RAG Stack (Qwen + Qdrant) on Ubuntu

End-to-end pipeline: File Share -> Crawler -> Text Extractor -> Embeddings (bge-m3) -> Qdrant -> RAG API -> llama.cpp (Qwen 7B/14B)

Components

• Ingestion worker (Python): crawls SMB/FS, extracts text with unstructured, chunks, embeds via BAAI/bge-m3, upserts into Qdrant.
• Vector DB: Qdrant (Docker).
• RAG API (FastAPI): retrieves from Qdrant and calls llama.cpp (/completion) on 7B (8080) or 14B (8081).
• Systemd services: qwen3-7b, qwen3-14b, rag-app, plus nightly rag-ingest.timer.

See config/settings.yaml for paths and knobs. EOF