -2,4 +2,4 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406

backend/python/autogptq/requirements.txt

@@ -1,7 +1,6 @@
 accelerate
 auto-gptq==0.7.1
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
-torch
 certifi
 transformers

backend/python/bark/requirements-cublas11.txt

@@ -0,0 +1,3 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch
+torchaudio

backend/python/bark/requirements-cublas12.txt

@@ -0,0 +1,2 @@
+torch
+torchaudio

backend/python/bark/requirements.txt

@@ -1,6 +1,6 @@
 accelerate
 bark==0.1.5
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
 certifi
 transformers

backend/python/common/libbackend.sh

@@ -122,6 +122,13 @@ function installRequirements() {
         requirementFiles+=("${MY_DIR}/requirements-${BUILD_PROFILE}.txt")
     fi
 
+    # if BUILD_TYPE is empty, we are a CPU build, so we should try to install the CPU requirements
+    if [ "x${BUILD_TYPE}" == "x" ]; then
+        requirementFiles+=("${MY_DIR}/requirements-cpu.txt")
+    fi
+
+    requirementFiles+=("${MY_DIR}/requirements-after.txt")
+
     for reqFile in ${requirementFiles[@]}; do
         if [ -f ${reqFile} ]; then
             echo "starting requirements install for ${reqFile}"

backend/python/common/template/requirements.txt

@@ -1,2 +1,2 @@
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf

backend/python/coqui/requirements-cublas11.txt

@@ -0,0 +1,3 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch
+torchaudio

backend/python/coqui/requirements-cublas12.txt

@@ -0,0 +1,2 @@
+torch
+torchaudio

backend/python/coqui/requirements-intel.txt

@@ -3,4 +3,4 @@ intel-extension-for-pytorch
 torch
 torchaudio
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406

backend/python/coqui/requirements.txt

@@ -1,6 +1,6 @@
 accelerate
 TTS==0.22.0
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
 certifi
 transformers

backend/python/diffusers/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/diffusers/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/diffusers/requirements.txt

@@ -3,11 +3,10 @@ accelerate
 compel
 peft
 diffusers
-grpcio==1.65.0
+grpcio==1.65.4
 opencv-python
 pillow
 protobuf
 sentencepiece
-torch
 transformers
 certifi

backend/python/exllama/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/exllama/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/exllama/requirements.txt

@@ -1,6 +1,5 @@
 grpcio==1.65.0
 protobuf
-torch
 transformers
 certifi
 setuptools

backend/python/exllama2/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/exllama2/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/exllama2/requirements.txt

@@ -1,7 +1,6 @@
 accelerate
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
 certifi
-torch
 wheel
 setuptools

backend/python/mamba/requirements-after.txt

@@ -0,0 +1,2 @@
+causal-conv1d==1.4.0
+mamba-ssm==2.2.2

backend/python/mamba/requirements-cpu.txt

@@ -0,0 +1 @@
+torch

backend/python/mamba/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/mamba/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/mamba/requirements-install.txt

@@ -3,5 +3,4 @@
 # https://github.com/Dao-AILab/causal-conv1d/issues/24
 packaging
 setuptools
-wheel
-torch==2.3.1
+wheel

backend/python/mamba/requirements.txt

@@ -1,6 +1,4 @@
-causal-conv1d==1.4.0
-mamba-ssm==2.2.2
-grpcio==1.65.0
+grpcio==1.65.1
 protobuf
 certifi
 transformers

backend/python/openvoice/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/openvoice/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/openvoice/requirements-intel.txt

@@ -2,7 +2,7 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-grpcio==1.64.1
+grpcio==1.65.4
 protobuf
 librosa==0.9.1
 faster-whisper==1.0.3

backend/python/openvoice/requirements.txt

@@ -1,4 +1,4 @@
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
 librosa
 faster-whisper

backend/python/openvoice/test.sh

@@ -5,7 +5,7 @@ source $(dirname $0)/../common/libbackend.sh
 
 # Download checkpoints if not present
 if [ ! -d "checkpoints_v2" ]; then
-    wget https://myshell-public-repo-hosting.s3.amazonaws.com/openvoice/checkpoints_v2_0417.zip -O checkpoints_v2.zip
+    wget https://myshell-public-repo-host.s3.amazonaws.com/openvoice/checkpoints_v2_0417.zip -O checkpoints_v2.zip
     unzip checkpoints_v2.zip
 fi
 

backend/python/parler-tts/requirements-cublas11.txt

@@ -0,0 +1,3 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch
+torchaudio

backend/python/parler-tts/requirements-cublas12.txt

@@ -0,0 +1,2 @@
+torch
+torchaudio

backend/python/parler-tts/requirements-intel.txt

@@ -3,4 +3,4 @@ intel-extension-for-pytorch
 torch
 torchaudio
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406

backend/python/parler-tts/requirements.txt

@@ -1,7 +1,6 @@
 accelerate
-grpcio==1.65.0
+grpcio==1.65.1
 protobuf
-torch
 git+https://github.com/huggingface/parler-tts.git@10016fb0300c0dc31a0fb70e26f3affee7b62f16
 certifi
 transformers

backend/python/petals/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/petals/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/petals/requirements-intel.txt

@@ -2,4 +2,4 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406

backend/python/rerankers/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/rerankers/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/rerankers/requirements-intel.txt

@@ -2,4 +2,4 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406

backend/python/rerankers/requirements.txt

@@ -1,6 +1,6 @@
 accelerate
 rerankers[transformers]
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
 certifi
 transformers

backend/python/sentencetransformers/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/sentencetransformers/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/sentencetransformers/requirements-intel.txt

