An open-source AI knowledge base for North American wholesale electricity market governing documents.
Live: rtointelligence.ai
rto intelligence answers questions about RTO governing documents in plain language. Every answer is cited — the source document, section, and excerpt are returned alongside the response so anyone can verify what they're reading.
The corpus covers three RTOs:
| RTO | Documents |
|---|---|
| MISO | Tariff, BPMs, MS-OPs, attachments, schedules |
| PJM | OATT, Operating Agreement, RAA, and 60+ more |
| SPP | OATT Parts I–IX, attachments, IMP, rate schedules |
584 governing documents. 6,882 indexed source excerpts.
- Ingestion — governing documents parsed, chunked, and embedded with OpenAI
text-embedding-3-small - Retrieval — hybrid search: HNSW vector similarity + full-text search, merged and re-ranked with authority tier boosts
- Generation — Claude Haiku produces a cited answer from the top chunks
- Verification — a groundedness judge scores each answer against its sources before delivery
- Frontend: Next.js 15, Tailwind CSS
- Database: Supabase (PostgreSQL + pgvector)
- Search: Hybrid HNSW + FTS with authority tier re-ranking
- Generation: Claude Haiku (Anthropic)
- Hosting: Vercel
Wholesale electricity market rules are public. They are also dense, fragmented across hundreds of documents, and difficult to navigate without years of institutional context. rto intelligence is built on the premise that this knowledge should be accessible to anyone — market participants, engineers, researchers, students, and regulators alike.
Everything here is open source. The corpus, the pipeline, and the methodology are published so the work can be verified, replicated, and improved.