Move files in a sub forlder dedicated to chat-ai.

chat-ai/Dockerfile

@@ -0,0 +1,71 @@
+# Étape 1 : build de llama.cpp
+FROM ubuntu:24.04 AS builder
+
+LABEL maintainer="vous" \
+      description="llama.cpp + Mistral-7B-Instruct (inférence locale isolée)"
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      lsb-release           \
+      gnupg                 \
+      ca-certificates       \
+      build-essential       \
+      cmake                 \
+      git                   \
+      libopenblas-dev       \
+      pkg-config            \
+      libvulkan-dev         \
+      vulkan-tools          \
+      wget
+
+# Télécharge et installe le Vulkan SDK
+RUN wget -qO - https://packages.lunarg.com/lunarg-signing-key-pub.asc | apt-key add - && \
+    echo "deb https://packages.lunarg.com/vulkan/ $(lsb_release -sc) main" > /etc/apt/sources.list.d/lunarg-vulkan.list && \
+    apt-get update && \
+    apt-get install -y vulkan-sdk && \
+    rm -rf /var/lib/apt/lists/*
+
+WORKDIR /build
+RUN git clone --depth 1 https://github.com/ggerganov/llama.cpp .
+
+RUN cmake -B build \
+       -DLLAMA_BUILD_TESTS=OFF \
+       -DLLAMA_BUILD_EXAMPLES=ON \
+       -DGGML_BLAS=ON \
+       -DGGML_BLAS_VENDOR=OpenBLAS \
+       -DCMAKE_INSTALL_PREFIX=/install \
+       -DGGML_VULKAN=ON \
+       -DVulkan_GLSLC_EXECUTABLE=/usr/bin/glslc \
+    && cmake --build build --config Release -j$(nproc) \
+    && cmake --install build --prefix /install
+
+# Étape 2 : image finale (runtime minimal)
+FROM ubuntu:24.04 AS runtime
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      libopenblas0-openmp   \
+      libgomp1              \
+      libvulkan1            \
+      mesa-vulkan-drivers   \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY --from=builder /install/ /usr/local/
+RUN ldconfig
+
+RUN useradd -m -u 1001 -s /bin/bash llama
+USER llama
+
+WORKDIR /home/llama
+RUN mkdir -p /home/llama/models
+
+EXPOSE 8080
+
+CMD ["llama-server", \
+     "--model",   "/home/llama/models/mistral-7b-instruct.gguf", \
+     "--host",    "0.0.0.0", \
+     "--port",    "8080", \
+     "-c",        "4096", \
+     "-ngl",      "33"]

chat-ai/compose.yml

@@ -0,0 +1,50 @@
+name: mistral-local
+
+services:
+  local-ai:
+    build:
+      context: .
+      dockerfile: Dockerfile
+    image: local-ai
+    container_name: local-ai
+    networks:
+      - mistral-net
+    ports:
+      - "127.0.0.1:8080:8080"
+    # Limites de ressources
+    deploy:
+      resources:
+        limits:
+          memory: 12g
+          cpus: "8"
+    # Sécurité
+    read_only: true
+    security_opt:
+      - no-new-privileges:true
+    tmpfs:
+      - /tmp:size=256m
+    devices:
+      - /dev/dri:/dev/dri
+    group_add:
+      - video
+    environment:
+      - GGML_VK_VISIBLE_DEVICES=0
+    volumes:
+      - ../ai-models/mistral-7b-instruct.gguf:/home/llama/models/mistral-7b-instruct.gguf:ro
+      - ../workspace:/home/llama/workspace
+
+    # Health check — vérifie que le serveur répond
+    # healthcheck:
+    #   test: ["CMD-SHELL", "curl -sf http://localhost:8080/health || exit 1"]
+    #   interval: 30s
+    #   timeout: 10s
+    #   retries: 3
+    #   start_period: 60s    # laisser le temps au modèle de charger
+
+networks:
+  mistral-net:
+    name: mistral-net
+    driver: bridge
+    internal: false         # pas de passerelle vers Internet
+    driver_opts:
+      com.docker.network.bridge.name: br-mistral

chat-ai/local-ai.bash

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+action=$1
+args=$2
+
+build_image() {
+    docker build $args -t local-ai .
+}
+
+start_container() {
+    # Premier démarrage (build + lancement)
+    docker compose up --build
+
+    # Voir les logs en temps réel
+    docker compose logs -f
+
+    # Arrêter
+    docker compose down
+
+    # Vérifier l'état du health check
+    docker compose ps
+}
+
+main() {
+    if [[ $action = 'build' ]] then
+        build_image
+    elif [[ $action = 'all' ]] then
+        build_image
+        start_container
+    else
+        start_container
+    fi
+}
+
+main

chat-ai/requirements.txt

@@ -0,0 +1,4 @@
+torch==2.2.0 --index-url https://download.pytorch.org/whl/cpu
+transformers==4.40.0
+accelerate==0.27.0
+sentencepiece