@@ -2,4 +2,4 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==69.5.1 # https://github.com/mudler/LocalAI/issues/2406

backend/python/sentencetransformers/requirements.txt

@@ -1,6 +1,6 @@
 accelerate
 sentence-transformers==3.0.1
 transformers
-grpcio==1.65.0
+grpcio==1.65.1
 protobuf
 certifi

backend/python/transformers-musicgen/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/transformers-musicgen/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/transformers-musicgen/requirements-intel.txt

@@ -2,4 +2,4 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==69.5.1 # https://github.com/mudler/LocalAI/issues/2406

backend/python/transformers-musicgen/requirements.txt

@@ -1,7 +1,6 @@
 accelerate
 transformers
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
-torch
 scipy==1.14.0
 certifi

backend/python/transformers/requirements-cublas11.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch

backend/python/transformers/requirements-cublas12.txt

@@ -0,0 +1 @@
+torch

backend/python/transformers/requirements-intel.txt

@@ -2,4 +2,3 @@
 intel-extension-for-pytorch
 torch
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406

backend/python/transformers/requirements.txt

@@ -1,9 +1,8 @@
 accelerate
 transformers
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
-torch
 certifi
 intel-extension-for-transformers
 bitsandbytes
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==69.5.1 # https://github.com/mudler/LocalAI/issues/2406

backend/python/vall-e-x/requirements-cublas11.txt

@@ -0,0 +1,3 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch
+torchaudio

backend/python/vall-e-x/requirements-cublas12.txt

@@ -0,0 +1,2 @@
+torch
+torchaudio

backend/python/vall-e-x/requirements-intel.txt

@@ -3,4 +3,4 @@ intel-extension-for-pytorch
 torch
 torchaudio
 optimum[openvino]
-setuptools==70.3.0 # https://github.com/mudler/LocalAI/issues/2406
+setuptools==72.1.0 # https://github.com/mudler/LocalAI/issues/2406

backend/python/vall-e-x/requirements.txt

@@ -1,4 +1,4 @@
 accelerate
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
 certifi

backend/python/vllm/requirements-cublas.txt

@@ -1 +0,0 @@
-flash-attn

backend/python/vllm/requirements-cublas11.txt

@@ -0,0 +1,3 @@
+--extra-index-url https://download.pytorch.org/whl/cu118
+torch
+flash-attn

backend/python/vllm/requirements-cublas12.txt

@@ -0,0 +1,2 @@
+torch
+flash-attn

backend/python/vllm/requirements.txt

@@ -1,6 +1,6 @@
 accelerate
 vllm
-grpcio==1.65.0
+grpcio==1.65.4
 protobuf
 certifi
 transformers

core/cli/federated.go

@@ -10,10 +10,12 @@ import (
 type FederatedCLI struct {
 	Address        string `env:"LOCALAI_ADDRESS,ADDRESS" default:":8080" help:"Bind address for the API server" group:"api"`
 	Peer2PeerToken string `env:"LOCALAI_P2P_TOKEN,P2P_TOKEN,TOKEN" name:"p2ptoken" help:"Token for P2P mode (optional)" group:"p2p"`
+	LoadBalanced   bool   `env:"LOCALAI_LOAD_BALANCED,LOAD_BALANCED" default:"false" help:"Enable load balancing" group:"p2p"`
 }
 
 func (f *FederatedCLI) Run(ctx *cliContext.Context) error {
-	fs := p2p.NewFederatedServer(f.Address, p2p.FederatedID, f.Peer2PeerToken)
+
+	fs := p2p.NewFederatedServer(f.Address, p2p.FederatedID, f.Peer2PeerToken, f.LoadBalanced)
 
 	return fs.Start(context.Background())
 }

core/cli/models.go

@@ -83,7 +83,9 @@ func (mi *ModelsInstall) Run(ctx *cliContext.Context) error {
 			return err
 		}
 
-		if !downloader.LooksLikeOCI(modelName) {
+		modelURI := downloader.URI(modelName)
+
+		if !modelURI.LooksLikeOCI() {
 			model := gallery.FindModel(models, modelName, mi.ModelsPath)
 			if model == nil {
 				log.Error().Str("model", modelName).Msg("model not found")

core/cli/run.go

@@ -119,7 +119,7 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 		}
 
 		log.Info().Msg("Starting P2P server discovery...")
-		if err := p2p.ServiceDiscoverer(context.Background(), node, token, "", func() {
+		if err := p2p.ServiceDiscoverer(context.Background(), node, token, "", func(serviceID string, node p2p.NodeData) {
 			var tunnelAddresses []string
 			for _, v := range p2p.GetAvailableNodes("") {
 				if v.IsOnline() {

core/cli/util.go

@@ -86,8 +86,8 @@ func (hfscmd *HFScanCMD) Run(ctx *cliContext.Context) error {
 		var errs error = nil
 		for _, uri := range hfscmd.ToScan {
 			log.Info().Str("uri", uri).Msg("scanning specific uri")
-			scanResults, err := downloader.HuggingFaceScan(uri)
-			if err != nil && !errors.Is(err, downloader.ErrNonHuggingFaceFile) {
+			scanResults, err := downloader.HuggingFaceScan(downloader.URI(uri))
+			if err != nil && errors.Is(err, downloader.ErrUnsafeFilesFound) {
 				log.Error().Err(err).Strs("clamAV", scanResults.ClamAVInfectedFiles).Strs("pickles", scanResults.DangerousPickles).Msg("! WARNING ! A known-vulnerable model is included in this repo!")
 				errs = errors.Join(errs, err)
 			}

