RAG Chatbot Developer - LangChain + Vector Search over Internal Docs

We have a growing set of internal documents (playbooks, templates, and reference material) that the team currently searches by hand. We want a chatbot that answers questions from that content accurately, with the source shown, instead of generic model output. This is small in scope but should be built properly. We'd rather have a clean, documented pipeline we can extend than a demo that breaks on the second edge case. WHAT THE BUILD COVERS - Document ingestion and chunking into a vector store (Pinecone or pgvector; tell us which you'd pick and why) - A retrieval pipeline: query, retrieval, then a grounded answer with a source reference on every response - A lightweight, self-contained web chat interface. We're not attached to a specific framework, so propose what you'd actually ship - Honest fallback behaviour: when the answer isn't in the knowledge base, the bot says so. No invented answers STACK LangChain or LangGraph, Pinecone or pgvector, Python or Node.js. We're open to your recommendation as long as it's something you'd put in production, not a weekend prototype. TIMELINE 5 to 7 days from kickoff. BUDGET $500 fixed price, split across two milestones: pipeline first ($300), then interface and polish ($200). Happy to discuss the split. WHO WE'RE LOOKING FOR - You've built real retrieval pipelines, not just wrapped an LLM API call - You can walk us through your chunking strategy and embedding choice before writing code - You hand off documented, readable work. We need to be able to maintain and extend it after you're gone TO APPLY Skip the generic pitch. Send us two things: 1. A short paragraph on how you'd approach chunking and retrieval for a mixed document set like this 2. One or two examples of RAG work you've shipped, with a note on what was hard about each Applications without these will not be reviewed.