Run offline AI commentary on a weak phone by borrowing a laptop's compute over a P2P link.
Study a working example of encrypted peer to peer inference offload with automatic failover.
Demo or judge the project as a hackathon submission for the Tether Developers Cup.
| edycutjong/gaffer | 00kaku/gallery-slider-block | 0xkinno/vellum | |
|---|---|---|---|
| Stars | 0 | — | 0 |
| Language | JavaScript | JavaScript | JavaScript |
| Last pushed | — | 2021-05-19 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | moderate | easy | hard |
| Complexity | 4/5 | 2/5 | 4/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Full AI commentary requires installing the @qvac/sdk and downloading a model, otherwise it runs with a disclosed simulated engine.
Gaffer is a hackathon project that generates spoken football match commentary entirely offline, without relying on any cloud service or API key. The idea it tackles is that expert commentary usually stops working the moment a phone loses its internet connection, and the fans who most need a simplified explanation of the game often have the weakest phones. Gaffer's answer is to run an AI model directly on a person's own device, and when that device is too weak to keep up, it borrows spare computing power from a nearby laptop over a direct, encrypted peer to peer connection, with no server in between. The project was built for a hackathon called the Tether Developers Cup, in the track combining two technologies: QVAC, which runs AI models on device, and Pear, a peer to peer networking stack built on Hyperswarm and Hyperdrive. In practice, a weak client device streams commentary at only a few tokens per second on its own, but once a laptop peer joins the same match session, that speed jumps dramatically because the laptop takes over the heavy computation while the phone still displays the result. If the laptop disconnects mid sentence, the phone finishes that sentence locally without losing its place. The README is notably transparent about what parts of the demo are real versus simulated: the peer to peer networking, encryption, and failover logic are described as genuinely real and covered by automated tests, including tests that run the actual peer to peer stack rather than mocks. The AI text generation and voice synthesis are real only when the QVAC software development kit is installed, otherwise the project falls back to a clearly labeled fake text generator so the networking and offload behavior can still be demonstrated on any machine. The underlying football match events themselves come from a simulated, seeded data feed rather than a licensed real data source. Gaffer is written in JavaScript, requires Node.js version 20 or newer, includes 220 automated tests, and is released under the Apache 2.0 license.
A hackathon project that generates offline football commentary by borrowing spare computing power from a nearby laptop over encrypted peer to peer networking.
Mainly JavaScript. The stack also includes JavaScript, Node.js, QVAC.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice and any stated changes.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.