core/config/backend_config.go

@@ -8,7 +8,6 @@ import (
 	"github.com/mudler/LocalAI/core/schema"
 	"github.com/mudler/LocalAI/pkg/downloader"
 	"github.com/mudler/LocalAI/pkg/functions"
-	"github.com/mudler/LocalAI/pkg/utils"
 )
 
 const (
@@ -72,9 +71,9 @@ type BackendConfig struct {
 }
 
 type File struct {
-	Filename string `yaml:"filename" json:"filename"`
-	SHA256   string `yaml:"sha256" json:"sha256"`
-	URI      string `yaml:"uri" json:"uri"`
+	Filename string         `yaml:"filename" json:"filename"`
+	SHA256   string         `yaml:"sha256" json:"sha256"`
+	URI      downloader.URI `yaml:"uri" json:"uri"`
 }
 
 type VallE struct {
@@ -213,28 +212,32 @@ func (c *BackendConfig) ShouldCallSpecificFunction() bool {
 // MMProjFileName returns the filename of the MMProj file
 // If the MMProj is a URL, it will return the MD5 of the URL which is the filename
 func (c *BackendConfig) MMProjFileName() string {
-	modelURL := downloader.ConvertURL(c.MMProj)
-	if downloader.LooksLikeURL(modelURL) {
-		return utils.MD5(modelURL)
+	uri := downloader.URI(c.MMProj)
+	if uri.LooksLikeURL() {
+		f, _ := uri.FilenameFromUrl()
+		return f
 	}
 
 	return c.MMProj
 }
 
 func (c *BackendConfig) IsMMProjURL() bool {
-	return downloader.LooksLikeURL(downloader.ConvertURL(c.MMProj))
+	uri := downloader.URI(c.MMProj)
+	return uri.LooksLikeURL()
 }
 
 func (c *BackendConfig) IsModelURL() bool {
-	return downloader.LooksLikeURL(downloader.ConvertURL(c.Model))
+	uri := downloader.URI(c.Model)
+	return uri.LooksLikeURL()
 }
 
 // ModelFileName returns the filename of the model
 // If the model is a URL, it will return the MD5 of the URL which is the filename
 func (c *BackendConfig) ModelFileName() string {
-	modelURL := downloader.ConvertURL(c.Model)
-	if downloader.LooksLikeURL(modelURL) {
-		return utils.MD5(modelURL)
+	uri := downloader.URI(c.Model)
+	if uri.LooksLikeURL() {
+		f, _ := uri.FilenameFromUrl()
+		return f
 	}
 
 	return c.Model

core/config/backend_config_loader.go

@@ -244,7 +244,7 @@ func (bcl *BackendConfigLoader) Preload(modelPath string) error {
 			// Create file path
 			filePath := filepath.Join(modelPath, file.Filename)
 
-			if err := downloader.DownloadFile(file.URI, filePath, file.SHA256, i, len(config.DownloadFiles), status); err != nil {
+			if err := file.URI.DownloadFile(filePath, file.SHA256, i, len(config.DownloadFiles), status); err != nil {
 				return err
 			}
 		}
@@ -252,10 +252,10 @@ func (bcl *BackendConfigLoader) Preload(modelPath string) error {
 		// If the model is an URL, expand it, and download the file
 		if config.IsModelURL() {
 			modelFileName := config.ModelFileName()
-			modelURL := downloader.ConvertURL(config.Model)
+			uri := downloader.URI(config.Model)
 			// check if file exists
 			if _, err := os.Stat(filepath.Join(modelPath, modelFileName)); errors.Is(err, os.ErrNotExist) {
-				err := downloader.DownloadFile(modelURL, filepath.Join(modelPath, modelFileName), "", 0, 0, status)
+				err := uri.DownloadFile(filepath.Join(modelPath, modelFileName), "", 0, 0, status)
 				if err != nil {
 					return err
 				}
@@ -269,10 +269,10 @@ func (bcl *BackendConfigLoader) Preload(modelPath string) error {
 
 		if config.IsMMProjURL() {
 			modelFileName := config.MMProjFileName()
-			modelURL := downloader.ConvertURL(config.MMProj)
+			uri := downloader.URI(config.MMProj)
 			// check if file exists
 			if _, err := os.Stat(filepath.Join(modelPath, modelFileName)); errors.Is(err, os.ErrNotExist) {
-				err := downloader.DownloadFile(modelURL, filepath.Join(modelPath, modelFileName), "", 0, 0, status)
+				err := uri.DownloadFile(filepath.Join(modelPath, modelFileName), "", 0, 0, status)
 				if err != nil {
 					return err
 				}

core/config/guesser.go

@@ -26,15 +26,17 @@ const (
 type settingsConfig struct {
 	StopWords      []string
 	TemplateConfig TemplateConfig
+	RepeatPenalty float64
 }
 
