JanDrishti — Indian Government Transparency Platform

A full-stack transparency platform covering every department of the Central Government and all 28 States + 8 Union Territories: leadership (ministers, secretaries), budgets (Union + all states), actual expenditure vs budget, scheme-wise utilisation, CAG audit findings, taxpayer vigilance red flags, and a live PIB press-release feed.

Stack

• Backend: Python + FastAPI (backend/main.py)
  • 15 JSON API endpoints under /api/*
  • POST /api/updates/refresh runs the PIB pipeline on demand
  • In-process scheduler (APScheduler) keeps the PIB feed fresh — no external cron needed
  • Serves the frontend at / — one server, one command
• Data: backend/data/*.json (snapshot as of 11 Sep 2026, sourced from pmindia.gov.in, indiabudget.gov.in, cag.gov.in, cga.nic.in, sansad.in, pib.gov.in, DoPT, RBI, PRS, ECI/ADR)
• Pipeline: backend/pipelines/fetch_pib.py — pulls PIB's official RSS feed (with an AllRelease.aspx fallback), auto-tags ministries (English + Hindi), writes pib_updates.json
• Frontend: frontend/index.html — single dynamic page, fetches all data from /api/all at load; dark/light theme toggle (persisted); live "Fetch latest releases" button with stale-feed error state

Run it (local)

# 1. Install dependencies (once)
pip install -r requirements.txt

# 2. Start the server
cd backend
uvicorn main:app --reload

# 3. Open http://localhost:8000

The PIB feed refreshes itself on a timer while the server runs (see PIB_REFRESH_INTERVAL_MINUTES below), and the "Fetch latest releases now" button in the Latest Updates section triggers an immediate refresh.

Configuration (environment variables)

Variable	Default	Purpose
DEBUG	off	1/true enables dev mode: CORS allows all origins
ALLOWED_ORIGINS	(empty)	Comma-separated CORS allowlist, e.g. https://jandrishti.in,https://www.jandrishti.in. If unset and not in DEBUG, only same-origin requests are allowed (the frontend is served by this app)
HOST	0.0.0.0	Bind address
PORT	8000	Bind port
PIB_REFRESH_INTERVAL_MINUTES	60	Fetch PIB releases every N minutes via the in-process scheduler. 0 disables
PIB_INITIAL_REFRESH_DELAY_SECONDS	30	Delay before the first scheduled fetch after startup

The PIB pipeline and bot-detection

pib.gov.in intermittently rejects scripted clients with 403 Forbidden (bot detection). The pipeline is built for this:

1. Tries the official RSS feed first, using a browser-like session (full header set, keep-alive).
2. Retries each source 3 times with exponential backoff (2s, 4s, …).
3. Falls back to the official AllRelease.aspx HTML listing if the RSS is blocked.
4. If everything fails, it keeps serving the last-known-good pib_updates.json, records the error in pib_status.json, and the API responds with a structured ok: false + last_good timestamp — never a raw 502. The frontend shows a visible "stale feed" state with the last successful fetch time.

Pipeline health is exposed at GET /api/updates/status (last_success, last_attempt, last_error).

API quick reference

Endpoint	What it returns
GET /api/health	Service status
GET /api/meta	Dataset counts & as-of date
GET /api/positions	19 constitutional positions
GET /api/ministers	PM + 30 Cabinet + 5 MoS(IC) + 35 MoS
GET /api/ministries	52 ministries w/ departments, ministers, secretaries
GET /api/states	28 states w/ full cabinets
GET /api/uts	8 union territories
GET /api/union-budget	Ministry allocations, scheme BE→RE trails
GET /api/state-budgets	All 28 state budgets (FY2025-26)
GET /api/expenditure	Actuals, CGA monthly, scheme utilisation, CAG
GET /api/vigilance	15 taxpayer red flags
GET /api/sources	14-source data registry
GET /api/updates	Latest PIB press releases
GET /api/updates/status	Pipeline health (last success/error)
POST /api/updates/refresh	Run the PIB pipeline now (structured result)
GET /api/all	Everything in one call (used by the frontend)

Interactive API docs are auto-generated at http://localhost:8000/docs.

Tests

pip install -r requirements.txt -r requirements-dev.txt
python -m pytest

The suite (11 tests) covers every GET endpoint's status code and JSON shape, the frontend being served at /, and the PIB refresh path with mocked PIB fetches — success, 403 rejection, and crash cases — so CI never depends on pib.gov.in. Pipeline parsers (RSS, HTML listing) are unit-tested against local fixtures.

Deploying

Docker

docker build -t jandrishti .
docker run -p 8000:8000 \
  -e ALLOWED_ORIGINS="https://yourdomain.in" \
  -e PIB_REFRESH_INTERVAL_MINUTES=60 \
  jandrishti

The image is a two-stage build: dependencies install in a builder stage, and the runtime stage is slim, runs as a non-root user (uid 10001), and contains no build toolchain. Set -e PORT=... if your platform needs a different port.

Render / Railway / Heroku (Procfile)

The repo ships a Procfile, so PaaS platforms can deploy it directly:

1. Push this repo to GitHub.
2. Create a Web Service (Render) / app (Railway) from the repo — the platform detects the Procfile and runs uvicorn main:app --host 0.0.0.0 --port $PORT.
3. Set ALLOWED_ORIGINS to your site's domain in the platform's environment settings.

The in-process scheduler means a single web dyno/instance keeps the PIB feed fresh — no separate worker or external cron required. (If you scale to multiple instances, set PIB_REFRESH_INTERVAL_MINUTES=0 on all but one to avoid redundant fetches.)

Project structure

jandrishti/
├── Dockerfile                 # two-stage, non-root runtime
├── Procfile                   # PaaS entrypoint
├── .dockerignore / .gitignore
├── pytest.ini
├── requirements.txt
├── backend/
│   ├── main.py                # FastAPI app + scheduler + config
│   ├── data/                  # 13 JSON datasets (the platform's database)
│   ├── pipelines/
│   │   └── fetch_pib.py       # Hardened PIB pipeline (RSS + fallback)
│   └── tests/                 # pytest suite (no network dependency)
└── frontend/
    └── index.html            # Dynamic SPA (fetches from the API)

Data provenance

Every record in backend/data/ carries sources and verification fields. Leadership data is anchored to the official PMO Council of Ministers list (25.07.2026), DoPT secretary lists (01.01.2026–01.02.2026), and MHA AGMUT lists (16.02.2026). Budget figures come from Union Budget 2025-26 documents, RBI State Finances (Jan 2026 edition), and PRS analyses. Expenditure actuals are from CGA monthly accounts and CAG Appropriation Accounts (Report No. 6 of 2026).

Next pipelines (same pattern as fetch_pib.py)

1. CGA monthly accounts — makes "money spent so far" current every month
2. CAG report tracker — new audit findings as they're tabled (weekly scrape)
3. Budget-day processing — Feb 1 auto-ingest of the new Union Budget
4. Parliament questions — daily during sessions (sansad.in)

Caveats

• PIB's RSS feed currently serves Hindi-language releases; the platform displays them as-is with automatic ministry tagging.
• Office-holder data changes with reshuffles — re-verify before production use (see the "needs-verification" flags in the data files).
• The pre-populated datasets are a snapshot dated 11 September 2026.
• The Dockerfile has not been built in CI yet (no Docker available during development); it is kept structurally simple — if the build surfaces an issue, it will be in the two-stage copy paths, which mirror the Procfile layout.