JalCheck

Find out what's contaminating your area's groundwater — arsenic, fluoride, nitrate, uranium — what it does to your body, and how to actually remove it.

India runs the world's largest groundwater-quality monitoring effort. The Central Ground Water Board and the Jal Shakti ministry know, district by district, where the water carries arsenic (West Bengal, Assam, Bihar), fluoride (Rajasthan, Telangana, Karnataka), nitrate (56% of districts exceed the safe limit), and uranium (Rajasthan, Punjab). Tens of millions of people drink it.

But that knowledge sits in survey portals and PDF annexures no resident will ever open — and the contaminants that matter most are invisible: you can't see, smell or taste arsenic or fluoride, and boiling doesn't remove them (it can even concentrate them). JalCheck turns the surveys into a plain question: "What's in my water, and what do I do about it?"

How it works

The data — real government contamination surveys

Two official sources, normalized into a structured dataset:

• Affected-habitation counts from the Ministry of Drinking Water & Sanitation's IMIS — the real per-state number of habitations and population affected by arsenic and fluoride (e.g. West Bengal: 10,846 arsenic-affected habitations, ~11.8 million people; Rajasthan: 6,087 fluoride-affected habitations, ~4.2 million people).
• The CGWB Annual Ground Water Quality Report (2024) for nitrate, uranium, iron and salinity — which states and districts exceed limits, and whether the source is geogenic or from fertiliser/industry.

Each of the 7 contaminants carries its IS 10500 safe limit, health effects, who's most vulnerable, and — the genuinely useful part — the correct removal method and whether boiling helps. Every state→contaminant record keeps its source citation, and a seed-time validator rejects any state that references an unknown contaminant.

The risk engine

backend/app/engine.py turns a state (and optional district) into a ranked, plain-language report. The opinionated parts:

• District awareness. Name a district that appears in a contaminant's affected list and that risk is flagged "in your district" and floated to the top — local specificity beats a state average. Nalgonda, Telangana → fluoride, flagged for your district.
• Boiling myth-busting. If a serious contaminant resists boiling (arsenic, fluoride, nitrate, uranium all do), the report leads with an explicit warning — because "just boil it" is the dangerous, widespread default.
• Actionable removal. It names the right fix: RO for the dissolved geogenic set, activated alumina as the cheaper fluoride-only option, aeration for iron — not generic "use a filter."

Example. A user in Rajasthan:

overall: SEVERE — Fluoride, Nitrate, Uranium
⚠ Boiling will NOT make this water safe — it does not remove Fluoride, Nitrate, Uranium.
💧 A reverse-osmosis (RO) unit removes the dissolved contaminants flagged here.
  Fluoride  · severe · 6,087 affected habitations (~4.2M people)
            dental & skeletal fluorosis · most at risk: children, all ages
  Nitrate   · severe · blue-baby syndrome in infants · most at risk: infants under 6 months
  Uranium   · severe · kidney toxicity · 42% of national uranium-exceeding samples

API

GET /api/states                 all states with worst-severity + contaminant list
GET /api/check?state=&district=  ranked risk report (district optional)
GET /api/contaminants           the reference table (limits, health, removal)
GET /api/stats                  dataset summary
GET /api/meta · /api/health

Consistent {"error": {code, message}} envelope; nothing about the query is stored.

Running it

docker compose up --build
# frontend → http://localhost:3200   API → http://localhost:8200/docs

Local dev:

cd backend
python3 -m venv .venv && .venv/bin/pip install -r requirements-dev.txt
DATABASE_URL=sqlite:///dev.db .venv/bin/python -m scripts.seed
DATABASE_URL=sqlite:///dev.db .venv/bin/uvicorn app.main:app --reload --port 8200
cd frontend && npm install && npm run dev

Tests

cd backend && .venv/bin/python -m pytest    # 21 tests

Tests assert the real IMIS figures load correctly, pin the engine's district-matching and severity ranking, and verify the boiling warning and RO recommendation fire for the right contaminants.

Data sourcing & verification

• Every state→contaminant record cites its source (IMIS habitation data or CGWB 2024).
• The two sources measure different things and JalCheck keeps them distinct: IMIS counts rural habitations not yet provided safe water (so a 0 doesn't mean a state is geogenically clean), while CGWB measures actual groundwater concentrations. Severity reflects whichever source applies, labelled.
• The UI repeats everywhere: this is an area-level guide, not a test of your specific tap — test your own source at a certified lab.

Limitations

• Resolution is state, with notable districts — not your street. Contamination varies within a district; only a lab test of your source is definitive.
• It's a snapshot of national reporting; a production build would sync CGWB's annual releases and add habitation-level WQMIS data.
• IMIS affected-habitation figures are from the last fully published cycle; concentrations shift with monsoon and extraction.

What I'd build next

1. Pincode / habitation resolution by ingesting the granular JJM WQMIS test records.
2. Map view — choropleth of India by contaminant, drill down to district.
3. "Test my water" routing — nearest government-empanelled water-testing lab and what to ask for.
4. Filter buying guide — match the certified filter spec (e.g. NSF/IS 16240 for arsenic) to the contaminants found.