Electify - Retrieval-Augmented Generation for Voter Information in the 2024 European Election
summary
Electify is an open source RAG pipeline that converts 70+ page German party manifestos into question‑answer pairs by PDF parsing with pypdf, 1,000‑character text chunks, embeddings stored in a Chroma database, and retrieval of the top 10 slices with maximal marginal relevance reranking. Responses run on Streamlit with Cloud Run/Azure hosting, use strict prompt refusals (about 15% refusal) to reach ~99% faithfulness, benchmark at 91% accuracy against Wahl‑O‑Mat, and served ~6,000 users for minimal cost.
AI-written summary · about
chapters (7)
- 0:00Electify and voter info challenges
- 4:01Limitations of voter tools
- 8:40How retrieval-augmented generation works in Electify
- 13:41Prompt trade-offs
- 17:35Live demo of Electify
- 20:28Key takeaways and open source
- 21:10Q&A on costs and bias
related talks
49mGenerating the Future
28mWhat the RAG? Demystifying Retrieval Augemented Generation
25mAdvanced RAG Systems: Scaling Retrieval-Augmented Generation in Production
32mRevolutionizing Automation: Retrieval-Augmentation with LLMs
39mAgentic RAG Platform for AI-Assisted Workflows
23m"Deliver AI at Lightspeed: Streamlined approach to AI strategy"
31mRetrieval Augmented Generation
44mThe RAG Really Ties the App Together