"""
LLM Serving Layer with RAG.

FastAPI-based HTTP API that wraps the existing Qwen3 model, tokenizer,
and generation code into a production-style serving layer. Includes:
- Streaming chat via Server-Sent Events (SSE)
- Retrieval-Augmented Generation (RAG) with FAISS + sentence-transformers
- Document ingestion with recursive text chunking
"""