Ask My Docs — RAG-Based AI Chat Agent

Project Title: Ask My Docs — RAG-Based AI Chat Agent Full case study description:

The problem: Most AI chatbots hallucinate — they answer confidently even when they don't actually know something, which makes them risky for real business use (FAQs, internal docs, support content).

What I built: A RAG-based (Retrieval-Augmented Generation) chat agent that answers questions strictly from content you provide — paste in a FAQ, policy doc, or knowledge base, and it becomes the only source the agent is allowed to use. If the answer isn't in the content, it says so clearly instead of guessing.

How it works: User-provided content is passed to Llama 3.3 (via Groq) alongside each question, with a system prompt that constrains the model to only use that content and cite the exact passage it drew from. Answers include an expandable "Exact Source Passage" so users can verify every response against the original text.

Why it's different: Most demo chatbots skip the hard part — knowing when not to answer. This one is built around that constraint, which is exactly what makes it usable for real business content instead of just a tech demo.

Tech stack tags: Groq, Llama 3.3, RAG, AI Agents, Python/Backend, Replit

Live link: Live demo available on request