Jeevan Rekha

Jeevan Rekha is a browser-based prototype for overflow flood helplines. When the phone lines are full, an AI call-taker collects a rescue request in the caller's language and sends human dispatchers a ranked queue on a live map. The demo shows a real Hindi call, a 24-call flood-night replay, and an English caller joining a group already reported. The agent reads every place back, confirms headcounts digit by digit ("30" as "three-zero"), asks for a landmark, and updates urgency as the water rises. While talking, it checks the live case board and asks the caller before joining an existing rescue request. Possible duplicates are flagged for a human instead of being guessed away. Dispatchers see the call transcripts, mapped cases, urgency and missing-information flags. They assign boats and mark rescues. Caller photos can be analysed with person-segmentation boxes, a headcount and scene notes. Merges and photo evidence cannot lower urgency or headcount. The application, call flow, case board, place matcher, evaluations, voice integration and map workflow were built during the hackathon. The current test suite passes 50 tests. Evaluation audio and the example rooftop photo are synthetic; accuracy on real distress calls is not established. This is a browser prototype, not a connected public emergency helpline. Live demo: https://sumerchoudhary78--jeevan-rekha-web.us-east.modal.run/?code=flood-11f855 Demo access code: flood-11f855 Judge walkthrough: Open the demo in Chrome, then open the caller and control room. Allow the microphone, use headphones, and speak after the tone. Location sharing is optional; disable it when reporting a rescue somewhere else. In the control room, "Simulate flood night" replays 24 recorded calls. No judge account is needed.