Update Dockerfile

Dockerfile

@@ -1,50 +1,64 @@
-# Start with a Python base image, which is convenient for running the Gradio app.
-FROM python:3.9
+# Use a base Ubuntu image
+FROM ubuntu:22.04
 
 # Set environment variables to prevent interactive prompts during apt operations
 ENV DEBIAN_FRONTEND=noninteractive
 
-# Install system dependencies: curl for Ollama and wget for model download.
-RUN apt-get update && apt-get install -y curl wget
+# Install system dependencies: curl for Ollama, wget for model download, git, python3 and pip
+RUN apt update && apt install -y curl wget git python3 python3-pip
 
-# Install Ollama using its official installation script.
+# --- PERMISSIONS FIX 1: CREATE USER ---
+# Create a non-root user named 'user' with user ID 1000.
+# This matches the user ID that Hugging Face Spaces uses to run the container.
+# The -m flag creates a home directory at /home/user.
+RUN useradd -m -u 1000 user
+
+# Install Ollama using its official installation script
+# This will be run as root, which is correct for system-wide installation.
 RUN curl -fsSL https://ollama.com/install.sh | sh
 
-# Set environment variables for the model for easy modification.
-ENV MODEL_REPO="unsloth/gemma-3-4b-it-qat-GGUF"
-ENV MODEL_FILENAME="gemma-3-4b-it-qat.Q4_K_M.gguf"
+# Set the working directory inside the container
+WORKDIR /app
 
-# Create a non-root user and switch to its home directory. This is a best practice for security and avoids permission errors with model caching.
-# As seen in the Hugging Face Docker guide. [1]
-RUN useradd -m -u 1000 user
-WORKDIR /home/user/app
+# Define environment variables for the model repository and filename
+ENV MODEL_REPO="unsloth/gemma-3-4b-it-GGUF"
+ENV MODEL_FILENAME="gemma-3-4b-it.Q4_K_M.gguf"
+ENV MODEL_enableDL="?download=true"
+
+# Download the specific GGUF model file directly from Hugging Face Hub.
+RUN wget -O ${MODEL_FILENAME} https://huggingface.co/${MODEL_REPO}/resolve/main/${MODEL_FILENAME}${MODEL_enableDL}
 
-# Download the GGUF model file directly from Hugging Face Hub.
-RUN wget "https://huggingface.co/unsloth/gemma-3-4b-it-qat-GGUF/resolve/main/gemma-3-4b-it-qat-Q4_K_M.gguf?download=true" -O ${MODEL_FILENAME}
+# This OK.... Download the GGUF model file directly from Hugging Face Hub.
+#RUN wget "https://huggingface.co/unsloth/gemma-3-4b-it-qat-GGUF/resolve/main/gemma-3-4b-it-qat-Q4_K_M.gguf?download=true" -O ${MODEL_FILENAME}
 
-# Copy the application files and set the correct ownership to the new user. [1]
-COPY --chown=user:user Modelfile .
-COPY --chown=user:user app.py .
-COPY --chown=user:user requirements.txt .
-COPY --chown=user:user run.sh .
+# --- PERMISSIONS FIX 2: COPY FILES WITH OWNERSHIP ---
+# Copy all application files into the container.
+# Use the --chown=user flag to set the owner of these files to the 'user' we created.
+# This is critical for giving the application write permissions.
+COPY --chown=user Modelfile .
+COPY --chown=user app.py .
+COPY --chown=user requirements.txt .
+COPY --chown=user run.sh .
 
-# Install Python dependencies.
+# Install Python dependencies required by your Gradio application
 RUN pip install --no-cache-dir -r requirements.txt
 
-# Make the startup script executable.
+# Make the startup script executable
 RUN chmod +x run.sh
 
-# Switch to the non-root user. [1]
+# --- PERMISSIONS FIX 3: SWITCH TO USER ---
+# Switch the context of the Dockerfile to our non-root user.
+# All subsequent commands (including the final CMD) will now be executed as 'user', not 'root'.
 USER user
 
-# Set environment variables for the user and for Ollama.
-# This ensures cached models are stored in a writable directory and that Ollama is accessible within the container network. [1]
-ENV HOME=/home/user
-ENV PATH=/home/user/.local/bin:$PATH
-ENV OLLAMA_HOST=127.0.0.1
+# Set environment variables for the user's home directory. This ensures that
+# tools like pip and transformers cache files in a location writable by the user (/home/user/.cache).
+ENV HOME=/home/user \
+	PATH=/home/user/.local/bin:$PATH
 
-# Expose the port for the Gradio application.
+# Expose the port that your Gradio application will listen on.
 EXPOSE 7860
 
-# Set the command to run our startup script.
+# Set the entrypoint for the container to execute our startup script.
+# This will now be run as the 'user'.
 CMD ["./run.sh"]