 // default settings to adopt with a given model family
 var defaultsSettings map[familyType]settingsConfig = map[familyType]settingsConfig{
 	Gemma: {
+		RepeatPenalty: 1.0,
 		StopWords: []string{"<|im_end|>", "<end_of_turn>", "<start_of_turn>"},
 		TemplateConfig: TemplateConfig{
-			Chat:        "{{.Input }}\n<|start_of_turn|>model\n",
-			ChatMessage: "<|start_of_turn|>{{if eq .RoleName \"assistant\" }}model{{else}}{{ .RoleName }}{{end}}\n{{ if .Content -}}\n{{.Content -}}\n{{ end -}}<|end_of_turn|>",
+			Chat:        "{{.Input }}\n<start_of_turn>model\n",
+			ChatMessage: "<start_of_turn>{{if eq .RoleName \"assistant\" }}model{{else}}{{ .RoleName }}{{end}}\n{{ if .Content -}}\n{{.Content -}}\n{{ end -}}<end_of_turn>",
 			Completion:  "{{.Input}}",
 		},
 	},
@@ -192,6 +194,9 @@ func guessDefaultsFromFile(cfg *BackendConfig, modelPath string) {
 		if len(cfg.StopWords) == 0 {
 			cfg.StopWords = settings.StopWords
 		}
+		if cfg.RepeatPenalty == 0.0 {
+			cfg.RepeatPenalty = settings.RepeatPenalty
+		}
 	} else {
 		log.Debug().Any("family", family).Msgf("guessDefaultsFromFile: no template found for family")
 	}
@@ -219,7 +224,7 @@ func identifyFamily(f *gguf.GGUFFile) familyType {
 	commandR := arch == "command-r" && eosTokenID == 255001
 	qwen2 := arch == "qwen2"
 	phi3 := arch == "phi-3"
-	gemma := strings.HasPrefix(f.Model().Name, "gemma")
+	gemma := strings.HasPrefix(arch, "gemma") || strings.Contains(strings.ToLower(f.Model().Name), "gemma")
 	deepseek2 := arch == "deepseek2"
 
 	switch {

core/dependencies_manager/manager.go

@@ -37,7 +37,8 @@ func main() {
 
 	// download the assets
 	for _, asset := range assets {
-		if err := downloader.DownloadFile(asset.URL, filepath.Join(destPath, asset.FileName), asset.SHA, 1, 1, utils.DisplayDownloadFunction); err != nil {
+		uri := downloader.URI(asset.URL)
+		if err := uri.DownloadFile(filepath.Join(destPath, asset.FileName), asset.SHA, 1, 1, utils.DisplayDownloadFunction); err != nil {
 			panic(err)
 		}
 	}

core/gallery/gallery.go

@@ -131,7 +131,8 @@ func AvailableGalleryModels(galleries []config.Gallery, basePath string) ([]*Gal
 
 func findGalleryURLFromReferenceURL(url string, basePath string) (string, error) {
 	var refFile string
-	err := downloader.DownloadAndUnmarshal(url, basePath, func(url string, d []byte) error {
+	uri := downloader.URI(url)
+	err := uri.DownloadAndUnmarshal(basePath, func(url string, d []byte) error {
 		refFile = string(d)
 		if len(refFile) == 0 {
 			return fmt.Errorf("invalid reference file at url %s: %s", url, d)
@@ -153,8 +154,9 @@ func getGalleryModels(gallery config.Gallery, basePath string) ([]*GalleryModel,
 			return models, err
 		}
 	}
+	uri := downloader.URI(gallery.URL)
 
-	err := downloader.DownloadAndUnmarshal(gallery.URL, basePath, func(url string, d []byte) error {
+	err := uri.DownloadAndUnmarshal(basePath, func(url string, d []byte) error {
 		return yaml.Unmarshal(d, &models)
 	})
 	if err != nil {
@@ -204,35 +206,34 @@ func DeleteModelFromSystem(basePath string, name string, additionalFiles []strin
 		log.Error().Err(err).Msgf("failed to read gallery file %s", configFile)
 	}
 
+	var filesToRemove []string
+
 	// Remove additional files
 	if galleryconfig != nil {
 		for _, f := range galleryconfig.Files {
 			fullPath := filepath.Join(basePath, f.Filename)
-			log.Debug().Msgf("Removing file %s", fullPath)
-			if e := os.Remove(fullPath); e != nil {
-				err = errors.Join(err, fmt.Errorf("failed to remove file %s: %w", f.Filename, e))
-			}
+			filesToRemove = append(filesToRemove, fullPath)
 		}
 	}
 
 	for _, f := range additionalFiles {
 		fullPath := filepath.Join(filepath.Join(basePath, f))
-		log.Debug().Msgf("Removing additional file %s", fullPath)
-		if e := os.Remove(fullPath); e != nil {
+		filesToRemove = append(filesToRemove, fullPath)
+	}
+
+	filesToRemove = append(filesToRemove, configFile)
+	filesToRemove = append(filesToRemove, galleryFile)
+
+	// skip duplicates
+	filesToRemove = utils.Unique(filesToRemove)
+
+	// Removing files
+	for _, f := range filesToRemove {
+		if e := os.Remove(f); e != nil {
 			err = errors.Join(err, fmt.Errorf("failed to remove file %s: %w", f, e))
 		}
 	}
 
-	log.Debug().Msgf("Removing model config file %s", configFile)
-
-	// Delete the model config file
-	if e := os.Remove(configFile); e != nil {
-		err = errors.Join(err, fmt.Errorf("failed to remove file %s: %w", configFile, e))
-	}
-
-	// Delete gallery config file
-	os.Remove(galleryFile)
-
 	return err
 }
 
