This repo has code, data, and flows for Langflow for the livestream https://www.datastax.com/events/build-more-accurate-ai-apps.
In this livestream and code you'll
- Generate a ground truth dataset for question & answer over 35 Wikipedia articles from the SQUaD dataset.
- Load the data to Astra DB for vector search
- Build a simple agentic chatbot that uses the SQUaD data
- Create and run experiments with different embedding models, LLMs, and prompts
- Use Arize to track experiments and understand how changes affect accuracy