Track 2: Visarjan Ghat Watch

Before Durga visarjan on 20 October 2026, the Bhopal Municipal Corporation officer who plans the ghats has to decide where to send extra cleaning, aeration and water testing. The evidence exists: MPPCB tests the water at every ghat before, during and one week after each Ganesh and Durga immersion. But it is buried in 90–120-page PDF reports that nobody compares with the legal limits or with earlier years. Those reports show dissolved oxygen at 0 mg/l four times at Bairagarh kund, 42 of 116 oxygen tests below the bathing limit, and 100 of 112 BOD tests above it. With the NGT ban on lake immersion and new BMC ghats coming, the officer is still deciding from memory.

Visarjan Ghat Watch ranks Bhopal's 7 immersion ghats by water-quality risk, using all 1,524 MPPCB readings from 2023–2025 checked against the CPCB bathing-water limits (class B).

Ranking with reasons: each ghat gets a plain-language reason in English and Hindi. Every number opens the original MPPCB PDF at the exact page. Missing data stays missing: untested readings show as "not sampled", never as clean. Gaps become "extra testing" suggestions. Action plan: the officer confirms aeration, cleaning or testing for each ghat and saves the plan. Brief: the officer shares a formatted Hindi + English PDF brief on WhatsApp. Day-one tools: a test schedule for 18, 20 and 27 October with a calendar file, a printable field sheet for each ghat's team, and a screen to enter this year's results so the ranking updates. Honest about limits: the tool states what the data can't show (no coliform tests, units not printed, an uncertain site name) and makes no health claims. Checked: we matched 29 readings against the source PDFs, and all 29 agree.

We built the project with Claude Code from start to finish, with the team setting direction, design and priorities at each step. Claude Code profiled the dataset and confirmed the guide's headline figures. It designed the scoring method with us, then wrote the backend, frontend and tests. It also wrote the Hindi text, which the team is reviewing. It checked readings against the PDFs with pdftotext, which caught and explained column-layout quirks in the reports. Along the way it flagged honesty issues for us to decide: missing data must never look clean, the brief makes no health claims, approximate map pins are labelled, the data doesn't come from BMC, and using the state emblem needs permission. The team supplied the design mockups and photos, chose features based on the judging criteria, and tested the app.