@@ -253,8 +254,8 @@ func SafetyScanGalleryModels(galleries []config.Gallery, basePath string) error
 
 func SafetyScanGalleryModel(galleryModel *GalleryModel) error {
 	for _, file := range galleryModel.AdditionalFiles {
-		scanResults, err := downloader.HuggingFaceScan(file.URI)
-		if err != nil && !errors.Is(err, downloader.ErrNonHuggingFaceFile) {
+		scanResults, err := downloader.HuggingFaceScan(downloader.URI(file.URI))
+		if err != nil && errors.Is(err, downloader.ErrUnsafeFilesFound) {
 			log.Error().Str("model", galleryModel.Name).Strs("clamAV", scanResults.ClamAVInfectedFiles).Strs("pickles", scanResults.DangerousPickles).Msg("Contains unsafe file(s)!")
 			return err
 		}

core/gallery/models.go

@@ -68,7 +68,8 @@ type PromptTemplate struct {
 
 func GetGalleryConfigFromURL(url string, basePath string) (Config, error) {
 	var config Config
-	err := downloader.DownloadAndUnmarshal(url, basePath, func(url string, d []byte) error {
+	uri := downloader.URI(url)
+	err := uri.DownloadAndUnmarshal(basePath, func(url string, d []byte) error {
 		return yaml.Unmarshal(d, &config)
 	})
 	if err != nil {
@@ -118,14 +119,14 @@ func InstallModel(basePath, nameOverride string, config *Config, configOverrides
 		filePath := filepath.Join(basePath, file.Filename)
 
 		if enforceScan {
-			scanResults, err := downloader.HuggingFaceScan(file.URI)
-			if err != nil && !errors.Is(err, downloader.ErrNonHuggingFaceFile) {
+			scanResults, err := downloader.HuggingFaceScan(downloader.URI(file.URI))
+			if err != nil && errors.Is(err, downloader.ErrUnsafeFilesFound) {
 				log.Error().Str("model", config.Name).Strs("clamAV", scanResults.ClamAVInfectedFiles).Strs("pickles", scanResults.DangerousPickles).Msg("Contains unsafe file(s)!")
 				return err
 			}
 		}
-
-		if err := downloader.DownloadFile(file.URI, filePath, file.SHA256, i, len(config.Files), downloadStatus); err != nil {
+		uri := downloader.URI(file.URI)
+		if err := uri.DownloadFile(filePath, file.SHA256, i, len(config.Files), downloadStatus); err != nil {
 			return err
 		}
 	}

core/http/app_test.go

@@ -73,8 +73,9 @@ func getModelStatus(url string) (response map[string]interface{}) {
 }
 
 func getModels(url string) (response []gallery.GalleryModel) {
+	uri := downloader.URI(url)
 	// TODO: No tests currently seem to exercise file:// urls. Fix?
-	downloader.DownloadAndUnmarshal(url, "", func(url string, i []byte) error {
+	uri.DownloadAndUnmarshal("", func(url string, i []byte) error {
 		// Unmarshal YAML data into a struct
 		return json.Unmarshal(i, &response)
 	})

core/http/elements/gallery.go

@@ -9,7 +9,6 @@ import (
 	"github.com/mudler/LocalAI/core/gallery"
 	"github.com/mudler/LocalAI/core/p2p"
 	"github.com/mudler/LocalAI/core/services"
-	"github.com/mudler/LocalAI/pkg/xsync"
 )
 
 const (
@@ -372,7 +371,12 @@ func dropBadChars(s string) string {
 	return strings.ReplaceAll(s, "@", "__")
 }
 
-func ListModels(models []*gallery.GalleryModel, processing *xsync.SyncedMap[string, string], galleryService *services.GalleryService) string {
+type ProcessTracker interface {
+	Exists(string) bool
+	Get(string) string
+}
+
+func ListModels(models []*gallery.GalleryModel, processTracker ProcessTracker, galleryService *services.GalleryService) string {
 	modelsElements := []elem.Node{}
 	descriptionDiv := func(m *gallery.GalleryModel) elem.Node {
 		return elem.Div(
@@ -396,7 +400,7 @@ func ListModels(models []*gallery.GalleryModel, processing *xsync.SyncedMap[stri
 
 	actionDiv := func(m *gallery.GalleryModel) elem.Node {
 		galleryID := fmt.Sprintf("%s@%s", m.Gallery.Name, m.Name)
-		currentlyProcessing := processing.Exists(galleryID)
+		currentlyProcessing := processTracker.Exists(galleryID)
 		jobID := ""
 		isDeletionOp := false
 		if currentlyProcessing {
@@ -404,7 +408,7 @@ func ListModels(models []*gallery.GalleryModel, processing *xsync.SyncedMap[stri
 			if status != nil && status.Deletion {
 				isDeletionOp = true
 			}
-			jobID = processing.Get(galleryID)
+			jobID = processTracker.Get(galleryID)
 			// TODO:
 			// case not handled, if status == nil : "Waiting"
 		}

core/http/endpoints/localai/welcome.go

@@ -17,7 +17,10 @@ func WelcomeEndpoint(appConfig *config.ApplicationConfig,
 		backendConfigs := cl.GetAllBackendConfigs()
 
 		galleryConfigs := map[string]*gallery.Config{}
+		modelsWithBackendConfig := map[string]interface{}{}
+
 		for _, m := range backendConfigs {
+			modelsWithBackendConfig[m.Name] = nil
 
 			cfg, err := gallery.GetLocalModelConfiguration(ml.ModelPath, m.Name)
 			if err != nil {
@@ -32,7 +35,7 @@ func WelcomeEndpoint(appConfig *config.ApplicationConfig,
 		modelsWithoutConfig := []string{}
 
 		for _, m := range models {
-			if _, ok := galleryConfigs[m]; !ok {
+			if _, ok := modelsWithBackendConfig[m]; !ok {
 				modelsWithoutConfig = append(modelsWithoutConfig, m)
 			}
 		}

core/http/endpoints/openai/chat.go

@@ -225,18 +225,16 @@ func ChatEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, startup
 			}
 
