Vinícius Raposo

Applied AI Reliability Engineer

AI systems you can leave on without babysitting.

Profile pin · Python

groundtruth

RAG application that cites every answer and refuses what its sources can't support: hybrid retrieval, versioned documents, human review.

331 API + 62 web tests; real-HTTP workflow and restart checks; checksum-verified offline evidence. Its held-out answer-coverage measurements missed their targets (it refuses more than it should, with zero unauthorized-source exposures), so no accuracy or answer-quality claim is made. Live PostgreSQL/Redis is not verified.

Language: Python Updated: 2026-08-15 Visibility: public
View on GitHub Back to catalog

Without corpus-grounded answers and explicit refusal, RAG systems hallucinate confidently—users get fluent wrong answers instead of citations or an honest refusal when the corpus does not support a claim.

Answers come from the corpus, or the system refuses.

Offline demo:

cd apps/web && npm run demo:ui

Walkthrough

Answer cites two corpus chunks with a visible retrieval trace.
Answer cites two corpus chunks with a visible retrieval trace. Sample data.
Question unsupported by the corpus gets an explicit refusal, not a guess.
Question unsupported by the corpus gets an explicit refusal, not a guess. Sample data.
The demo corpus and version history the answers draw from.
The demo corpus and version history the answers draw from. Sample data.

Why it matters

RAG application that cites every answer and refuses what its sources can't support: hybrid retrieval, versioned documents, human review.

Verified proof

331 API + 62 web tests; real-HTTP workflow and restart checks; checksum-verified offline evidence. Its held-out answer-coverage measurements missed their targets (it refuses more than it should, with zero unauthorized-source exposures), so no accuracy or answer-quality claim is made. Live PostgreSQL/Redis is not verified.

Repo facts

Default branch: main. Stars: 0. Forks: 0. Created: 2026-05-31.