I built the Customer-Complaint-Management-System repository on my GitHub handle @AnshuKumari09. It is a Python backend with a React/Redux Toolkit frontend for AI‑powered intake and triage of pharma (API/FDF) customer complaints.
Repository name: Customer-Complaint-Management-System
Language: Python
Started 2026-09, last updated 2026-09
/api/complaints/extract-text -> pulls raw text out of uploaded PDF/txt /api/complaints/analyze -> runs the LangGraph pipeline /api/complaints (GET/POST) -> persists to Postgres/SQLite via SQLAlchemy /api/complaints/upload-image -> OCR route using Groq vision model qwen/qwen3.8-27b
/api/complaints/upload-image uses app/ocr_service.py, which sends the uploaded photo/scan directly to Groq's vision‑capable model (qwen/qwen3.8-27b)
LangGraph models the QMS complaint‑triage process as an explicit graph of steps, matching how a real quality reviewer works through a complaint (extract -> check completeness -> assess risk -> find root cause -> propose CAPA -> summarize). Each node is independently testable and swappable.
extract_info -> completeness_check -> risk_classification -> root_cause_suggestion -> capa_recommendation -> summary
I chose LangGraph over a single prompt to improve auditability and to mirror the real QMS complaint‑triage workflow. The deterministic completeness_check node guarantees reproducible mandatory‑field verification, and the /analyze endpoint stays human‑in‑the‑loop by returning AI suggestions without persisting them.
For local development I run the backend with uvicorn app.main:app --reload --port 8000 and the frontend with npm run dev, accessing the UI at http://localhost:5173.