Privacy note: Documents are stored on the API host. When configured with hosted providers (e.g. OpenAI), document chunks and tribunal prompts are sent to third-party APIs. Use Ollama or local embeddings for fully private inference.

The Hallucination Tribunal

A RAG-powered courtroom for AI answers. The Witness generates grounded responses, the Prosecutor challenges claims, and the Judge delivers a transparent verdict with citations.

Why it matters

Enterprise teams need answers they can trust. Claim-level verification catches hallucinations before they reach users.

Architecture

Hybrid retrieval, structured agent pipeline, local-first privacy with optional hosted LLM providers.

Courtroom metaphor

Witness, Prosecutor, and Judge roles make AI verification explainable to hiring managers and compliance teams.