Parchi

Government paperwork, in plain language.

Upload a government notice, scheme, form or letter — as a PDF or just a photo of the page — and get back:

• a plain-language summary and an "explain it like I'm 10" version
• an eligibility checklist: who can apply, documents needed, conditions, exclusions
• every key date, pulled out and dated
• any tables in the document, rebuilt as real tables
• a chat that answers your questions using only what's in the document
• all of it in English, हिंदी, ಕನ್ನಡ, தமிழ், తెలుగు, मराठी or বাংলা

---

Stack

Layer	Tech
Frontend	Next.js 16 (App Router), TypeScript, Tailwind CSS
Backend	Python, FastAPI, Pydantic — REST + WebSockets
Data / AI	pandas, NumPy, scikit-learn (LSA retrieval), Groq LLM (text + vision), RAG
Database	PostgreSQL + pgvector
Auth	JWT access + rotating refresh tokens
Infra	Docker + Docker Compose, GitHub Actions
Tests	Pytest + FastAPI TestClient

frontend (Next.js)  ──REST + WebSocket──▶  backend (FastAPI)  ──▶  PostgreSQL + pgvector
                                                  ▲                        │
                                          NOTIFY progress          FOR UPDATE SKIP LOCKED
                                                  │                        ▼
                                            worker (python -m app.worker) runs the pipeline:
                                            extract → simplify → embed → tables

There is no Node backend and no Redis — the job queue is a Postgres table, the worker is a Python process, and multimodal extraction (photos of documents) goes through Groq's vision model with Tesseract OCR as an offline fallback.

Repo layout

path	what
frontend/	Next.js app — see frontend/explanation.md
backend/	FastAPI API + worker + document pipeline — see backend/explanation.md
database/schema.sql	the whole schema — see database/explanation.md
samples/	a sample notice (PDF + image) for testing
docker-compose.yml	postgres + api + worker + web
commands.md	every command to run, develop, test and ship

Quick start

cp backend/.env.example backend/.env
# edit backend/.env:  GROQ_API_KEY=gsk_...   (free, no card: https://console.groq.com/keys)
#                     JWT_SECRET=<64 hex chars>   python -c "import secrets; print(secrets.token_hex(32))"

docker compose up --build

• Web → http://localhost:3000
• API docs → http://localhost:8000/docs

Then sign up, upload samples/sample-notice.pdf, and watch the pipeline run live.

Full command reference — local dev per service, tests, lint, database, CI — is in commands.md. The full implementation write-up — process model, data model, every pipeline stage, auth, live progress, trade-offs — is in IMPLEMENTATION.md.

Notes before deploying anywhere

• Rotate JWT_SECRET and GROQ_API_KEY; never ship dev values.
• Redact PII (Aadhaar / PAN) before sending document text to any LLM.
• Put the API behind TLS and lock CORS_ORIGINS to your real origin.