rag-knowledge-baselisted

# Semantic Search & RAG Instructions 1. Load document strings and clean HTML/markdown syntax. 2. Chunk text using recursive character splitting (target chunk size: 500, overlap: 50). 3. Compute embeddings using model API. 4. Insert chunks and embeddings into local vector store (e.g. Chroma, FAISS). 5. For queries, embed query string and retrieve top 3 nearest chunks. 6. Format prompt template: Context + Query -> Answer.