 			// Update input grammar
-			// Handle if we should return "name" instead of "functions"
-			if config.FunctionsConfig.FunctionName {
-				jsStruct := funcs.ToJSONNameStructure()
-				config.Grammar = jsStruct.Grammar(config.FunctionsConfig.GrammarConfig.Options()...)
-			} else {
-				jsStruct := funcs.ToJSONFunctionStructure()
-				config.Grammar = jsStruct.Grammar(config.FunctionsConfig.GrammarConfig.Options()...)
+			jsStruct := funcs.ToJSONStructure(config.FunctionsConfig.FunctionNameKey, config.FunctionsConfig.FunctionNameKey)
+			g, err := jsStruct.Grammar(config.FunctionsConfig.GrammarOptions()...)
+			if err == nil {
+				config.Grammar = g
 			}
 		case input.JSONFunctionGrammarObject != nil:
-			config.Grammar = input.JSONFunctionGrammarObject.Grammar(config.FunctionsConfig.GrammarConfig.Options()...)
-		case input.JSONFunctionGrammarObjectName != nil:
-			config.Grammar = input.JSONFunctionGrammarObjectName.Grammar(config.FunctionsConfig.GrammarConfig.Options()...)
+			g, err := input.JSONFunctionGrammarObject.Grammar(config.FunctionsConfig.GrammarOptions()...)
+			if err == nil {
+				config.Grammar = g
+			}
 		default:
 			// Force picking one of the functions by the request
 			if config.FunctionToCall() != "" {

core/http/routes/ui.go

@@ -21,6 +21,40 @@ import (
 	"github.com/google/uuid"
 )
 
+type modelOpCache struct {
+	status *xsync.SyncedMap[string, string]
+}
+
+func NewModelOpCache() *modelOpCache {
+	return &modelOpCache{
+		status: xsync.NewSyncedMap[string, string](),
+	}
+}
+
+func (m *modelOpCache) Set(key string, value string) {
+	m.status.Set(key, value)
+}
+
+func (m *modelOpCache) Get(key string) string {
+	return m.status.Get(key)
+}
+
+func (m *modelOpCache) DeleteUUID(uuid string) {
+	for _, k := range m.status.Keys() {
+		if m.status.Get(k) == uuid {
+			m.status.Delete(k)
+		}
+	}
+}
+
+func (m *modelOpCache) Map() map[string]string {
+	return m.status.Map()
+}
+
+func (m *modelOpCache) Exists(key string) bool {
+	return m.status.Exists(key)
+}
+
 func RegisterUIRoutes(app *fiber.App,
 	cl *config.BackendConfigLoader,
 	ml *model.ModelLoader,
@@ -29,7 +63,7 @@ func RegisterUIRoutes(app *fiber.App,
 	auth func(*fiber.Ctx) error) {
 
 	// keeps the state of models that are being installed from the UI
-	var processingModels = xsync.NewSyncedMap[string, string]()
+	var processingModels = NewModelOpCache()
 
 	// modelStatus returns the current status of the models being processed (installation or deletion)
 	// it is called asynchonously from the UI
@@ -232,6 +266,8 @@ func RegisterUIRoutes(app *fiber.App,
 			return c.SendString(elements.ProgressBar("100"))
 		}
 		if status.Error != nil {
+			// TODO: instead of deleting the job, we should keep it in the cache and make it dismissable
+			processingModels.DeleteUUID(jobUID)
 			return c.SendString(elements.ErrorProgress(status.Error.Error(), status.GalleryModelName))
 		}
 
@@ -246,12 +282,7 @@ func RegisterUIRoutes(app *fiber.App,
 		status := galleryService.GetStatus(jobUID)
 
 		galleryID := ""
-		for _, k := range processingModels.Keys() {
-			if processingModels.Get(k) == jobUID {
-				galleryID = k
-				processingModels.Delete(k)
-			}
-		}
+		processingModels.DeleteUUID(jobUID)
 		if galleryID == "" {
 			log.Debug().Msgf("no processing model found for job : %+v\n", jobUID)
 		}

core/http/views/p2p.html

@@ -16,7 +16,16 @@
                 </a> 
             </h2> 
             <h5 class="mb-4 text-justify">LocalAI uses P2P technologies to enable distribution of work between peers. It is possible to share an instance with Federation and/or split the weights of a model across peers (only available with llama.cpp models). You can now share computational resources between your devices or your friends!</h5>
-            
+            <!-- Warning box if p2p token is empty and p2p is enabled -->
+            {{ if and .IsP2PEnabled (eq .P2PToken "") }}
+            <div class="bg-red-500 p-4 rounded-lg shadow-lg mb-12 text-left">
+                <p class="text-xl font-semibold text-white"> <i class="fa-solid fa-exclamation-triangle"></i> Warning: P2P mode is disabled or no token was specified</p>
+                <p class="mb-4">You have to enable P2P mode by starting LocalAI with <code>--p2p</code>. Please restart the server with <code>--p2p</code> to generate a new token automatically that can be used to automatically discover other nodes. If you already have a token specify it with <code>export TOKEN=".."</code> <a href="https://localai.io/features/distribute/" target="_blank">
+                    Check out the documentation for more information.
+                </a> </p>
+            </div>
+            {{ else }}
+
             <!-- Federation Box -->
             <div class="bg-gray-800 p-6 rounded-lg shadow-lg mb-12 text-left">
 
