What it is
Android co-pilot that reads chat via OCR/accessibility, then asks Jev for intent, risk, and reply-now judgments before any draft text is shown.
What it does: Closed judgments drive whether to reply; a generative model only produces candidate text; the human always sends.
Why it belongs in the showcase: High-star public app with a clear judgment-before-generation split (distinct from social draft vibe checks).
Caveats: Needs a11y/OCR permissions; README says it never auto-sends; provider choice must be verified in your build.
Evidence
Public MIT README + https://chatjevs.com
Caveats
- Accessibility/OCR permissions
- Never auto-sends — product claim from README
- Supports multiple Jev providers; verify which you configure
Unofficial case page. Stars / list position observed for heat; do not treat heat as a quality SLA. Review the upstream README before you depend on it.