# Uplink

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Taraka Yarlagadda](https://cerebralvalley.ai/u/Tarak0910)
- **GitHub:** https://github.com/tarak-09/Uplink
- **Demo video:** https://youtu.be/ZDuev4sqAdQ
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2387

Uplink tells the people who keep community Wi-Fi running what broke, why, and what to check first, in English, German or any other language.

Freifunk München runs over 2,000 routers, all looked after by volunteers. When routers fail, the public map turns them red but never says why. It can't tell a power cut that took out a rooftop antenna, and every antenna behind it, from someone unplugging a router at home.

Uplink reads the same public map data every minute and links failures that share a cause: which nodes reached the internet through a node that failed first, which shared a neighbour, which booted in the same minute (so likely share a power supply), and which the operator named as a set. On one real day, 79 routers went quiet. Uplink found 6 outages among them, so it sends 6 alerts instead of 79 (92% less noise). It notices a failure about 7 minutes before the map does: it treats 3 minutes of silence as down, while the map waits about 10. It runs live on the Munich network today, sending alerts and all-clears.

Muse Spark writes every alert in the volunteer's language from facts the code has already settled, and a checker rejects any draft that invents a number. Volunteers can ask by voice note, transcribed by Muse Voice Transcribe; I tested it in Telugu, my mother tongue. An outage page (uplink analyze --html) draws each outage as a picture: what stayed online, what failed and when, and what to check first.

Built for my world: last-mile Wi-Fi with no operations team. India's PM-WANI program has over 4 lakh public hotspots, many run by small shop owners, and the regulator TRAI has proposed outside companies to manage them. Uplink automates part of that job: knowing what broke, and why. Pure Python standard library, tested in CI on Linux and Windows, with the real Munich data in the tests.

## More from Meta Global AI Developer Hackathon

- [RecallLoop Team](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2384)
- [BRACU ABYSS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2385)
- [Countersign](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2386)
- [StatWise Innovators](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2388)
- [ReproTrace AI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2389)
- [SmearLens](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2390)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2387. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt