Jev 101
Start Guides
Search
中文 EN
Agents ✦ 72outlineNot a production recipe — calibrate thresholds; a11y/OCR setup varies by device

Chat reply judgment then draft

Who: Messaging power users. Steps: (1) Capture chat via OCR/a11y. (2) Closed Choice/Score/Noul on intent, danger, reply-now. (3) Code ranks; generative model drafts text only. (4) Human sends. Expected effect: judgment before generation; no auto-send.

Source: https://github.com/jev-chat/jev-chat-jarvis ↗ Docs / source ↗
1

Treat this as an outline — adapt state and questions to your data.

2

Implement in code — call System One / Jev; compose answers yourself.

3

Gate on confidence — act, confirm, or escalate before side effects.

Outline sketch

Prompt 
# Chat reply judgment then draft
State: recent chat text captured via OCR or accessibility (desensitize PII).
Ask: Choice/Score/Noul on intent, danger, and whether to reply now.
Then: rank options in code; let a generative model draft candidate text only; human sends.

Needs access to: typesafe-sdkandroid-a11y

Who it's for

Mobile / messaging power users

Steps / how it's set up

Steps

  1. Capture chat via OCR / accessibility.
  2. Closed Choice / Score / Noul on intent, danger, reply-now.
  3. Code ranks; generative model drafts text only.
  4. Human sends — never auto-send.

Sources

Distinct from social-draft-vibe-check (X draft gate). Metrics and product claims are author-reported unless you measure them yourself. Calibrate on your data.

Expected effect

Judgment before generation; human remains the sender

Unofficial outline for learning. Paraphrased from public docs and tutorials — not a production recipe. Review sources before you automate anything.