Update Dockerfile

Dockerfile

@@ -1,5 +1,3 @@
-# Use Python 3.11 Slim (Debian-based)
-# WHY: The pre-built AI engine is built for Debian (glibc), not Alpine (musl).
 FROM python:3.11-slim
 
 # Install Node.js (for promptfoo) and system tools
@@ -15,14 +13,20 @@ RUN npm install -g promptfoo
 
 WORKDIR /app
 
-# --- CRITICAL INSTALLATION STEP ---
-# We install the AI engine separately with three safety flags:
-# 1. --extra-index-url: Points to the pre-built CPU library.
-# 2. --only-binary :all: STRICTLY forbids compilation. If the pre-built file isn't found, it will error immediately (fail fast) rather than trying to compile for 30 mins.
-# 3. --no-cache-dir: Keeps the image small.
-RUN pip install --no-cache-dir --only-binary :all: \
+# --- ULTIMATE FIX: Force Specific Platform ---
+# We do not trust pip to choose. We force it to download the 'manylinux' (Debian) version.
+# 1. Download the specific wheel for Debian (manylinux_2_17_x86_64).
+# 2. Install the downloaded file directly.
+RUN pip download --only-binary :all: \
+    --platform manylinux_2_17_x86_64 \
+    --python-version 311 \
+    --implementation cp \
+    --abi cp311 \
     "llama-cpp-python" \
-    --extra-index-url "https://abetlen.github.io/llama-cpp-python/whl/cpu"
+    --extra-index-url "https://abetlen.github.io/llama-cpp-python/whl/cpu" \
+    -d /tmp/wheels
+
+RUN pip install --no-cache-dir /tmp/wheels/llama*.whl && rm -rf /tmp/wheels
 
 # --- Install Standard Requirements ---
 COPY requirements.txt .
@@ -33,5 +37,4 @@ COPY . .
 
 EXPOSE 7860
 
-# Run server
 CMD ["uvicorn", "server:app", "--host", "0.0.0.0", "--port", "7860"]