Update Dockerfile

Dockerfile

@@ -1,24 +1,28 @@
-# 1. Use Python 3.9 (This is the version that worked for you)
-FROM python:3.9
+# 1. Use Python 3.9 (Standard, Safe)
+FROM python:3.9-slim
 
-# 2. Upgrade pip (Essential to find the correct wheels)
+# 2. Install wget so we can download the file manually
+# We also update system tools to prevent weird glitches
+RUN apt-get update && apt-get install -y \
+    wget \
+    build-essential \
+    && rm -rf /var/lib/apt/lists/*
+
+# 3. CRITICAL: Upgrade pip
 RUN pip install --no-cache-dir --upgrade pip
 
-# 3. Install the specific Hugging Face version (Prevents the crash)
-RUN pip install "huggingface_hub<0.25.0"
+# 4. Manually Download the CORRECT Wheel (Bypassing pip's search)
+# We download the 'manylinux' (Standard Linux) version to a file named 'llama.whl'
+RUN wget -O llama.whl https://github.com/abetlen/llama-cpp-python/releases/download/v0.2.56/llama_cpp_python-0.2.56-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
 
-# 4. Install llama-cpp-python
-# We use the EXACT method that worked in Turn 16.
-# --prefer-binary forces it to download the wheel (fast) instead of building.
-RUN pip install --no-cache-dir \
-    llama-cpp-python \
-    --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu \
-    --prefer-binary
+# 5. Install the wheel we just downloaded
+RUN pip install llama.whl
 
-# 5. Install Gradio
-RUN pip install --no-cache-dir gradio
+# 6. Install the rest (Pinned to prevent the HfFolder crash)
+RUN pip install "huggingface_hub<0.25.0" gradio
 
-# 6. Run the app
+# 7. Cleanup and Run
+RUN rm llama.whl
 WORKDIR /app
 COPY . .
 CMD ["python", "app.py"]