Update Dockerfile

Dockerfile

@@ -1,32 +1,40 @@
-# Use a base Ubuntu image
-FROM ubuntu:22.04
+# Use the official Ollama base image, which already includes Ollama
+FROM ollama/ollama:latest
 
 # Set environment variables to prevent interactive prompts during apt operations
 ENV DEBIAN_FRONTEND=noninteractive
 
-# Install system dependencies: curl for Ollama, wget for model download, git (optional, but good for debugging), python3 and pip for the application
-RUN apt update && apt install -y curl wget git python3 python3-pip
+# Install system dependencies: git (optional), python3 and pip for the application
+RUN apt update && apt install -y git python3 python3-pip
 
-# Install Ollama using its official installation script
-# This script automatically detects the system architecture and installs the correct binary.
-RUN curl -fsSL https://ollama.com/install.sh | sh
+# Set up a non-root user as recommended for Hugging Face Spaces
+RUN useradd -m -u 1000 user
+USER user
+ENV HOME=/home/user \
+	PATH=/home/user/.local/bin:$PATH \
+    OLLAMA_HOST=0.0.0.0 # Allow Ollama to be accessed from outside localhost
 
 # Set the working directory inside the container
-WORKDIR /app
-
-# Define environment variables for the model repository and filename
-# This makes it easy to change the model later without editing the RUN command directly.
-ENV MODEL_REPO="unsloth/gemma-3-4b-it-GGUF"
-# CORRECTED FILENAME: Removed '-qat'
-ENV MODEL_FILENAME="gemma-3-4b-it.Q4_K_M.gguf"
-
-# Download the specific GGUF model file directly from Hugging Face Hub.
-# We use 'wget -O' to save the file with the desired filename in the current directory.
-# The 'resolve/main' path ensures we get the raw file content.
-RUN wget -O ${MODEL_FILENAME} https://huggingface.co/${MODEL_REPO}/resolve/main/${MODEL_FILENAME}
-
-# Copy the Modelfile, application script, Python requirements, and the startup script into the container
-COPY Modelfile .
+WORKDIR $HOME/app
+
+# Define the model to be pulled
+ENV OLLAMA_HF_MODEL="hf.co/unsloth/gemma-3-4b-it-qat-GGUF:Q4_K_M"
+
+# Pull the model during the build process
+# Start ollama serve in background, wait for it, pull the model, then kill ollama.
+# This ensures the model is downloaded and cached within the image.
+RUN ollama serve & \
+    echo "Waiting for Ollama to start for model pull..." && \
+    sleep 5 && \
+    while ! curl -s http://localhost:11434 > /dev/null; do \
+        sleep 1; \
+    done && \
+    echo "Ollama started. Pulling model: ${OLLAMA_HF_MODEL}" && \
+    ollama pull ${OLLAMA_HF_MODEL} && \
+    echo "Model pull complete. Stopping Ollama for build process." && \
+    pkill ollama || true # '|| true' to prevent build failure if pkill returns non-zero when ollama already stopped
+
+# Copy the application files
 COPY app.py .
 COPY requirements.txt .
 COPY run.sh .