@@ -128,7 +137,8 @@
                     </div>
                 </div>
             </div>
-            <!-- Llama.cpp Box END -->       
+            <!-- Llama.cpp Box END -->    
+            {{ end }}   
         </div>
     </div>
 

core/p2p/federated.go

@@ -0,0 +1,47 @@
+package p2p
+
+const FederatedID = "federated"
+
+type FederatedServer struct {
+	listenAddr, service, p2ptoken string
+	requestTable                  map[string]int
+	loadBalanced                  bool
+}
+
+func NewFederatedServer(listenAddr, service, p2pToken string, loadBalanced bool) *FederatedServer {
+	return &FederatedServer{
+		listenAddr:   listenAddr,
+		service:      service,
+		p2ptoken:     p2pToken,
+		requestTable: map[string]int{},
+		loadBalanced: loadBalanced,
+	}
+}
+
+func (fs *FederatedServer) SelectLeastUsedServer() string {
+	// cycle over requestTable and find the entry with the lower number
+	// if there are multiple entries with the same number, select one randomly
+	// if there are no entries, return an empty string
+	var min int
+	var minKey string
+	for k, v := range fs.requestTable {
+		if min == 0 || v < min {
+			min = v
+			minKey = k
+		}
+	}
+	return minKey
+}
+
+func (fs *FederatedServer) RecordRequest(nodeID string) {
+	// increment the counter for the nodeID in the requestTable
+	fs.requestTable[nodeID]++
+}
+
+func (fs *FederatedServer) EnsureRecordExist(nodeID string) {
+	// if the nodeID is not in the requestTable, add it with a counter of 0
+	_, ok := fs.requestTable[nodeID]
+	if !ok {
+		fs.requestTable[nodeID] = 0
+	}
+}

core/p2p/federatedServer.go

@@ -1,13 +0,0 @@
-package p2p
-
-type FederatedServer struct {
-	listenAddr, service, p2ptoken string
-}
-
-func NewFederatedServer(listenAddr, service, p2pToken string) *FederatedServer {
-	return &FederatedServer{
-		listenAddr: listenAddr,
-		service:    service,
-		p2ptoken:   p2pToken,
-	}
-}

core/p2p/federated_server.go

@@ -7,34 +7,35 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"io"
 	"net"
 	"time"
 
-	"github.com/rs/zerolog/log"
-
 	"math/rand/v2"
 
 	"github.com/mudler/edgevpn/pkg/node"
 	"github.com/mudler/edgevpn/pkg/protocol"
 	"github.com/mudler/edgevpn/pkg/types"
+	"github.com/rs/zerolog/log"
 )
 
-func (fs *FederatedServer) Start(ctx context.Context) error {
-	n, err := NewNode(fs.p2ptoken)
+func (f *FederatedServer) Start(ctx context.Context) error {
+
+	n, err := NewNode(f.p2ptoken)
 	if err != nil {
 		return fmt.Errorf("creating a new node: %w", err)
 	}
 	err = n.Start(ctx)
 	if err != nil {
-		return fmt.Errorf("starting a new node: %w", err)
+		return fmt.Errorf("creating a new node: %w", err)
 	}
 
-	if err := ServiceDiscoverer(ctx, n, fs.p2ptoken, FederatedID, nil); err != nil {
+	if err := ServiceDiscoverer(ctx, n, f.p2ptoken, f.service, func(servicesID string, tunnel NodeData) {
+		log.Debug().Msgf("Discovered node: %s", tunnel.ID)
+	}); err != nil {
 		return err
 	}
 
-	return fs.proxy(ctx, n)
+	return f.proxy(ctx, n)
 }
 
 func (fs *FederatedServer) proxy(ctx context.Context, node *node.Node) error {
@@ -84,44 +85,56 @@ func (fs *FederatedServer) proxy(ctx context.Context, node *node.Node) error {
 			}
 
 			// Handle connections in a new goroutine, forwarding to the p2p service
-			go handleConn(conn)
+			go func() {
+				var tunnelAddresses []string
+				for _, v := range GetAvailableNodes(fs.service) {
+					if v.IsOnline() {
+						tunnelAddresses = append(tunnelAddresses, v.TunnelAddress)
+					} else {
+						log.Info().Msgf("Node %s is offline", v.ID)
+					}
+				}
+
+				if len(tunnelAddresses) == 0 {
+					log.Error().Msg("No available nodes yet")
+					return
+				}
+
+				tunnelAddr := ""
+
+				if fs.loadBalanced {
+					for _, t := range tunnelAddresses {
+						fs.EnsureRecordExist(t)
+					}
+
+					tunnelAddr = fs.SelectLeastUsedServer()
+					log.Debug().Msgf("Selected tunnel %s", tunnelAddr)
+					if tunnelAddr == "" {
+						tunnelAddr = tunnelAddresses[rand.IntN(len(tunnelAddresses))]
+					}
+
+					fs.RecordRequest(tunnelAddr)
+				} else {
+					tunnelAddr = tunnelAddresses[rand.IntN(len(tunnelAddresses))]
+				}
+
+				tunnelConn, err := net.Dial("tcp", tunnelAddr)
+				if err != nil {
+					log.Error().Err(err).Msg("Error connecting to tunnel")
+					return
+				}
+
+				log.Info().Msgf("Redirecting %s to %s", conn.LocalAddr().String(), tunnelConn.RemoteAddr().String())
+				closer := make(chan struct{}, 2)
+				go copyStream(closer, tunnelConn, conn)
+				go copyStream(closer, conn, tunnelConn)
+				<-closer
+
+				tunnelConn.Close()
+				conn.Close()
+				//	ll.Infof("(service %s) Done handling %s", serviceID, l.Addr().String())
+			}()
 		}
 	}
 
