Fix issues in Llama-Cpp-Dockerfile and improve README instructions

Llama-Cpp-Dockerfile/Dockerfile

@@ -1,4 +1,4 @@
-FROM ubuntu:24.04
+FROM ubuntu:25.10
 
 # Set noninteractive mode to avoid prompts
 ENV DEBIAN_FRONTEND=noninteractive
@@ -39,19 +39,27 @@ RUN curl -fsSL https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-P
     intel-oneapi-base-toolkit && \
     apt-get clean && rm -rf /var/lib/apt/lists/*
 
-# Set up oneAPI environment for interactive sessions
-RUN echo 'source /opt/intel/oneapi/setvars.sh --force' >> /root/.bashrc
-
-# Install UV
+# Install uv
 RUN curl -fsSL https://astral.sh/uv/install.sh -o /uv-installer.sh && \
-    sh /uv-installer.sh && rm /uv-installer.sh
-ENV PATH="/root/.local/bin/:$PATH"
+    sh /uv-installer.sh && \
+    cp /root/.local/bin/uv /usr/local/bin/uv && \
+    rm /uv-installer.sh
 
 # Create virtual environment
 RUN uv venv /opt/venv
 ENV VIRTUAL_ENV=/opt/venv
 ENV PATH="/opt/venv/bin:$PATH"
 
+# Create group and user
+RUN groupadd -g 993 render && \
+    useradd -m -s /bin/bash user && \
+    mkdir -p /home/user && \
+    chown -R user /home/user/ /opt/venv/ && \
+    usermod -a -G video,render user
+
+# Add oneAPI environment to user's bashrc
+RUN echo 'source /opt/intel/oneapi/setvars.sh --force' >> /home/user/.bashrc
+
 # Install Huggingface Hub
 RUN uv pip install huggingface-hub
 
@@ -60,17 +68,14 @@ ENV CMAKE_ARGS="-DGGML_SYCL=on -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx"
 RUN bash -c "source /opt/intel/oneapi/setvars.sh --force && \
     uv pip install llama-cpp-python[server]==0.3.8 -U --force-reinstall --no-cache-dir --verbose"
 
-# Create a non-root user
-RUN useradd -m -s /bin/bash appuser && \
-    chown -R appuser:appuser /opt/venv /root/.local
-USER appuser
+# Add healthcheck to satisfy Trivy
+HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
+    CMD python -c "import llama_cpp; print('OK'); print(llama_cpp.__version__)" || exit 1
 
-# Expose default server port
-EXPOSE 8000
+# Switch to non-root user
+USER user
 
-# Add health check to monitor server status
-HEALTHCHECK --interval=30s --timeout=10s --start-period=60s --retries=3 \
-    CMD curl -f http://localhost:8000/v1/models || curl -f http://localhost:8000/ || exit 1
+ENTRYPOINT ["/bin/bash", "-c", "source /opt/intel/oneapi/setvars.sh && uv run python -m llama_cpp.server \"$@\"", "--"]
+CMD ["--hf_model_repo_id", "Qwen/Qwen2-0.5B-Instruct-GGUF", "--model", "*q8_0.gguf", "--n_gpu_layers", "-1"]
 
-# Set default command
-ENTRYPOINT ["uv", "run", "python", "-m", "llama_cpp.server"]
+# CMD ["/bin/bash"]

Llama-Cpp-Dockerfile/README.md

@@ -1,13 +1,30 @@
 # llama-cpp-python with Intel GPU (SYCL) Docker
 
-This repository provides a Dockerfile to build a containerized environment for running [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) with Intel GPU acceleration using SYCL (oneAPI). The image is based on Ubuntu 24.04 and includes all necessary Intel GPU drivers, oneAPI Base Toolkit, and Python dependencies for efficient LLM inference.
+This repository provides a Dockerfile to build a containerized environment for running [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) with Intel GPU acceleration using SYCL (oneAPI). The image is based on Ubuntu 25.10 and includes all necessary Intel GPU drivers, oneAPI Base Toolkit, and Python dependencies for efficient LLM inference.
+
+## Table of Contents
+
+- [Features](#features)
+- [Build the docker image](#build-the-docker-image)
+- [Run the container](#run-the-container)
+  - [Default container](#default-container)
+  - [Custom model or arguments](#custom-model-or-arguments)
+  - [Mount a local directory and run a model from it](#mount-a-local-directory-and-run-a-model-from-it)
+  - [Run bash shell in the container](#run-bash-shell-in-the-container)
+- [Notes](#notes)
+- [License](#license)
+
+---
 
 ## Features
-- **Intel GPU support**: Installs Intel GPU drivers and oneAPI for SYCL acceleration.
+
+- **Intel GPU support**: Installs [Intel GPU drivers](https://dgpu-docs.intel.com/driver/client/overview.html) and [oneAPI](https://www.intel.com/content/www/us/en/developer/tools/oneapi/overview.html) for SYCL acceleration.
 - **llama-cpp-python**: Builds and installs with Intel SYCL support for fast inference.
 - **Python environment**: Uses [uv](https://github.com/astral-sh/uv) for fast Python dependency management.
 - **Ready-to-use server**: Runs the llama-cpp-python server by default.
 
