Show HN: I built a deterministic check for fabricated quotes in LLM output
Category: library
Tags: llm, verification, rag, hallucination-detection, python
Score: 6.8/10 (Innovation: 7, Technical: 6, Documentation: 7, Utility: 7)
verbatimeter provides deterministic, lexical verification of how much LLM output is directly copied from a source text, supporting both verbatim reuse and quotation checking for RAG pipelines. Its novel approach of enforcing contiguous matching (n-gram overlap) and offering a decorator-based API makes it a practical tool for grounding verification and hallucination detection. The project fills a known gap for lightweight, offline, and judge-free factuality baselines in production AI systems.
Target audience: backend devs, ai engineers, data engineers
Repository: https://github.com/pierreolivierbonin/verbatimeter · Python · MIT · 2 stars
View on Hacker News