Jev Chat Assistant
An Android conversation copilot that uses Jev to interpret chats and rank reply candidates without sending for the user.
What Jev decides
The likely intent, conversation risk, reply need, best next action, unresolved tension, and which of three candidate replies fits best.
The app reads the visible conversation through Android accessibility or local OCR, asks Jev a fixed set of bounded questions, uses a separate generative model to draft three replies, and asks Jev to rank them before the user chooses whether to fill one into the chat box.
- Jev role
- Interpret the current conversation through bounded judgments and rank three already-drafted reply candidates.
- Input
- The relationship, up to ten recent visible messages, and optional local contact, knowledge-base, and older-chat context.
- Output
- Choice distributions, Score output, Noul probabilities, and a Choice ranking over three reply candidates.
- Decision frequency
- When the user or automatic analysis triggers on a supported chat screen, followed by one ranking call after replies are drafted.
- Fallback
- Decision errors are surfaced without sending anything; the user keeps control of the chat input and can reply manually.
How it works
Overview
Jev Chat Assistant is an Android overlay for conversations in apps including WeChat, QQ, X, and Lark. Platform adapters turn the visible chat into a small message snapshot; optional local contact notes and chat history can add context. Jev then evaluates seven typed questions covering subtext, intent, conversational risk, whether a substantive reply is warranted, the best action, what the other person needs, and whether tension has been resolved.
A separate OpenAI-compatible model drafts exactly three Chinese reply candidates, and Jev ranks those candidates. The app can copy or fill a selected reply but does not press send. The repository documents limits including platform-version sensitivity, incomplete validation for some chat modes and languages, Android background restrictions, OCR errors, and the need to calibrate Jev for Chinese conversations.
- Language
- Kotlin
- License
- MIT
- Author
- Jev Chat contributors
Sources
Related projects
Distill
Route coding-agent work, reasoning effort, and context retention through bounded Jev decisions.
What Jev decides
Which configured model and effort should handle a call, what evidence deserves attention, whether another attempt is useful, and what context should be retained.
jev-model-router
Route Claude Code model tiers and reasoning effort with one parallel Jev classification.
What Jev decides
Which fast, balanced, or deep tier fits the work, how much reasoning it needs, and whether production risk requires escalation.
typesafe-computer-use
A native macOS computer-use loop built from local OCR, accessibility data, and bounded Jev actions.
What Jev decides
Which computer action, visible item, site, or off-screen accessibility control to use next.