SID

Description Python LangGraph agent that turns messy procurement intake (files/text) into a structured Fairmarkit event payload. Problem statement Procurement requests arrive in inconsistent formats and channels, so teams spend manual effort deciding if a request is processable, extracting fields, choosing categories, and mapping into API-ready schema. Solution This repo automates that pipeline end-to-end: route/qualify intake, normalize content, classify event intent, extract event fields, enrich with category retrieval, and output Fairmarkit-compatible JSON. How it works A staged graph runs nodes in sequence with early-stop gates: intake_router -> content_normalizer -> event_intent_classifier -> event_field_extractor -> enriched_event_fields (with category search tool + FAISS index) -> fairmarkit_payload. If input is not processable or not an event-creation request, it exits through result_builder with a structured reason. Why it wins Deterministic workflow + strict JSON contracts + retrieval-backed category selection reduces hallucinations, improves consistency, and converts unstructured intake into actionable payloads with minimal human triage. GIT https://gitlab.fmdev.io/sergei.sergeenko/hack27_sid