RAG and agent prototypes

Retrieval and agent prototypes with LangChain and LangGraph, working up from plain RAG to agentic graphs, all on local models.

Things I implemented:

• Chunking and embedding local corpora into FAISS
• LangChain retrieval chains over private documents
• LangGraph state machines that decide when to retrieve, re-query or answer
• Everything runnable against local models through Ollama
• Notes along the way: chunking is the real quality lever, not the framework