perf(llm): install prebuilt AVX2 llama-cpp wheel (no HF-build compile) + Qwen2.5-3B for citations

a406137

verified

MHamdan commited on 23 days ago