Update Dockerfile

Dockerfile

@@ -1,7 +1,6 @@
-# Use Python 3.11 slim image
 FROM python:3.11-slim
 
-# 1. Install Node.js (for promptfoo) and basic tools
+# Install Node.js (for promptfoo) and system tools
 RUN apt-get update && apt-get install -y \
     curl \
     git \
@@ -9,25 +8,24 @@ RUN apt-get update && apt-get install -y \
     npm \
     && rm -rf /var/lib/apt/lists/*
 
-# 2. Install promptfoo globally
+# Install promptfoo globally
 RUN npm install -g promptfoo
 
-# Set working directory
 WORKDIR /app
 
-# 3. Install Python dependencies (FastAPI, etc.)
+# --- CRITICAL FIX: Install the AI Engine separately FIRST ---
+# We use the specific "CPU" URL to ensure we get the pre-built binary.
+RUN pip install --no-cache-dir \
+    "llama-cpp-python==0.2.90" \
+    --extra-index-url "https://abetlen.github.io/llama-cpp-python/whl/cpu"
+
+# --- Install the rest of your requirements ---
 COPY requirements.txt .
 RUN pip install --no-cache-dir -r requirements.txt
 
-# 4. CRITICAL FIX: Install the Pre-Built AI Engine
-# We directly download the compiled wheel file so we don't have to compile it (saves 30 mins)
-RUN curl -L "https://github.com/abetlen/llama-cpp-python/releases/download/v0.2.90/llama_cpp_python-0.2.90-cp311-cp311-manylinux_2_17_x86_64.whl" -o llama.whl && \
-    pip install llama.whl
-
-# 5. Copy your application code
+# Copy application code
 COPY . .
 
-# Expose port
 EXPOSE 7860
 
 # Run server