ChatGPT: Chicago City RAG

This talk demonstrates a retrieval-augmented generation pipeline using vector search and prompt engineering to answer Chicago residents' questions about city services from official documents.

Overview

Building a conversational AI that answers Chicago residents’ questions about city services - inspired by Landon’s newsletter idea.

Live demo: Ask “How do I report a pothole?” or “What are parking rules on Michigan Ave?” and watch it pull answers from city documents and data.

Technical walkthrough:

RAG pipeline: PDFs → chunks → vector embeddings → retrieval
Pinecone for vector storage of city procedures/FAQs
Tool calling to external APIs (future: Chicago Open Data Portal)
Prompt engineering for helpful government tone

It will be live by demo day and in early beta!!

Links

Tech stack