Show HN: GroundCite-An open-source package to fix broken Gemini API citations (github.com)

GroundCite is an open‑source toolkit designed to fix broken, irrelevant or missing citations produced by Google’s Gemini (and OpenAI) outputs. It provides an end‑to‑end pipeline—CLI, FastAPI service, and Python library—that performs intelligent grounding search (with site include/exclude filtering), AI‑powered citation validation, structured JSON parsing with custom schemas, and multi‑agent graph orchestration (built on LangGraph). The project also includes retry/error handling, comprehensive logging, token usage and correlation tracking, and configurable settings/validation to make citation generation reproducible and auditable. For the AI/ML community this matters because citation and grounding quality are central to trust, reproducibility, and safe deployment of LLM systems (especially RAG and automated analysis workflows). GroundCite’s technical stack (Python 3.12+, Pydantic, FastAPI, Click, Rich) supports Gemini and OpenAI providers, gives automated validation of sources, and emits structured outputs so downstream systems can parse and verify provenance. It’s developer‑friendly (pip installable, CLI commands like gemini-groundcite analyze, and a FastAPI endpoint for HTTP integration), MIT‑licensed, and aimed at reducing hallucinations, improving audit trails, and lowering debugging/cost overhead for production LLM applications.