upgrade dockerfile

Dockerfile

@@ -16,15 +16,39 @@ RUN apt-get update && \
     cmake \
     && rm -rf /var/lib/apt/lists/*
 
-# Обновление pip и установка setuptools/wheel
+WORKDIR /app
+
+# Кэшируем установку зависимостей Python
+COPY requirements.txt .
 RUN pip install --upgrade pip && \
-    pip install setuptools wheel
+    pip install setuptools wheel \
+    pip install --no-cache-dir -r requirements.txt
 
-# Проверка версии Python и pip
-RUN python --version && pip --version
+# Сборка RKLLM runtime
+RUN git clone --depth 1 https://git.home/drholy/ezrknn-llm /tmp/rkllm-build && \
+    cp -r /tmp/rkllm-build/rkllm-runtime/runtime/Linux/librkllm_api/aarch64/* /usr/lib && \
+    cp -r /tmp/rkllm-build/rkllm-runtime/runtime/Linux/librkllm_api/include/* /usr/local/include && \
+    cd /tmp/rkllm-build/rkllm-runtime/examples/rkllm_api_demo && \
+    bash build-linux.sh && \
+    cp ./build/build_linux_aarch64_Release/llm_demo /usr/bin/rkllm && \
+    # Очистка временных файлов сборки
+    ldconfig && \
+    rm -rf /tmp/rkllm-build && \
+    apt-get purge -y git cmake build-essential libssl-dev && \
+    apt-get autoremove -y && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
 
-RUN pip install --no-cache-dir -r requirements.txt
+# Копируем приложение (исключая кэш и модели)
+COPY . .
 
-# Очистка кэша
-RUN apt-get clean && \
-    rm -rf /tmp/* /var/tmp/*
+# Исключаем модели из образа — монтируем как volume при запуске
+RUN rm -rf /app/models/* 2>/dev/null || true && \
+    find /app -type d -name __pycache__ -exec rm -rf {} + 2>/dev/null || true && \
+    chmod +x /app/entrypoint.sh
+
+# Порт Gradio по умолчанию
+EXPOSE 8080
+
+WORKDIR /app
+ENTRYPOINT ["/app/entrypoint.sh"]

app/entrypoint.sh

@@ -1,34 +1,18 @@
 #!/bin/bash
-message_print() {
-  echo
-  echo "#########################################"
-  echo $1
-  echo "#########################################"
-  echo
-}
+set -e  # Автоматический выход при ошибке
 
-cd /
+echo
+echo "#########################################"
+echo "Starting RKLLM Gradio Server"
+echo "#########################################"
+echo
 
-message_print "Changing to repository..."
-git clone https://git.home/drholy/ezrknn-llm
-cd ezrknn-llm/
-cp ./rkllm-runtime/runtime/Linux/librkllm_api/aarch64/* /usr/lib
-cp ./rkllm-runtime/runtime/Linux/librkllm_api/include/* /usr/local/include
+# Проверка наличия моделей (монтируются через volume)
+if [ ! -d "/app/models" ] || [ -z "$(ls -A /app/models 2>/dev/null)" ]; then
+    echo "⚠️  WARNING: Models directory is empty!"
+    echo "    Mount models volume: -v ./models:/app/models"
+    sleep 5
+fi
 
-message_print "Compiling LLM runtime for Linux..."
-cd ./rkllm-runtime/examples/rkllm_api_demo/
-bash build-linux.sh
-
-message_print "Moving rkllm to /usr/bin..."
-cp ./build/build_linux_aarch64_Release/llm_demo /usr/bin/rkllm
-echo "* soft nofile 16384" >> /etc/security/limits.conf
-echo "* hard nofile 1048576" >> /etc/security/limits.conf
-
-message_print "Increasing file limit for all users (needed for LLMs to run)..."
-echo "root soft nofile 16384" >> /etc/security/limits.conf
-echo "root hard nofile 1048576" >> /etc/security/limits.conf
-message_print "Done installing ezrknn-llm!"
-
-message_print "Start Gradio server"
-cd /app
-python rkllm_server_gradio.py
+# Запуск сервера
+exec python rkllm_server_gradio.py

docker-compose.yaml

@@ -1,16 +1,16 @@
 services:
   rkllm:
-    build:
-      context: .
-      dockerfile: Dockerfile
+    build: .
     platform: linux/arm64/v8
     container_name: rkllm
     restart: always
     privileged: true
     volumes:
-      - ./app:/app
-      - ./ezrknn-llm:/ezrknn-llm
-    entrypoint: ./app/entrypoint.sh
+      - ./models:/app/models:ro
+    ulimits:
+      nofile:
+        soft: 16384
+        hard: 1048576
     ports:
       - "8080:8080"
     networks: