Community · E2 · artifact verified
Control Chrome by voice with typed commands
A Chrome MV3 extension that turns speech into browser actions: Jev routes each spoken command to open a site, search, click a link, fill a form field or go back, with typed answers instead of parsed free text; ships with tests, CI and a side-panel command log.
01 · Role in the system
What Jev does here
Spoken input is transcribed in the side panel and turned into candidate commands; Jev answers typed questions about what the user wants - a choice over the action set, plus yes/no checks - and the extension executes the winning action through Chrome's normal APIs: opening sites, running searches, clicking links, filling named form fields and navigating back. Because Jev returns calibrated answers rather than free text, no intent parser or string matching sits between the transcript and the action, and unmatched commands fail loudly instead of guessing. The same client library also carries score questions and maps yes/no questions to Nouls on the wire, and the repo keeps a CI workflow with unit tests over the question mapping and the action dispatch.
02 · Control boundary
Where Jev sits
Jev as the intent router for voice commands: transcript in, typed choice over a fixed action set plus yes/no guards, and Chrome APIs execute only the winning action - no free-text parsing in between.
Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.
03 · Known limits
What this evidence does not prove
- Chrome 116+ desktop; transcription and actions run locally, Jev calls need an API key.
- Command set is fixed to the supported actions; arbitrary page manipulation is out of scope.
- No hosted demo; install from source.
04 · Attribution
Public sources
This is a Community record: the project was published by a third-party community author.
- dgr8akki ↗Community · github · public · checked 2026-09-28