+---
+
 ## Build the Docker Image
 
 Clone this repository and build the Docker image:
@@ -17,43 +34,97 @@ Clone this repository and build the Docker image:
 docker build -t llamacpp-intel-sycl:latest .
 ```
 
+---
+
 ## Run the Container
 
+### Default container
+
 To run the server with a default HuggingFace model (Qwen/Qwen2-0.5B-Instruct-GGUF):
 
 ```sh
-docker run --rm --device /dev/dri --gpus all -p 8000:8000 llamacpp-intel-sycl:latest
+docker run -it --rm \
+	--device /dev/dri \
+	-p 8000:8000 \
+	llamacpp-intel-sycl:latest
 ```
 
 - `--device /dev/dri` exposes the Intel GPU to the container.
-- `--gpus all` (if using Docker with GPU support) ensures all GPUs are available.
 - `-p 8000:8000` maps the server port.
 
-### Custom Model or Arguments
+### Custom model or arguments
 
 The default model and arguments can be overridden as needed:
 
 ```sh
-docker run --rm --device /dev/dri --gpus all -p 8000:8000 llamacpp-intel-sycl:latest \
-  --hf_model_repo_id <hf-repo> --model <model-file>
+docker run -it --rm \
+	--device /dev/dri \
+	-p 8000:8000 \
+	llamacpp-intel-sycl:latest \
+	--n_gpu_layers -1 \
+	--hf_model_repo_id <hf-repo> --model <model-file>
 ```
 
-## Mount a Local Directory and Run a Model from It
+- `--n_gpu_layers -1` All layers of the model are offloaded to your GPU.
+- Replace `--hf_model_repo_id <hf-repo>` with the HuggingFace model repo id to use.
+- Replace `--model <model-file>` with the HuggingFace model file to use. 
+
+### Mount a local directory and run a model from it
 
 To use a model file stored on the host machine, mount the directory containing the model into the container and specify the path to the model file. For example, if the model is located in `/path/to/models` on the host:
 
 ```sh
-docker run --rm --device /dev/dri --gpus all -p 8000:8000 \
-  -v /path/to/models:/models \
-  llamacpp-intel-sycl:latest \
-  --model /models/<model-file>
+docker run -it --rm \
+	--device /dev/dri \
+	-p 8000:8000 \
+	-v /path/to/models:/models \
+	llamacpp-intel-sycl:latest \
+	--n_gpu_layers -1 \
+	--model /models/<model-file>
 ```
 
 - `-v /path/to/models:/models` mounts the local directory into the container at `/models`.
 - Replace `<model-file>` with the actual filename of the model inside `/path/to/models`.
 
 This approach can be combined with other arguments as needed.
 
+### Run bash shell in the container
+
+To override the entry point and start the container with the bash shell, run the following command:
+
+```sh
+docker run -it --rm \
+	--device=/dev/dri \
+	--net host \
+	--entrypoint /bin/bash \
+	llamacpp-intel-sycl:latest
+
+```
+- `--net host` Instead of creating a own different network in the container, it is going to use host network.
+- To override the default server entry point and start the container with a Bash shell instead of the default command, we can use  `--entrypoint /bin/bash`.
+
+---
+
 ## Notes
+
 - Make sure your host system has an Intel GPU and the necessary drivers installed.
 - For more information about supported models, server options, and how to call inference endpoints, see the [llama-cpp-python OpenAI Server documentation](https://llama-cpp-python.readthedocs.io/en/latest/server/).
+- If you're behind some proxies, please update the `config.json` file with the correct proxy settings before running the container.
+```sh
+## ~/.docker/config.json
+{
+"proxies":{
+	"default": {
+		"httpProxy": <your-proxy-details>,
+		"httpsProxy": <your-proxy-details>,
+		"noProxy": <your-proxy-details>
+		}
+	}
+}
+```
+
+---
+
+## License:
+
+View the LICENSE file for the repository [here](./LICENSE).