Skip to content
← The showcaseComplete free read
Creator report with public repository

Draft replies beside the conversation, but leave sending to the person

An Android conversation copilot reads visible chat context, ranks reply drafts, and fills the chosen text without pressing send.

Editorial workflow map
  1. Read visible messages
  2. Rank candidate replies
  3. Fill but do not send
View the author’s original post (opens in a new tab)
Our interpretation · not the author’s screenshot or a reproduced result

The consequential gate stays human

The repository separates analysis, drafting, ranking, and text insertion. It explicitly keeps the send action—and payment-related actions—outside automation.

Supporting documentation (opens in a new tab)

How the stack is divided

Android accessibility or OCR supplies visible context; Jev judges intent, risk, and candidate fit; a separate model drafts three replies. Local adapters handle each chat application.

What to check before trying it

Accessibility and screenshot permissions expose sensitive conversations. The repository lists OCR, language, app-version, and background-process limits; no independent privacy or device test was run here.

Use the same checks on another demo.

Read the complete free evaluation guide →