 }
-
-func handleConn(conn net.Conn) {
-	var tunnelAddresses []string
-	for _, v := range GetAvailableNodes(FederatedID) {
-		if v.IsOnline() {
-			tunnelAddresses = append(tunnelAddresses, v.TunnelAddress)
-		} else {
-			log.Info().Msgf("Node %s is offline", v.ID)
-		}
-	}
-
-	// open a TCP stream to one of the tunnels
-	// chosen randomly
-	// TODO: optimize this and track usage
-	tunnelAddr := tunnelAddresses[rand.IntN(len(tunnelAddresses))]
-
-	tunnelConn, err := net.Dial("tcp", tunnelAddr)
-	if err != nil {
-		log.Error().Err(err).Msg("Error connecting to tunnel")
-		return
-	}
-
-	log.Info().Msgf("Redirecting %s to %s", conn.LocalAddr().String(), tunnelConn.RemoteAddr().String())
-	closer := make(chan struct{}, 2)
-	go copyStream(closer, tunnelConn, conn)
-	go copyStream(closer, conn, tunnelConn)
-	<-closer
-
-	tunnelConn.Close()
-	conn.Close()
-}
-
-func copyStream(closer chan struct{}, dst io.Writer, src io.Reader) {
-	defer func() { closer <- struct{}{} }() // connection is closed, send signal to stop proxy
-	io.Copy(dst, src)
-}

core/p2p/node.go

@@ -6,7 +6,6 @@ import (
 )
 
 const defaultServicesID = "services_localai"
-const FederatedID = "federated"
 
 type NodeData struct {
 	Name          string

core/p2p/p2p.go

@@ -7,6 +7,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"io"
 	"net"
 	"os"
 	"sync"
@@ -138,7 +139,7 @@ func allocateLocalService(ctx context.Context, node *node.Node, listenAddr, serv
 
 // This is the main of the server (which keeps the env variable updated)
 // This starts a goroutine that keeps LLAMACPP_GRPC_SERVERS updated with the discovered services
-func ServiceDiscoverer(ctx context.Context, n *node.Node, token, servicesID string, discoveryFunc func()) error {
+func ServiceDiscoverer(ctx context.Context, n *node.Node, token, servicesID string, discoveryFunc func(serviceID string, node NodeData)) error {
 	if servicesID == "" {
 		servicesID = defaultServicesID
 	}
@@ -160,7 +161,7 @@ func ServiceDiscoverer(ctx context.Context, n *node.Node, token, servicesID stri
 			case tunnel := <-tunnels:
 				AddNode(servicesID, tunnel)
 				if discoveryFunc != nil {
-					discoveryFunc()
+					discoveryFunc(servicesID, tunnel)
 				}
 			}
 		}
@@ -390,3 +391,8 @@ func newNodeOpts(token string) ([]node.Option, error) {
 
 	return nodeOpts, nil
 }
+
+func copyStream(closer chan struct{}, dst io.Writer, src io.Reader) {
+	defer func() { closer <- struct{}{} }() // connection is closed, send signal to stop proxy
+	io.Copy(dst, src)
+}

core/p2p/p2p_disabled.go

@@ -14,11 +14,11 @@ func GenerateToken() string {
 	return "not implemented"
 }
 
-func (fs *FederatedServer) Start(ctx context.Context) error {
+func (f *FederatedServer) Start(ctx context.Context) error {
 	return fmt.Errorf("not implemented")
 }
 
-func ServiceDiscoverer(ctx context.Context, node *node.Node, token, servicesID string, fn func()) error {
+func ServiceDiscoverer(ctx context.Context, node *node.Node, token, servicesID string, fn func(string, NodeData)) error {
 	return fmt.Errorf("not implemented")
 }
 

core/schema/openai.go

@@ -179,8 +179,7 @@ type OpenAIRequest struct {
 	// A grammar to constrain the LLM output
 	Grammar string `json:"grammar" yaml:"grammar"`
 
-	JSONFunctionGrammarObject     *functions.JSONFunctionStructureFunction `json:"grammar_json_functions" yaml:"grammar_json_functions"`
-	JSONFunctionGrammarObjectName *functions.JSONFunctionStructureName     `json:"grammar_json_name" yaml:"grammar_json_name"`
+	JSONFunctionGrammarObject *functions.JSONFunctionStructure `json:"grammar_json_functions" yaml:"grammar_json_functions"`
 
 	Backend string `json:"backend" yaml:"backend"`
 

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

@@ -152,7 +152,8 @@ function:
     replace_function_results: [] # Placeholder to replace function call results with arbitrary strings or patterns.
     replace_llm_results: [] # Replace language model results with arbitrary strings or patterns.
     capture_llm_results: [] # Capture language model results as text result, among JSON, in function calls. For instance, if a model returns a block for "thinking" and a block for "response", this will allow you to capture the thinking block.
-    return_name_in_function_response: false # Some models might prefer to use "name" rather then "function" when returning JSON data. This will allow to use "name" as a key in the JSON response.
+    function_name_key: "name"
+    function_arguments_key: "arguments"
 
 # Feature gating flags to enable experimental or optional features.
 feature_flags: {}