Experienced Developer for Production-Quality Rag Chatbot with Internal Knowledge Base

We are seeking an experienced developer to build a small but production-quality Retrieval-Augmented Generation (RAG) chatbot. This chatbot will be connected to our internal knowledge base, comprising various internal documents such as playbooks, templates, and reference materials. The primary goal is to enable accurate, grounded answers to user questions based on this content, avoiding generic AI responses or hallucinations.

Project Scope:

*  Document Ingestion and Vector Store Setup: The chosen freelancer will be responsible for ingesting and chunking our internal documents into a vector store. We are open to recommendations between Pinecone or pgvector for this component.
*  RAG Pipeline Development: Build a robust RAG pipeline that processes user queries, performs efficient retrieval from the knowledge base, and generates grounded answers with clear source references.
*  Simple Chat Interface: Develop a lightweight, self-contained, and easy-to-use web-based chat user interface for interacting with the chatbot. Your recommended approach for the UI is welcome.
*  Intelligent Fallback: Implement a sensible fallback mechanism for instances where the answer is not found within the knowledge base, ensuring no hallucinated responses are provided.

Technical Preferences:

We prefer development using LangChain or LangGraph. For backend development, Python or Node.js are preferred. We are open to your recommended stack, provided it is production-grade and ensures reliability and maintainability.

Timeline: The estimated timeline for this project is 5-7 days.

Budget: The budget for this fixed-price project is $500. We are open to discussing a milestone-based payment structure if preferred, for example, splitting payments between setup/pipeline development and interface/polish.

What We Are Looking For in a Freelancer:

*  Demonstrated real-world experience with rag pipeline implementation, beyond just direct llm api calls.
*  Ability to clearly articulate your proposed retrieval approach, including chunking strategies and embedding choices, before commencing work.
*  Commitment to a clean handoff, providing well-documented code rather than a black box solution.