NewJev Chat Jarvis
A phone conversation copilot that reads visible chat and offers suggested replies while the user stays in control.
Added to Jevfast
NewHow it uses Jev
The Android assistant reads visible chat content through accessibility or OCR, then sends conversation context and matched local contact/history notes to Jev for intent, risk, and reply-timing judgments. A separate generation model drafts replies, which Jev can rank.
What Jev decides
Conversation judgment requestExample answers · not a recorded Jev response · Source ↗
Question 1 · literal question
YOUR APP
INSTRUCTION
Is the other person’s latest message purely literal, without subtext?
STATE
The state includes a Chinese chat snapshot and optional relationship context. These are source-defined judgment questions, not a recorded conversation.
JEV · NOUL
YesNo
Question 2 · true intent
YOUR APP
INSTRUCTION
What is the other person’s intent in the latest message?
STATE
The state includes a Chinese chat snapshot and optional relationship context. These are source-defined judgment questions, not a recorded conversation.
JEV · CHOICE
- confirm_you_care
- vent_anger
- request_action
- seek_explanation
- casual_chat
- close_topic
App workflow
- Read the visible conversation
Collect displayed messages through accessibility adapters, with OCR fallback for supported interfaces.
- Build contextual state
Combine the exchange with matching contact notes and conversation history when available.
- Judge before drafting
Ask Jev for intent, risk, and whether to respond; a generation model drafts candidate replies and Jev ranks them.
- Present user-controlled suggestions
Show analysis and candidates in the overlay; the user may fill a reply into the input and decides whether to send.
Why it is interesting
It separates interpretation from text generation and keeps sending under user control: suggestions can be filled into the chat input but are never sent automatically. Local knowledge notes provide relevant relationship context to the request.