Upload 12 files

Dockerfile

@@ -14,9 +14,10 @@
 
 FROM python:3.11-slim
 
-# curl/ca-certs only; the prebuilt wheel means we need NO compiler toolchain.
+# build-essential + cmake so llama-cpp-python compiles from source if no
+# prebuilt wheel matches (python:slim has no compiler -> CMake "gcc not found").
 RUN apt-get update && apt-get install -y --no-install-recommends \
-        curl ca-certificates \
+        build-essential cmake git curl ca-certificates \
     && rm -rf /var/lib/apt/lists/*
 
 # Hugging Face Spaces run the container as a non-root user with UID 1000.
@@ -25,7 +26,9 @@ USER user
 ENV HOME=/home/user \
     PATH=/home/user/.local/bin:$PATH
 
-# --- Python deps: prebuilt CPU wheel installs in seconds (no source build) ---
+# --- Python deps. GGML_NATIVE=OFF -> portable CPU binary (no -march=native),
+# so it can't crash with "illegal instruction" if build CPU != runtime CPU. ---
+ENV CMAKE_ARGS="-DGGML_NATIVE=OFF" FORCE_CMAKE=1
 RUN pip install --no-cache-dir --user "llama-cpp-python[server]" huggingface_hub
 
 # --- Model: which GGUF to serve ---------------------------------------------

hf-space-rag/Dockerfile

@@ -9,8 +9,10 @@
 
 FROM python:3.11-slim
 
+# build-essential + cmake so llama-cpp-python can compile if no prebuilt wheel
+# matches (python:slim has no compiler otherwise -> CMake "gcc not found").
 RUN apt-get update && apt-get install -y --no-install-recommends \
-        curl ca-certificates && rm -rf /var/lib/apt/lists/*
+        build-essential cmake git curl ca-certificates && rm -rf /var/lib/apt/lists/*
 
 # Hugging Face Spaces run the container as non-root UID 1000.
 RUN useradd -m -u 1000 user
@@ -19,8 +21,11 @@ ENV HOME=/home/user \
     PATH=/home/user/.local/bin:$PATH
 WORKDIR /home/user/app
 
-# --- Python deps (prebuilt wheels -> fast, no build toolchain) ---
+# --- Python deps ---
+# GGML_NATIVE=OFF -> portable CPU binary (no -march=native), so it can't crash
+# with "illegal instruction" if the build CPU != the runtime CPU.
 COPY --chown=user requirements.txt .
+ENV CMAKE_ARGS="-DGGML_NATIVE=OFF" FORCE_CMAKE=1
 RUN pip install --no-cache-dir --user -r requirements.txt
 
 # --- Model config ------------------------------------------------------------

hf-space-rag/requirements.txt

@@ -1,4 +1,5 @@
-# Prebuilt CPU wheel for llama.cpp -> fast build, no compiler needed.
+# Try a prebuilt CPU wheel first (fast); if none matches, the Dockerfile's
+# build-essential/cmake lets it compile from source instead of failing.
 --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
 llama-cpp-python