jev-triage
Creator inbox sorter: Jev answers five questions per comment or DM and code decides to notify, draft, like, hide, archive or escalate.
Overview
jev-triage (peterp22) is a sorting desk for a content creator's comments and DMs. Every message is read by Jev in one API call with five questions: what the person wants, how much a personal reply is worth, and whether the message is spam, abusive or mentions a health issue. Plain code rules then pick an action: notify now for payment or brand collaboration, draft a reply for real questions, like for praise, hide spam or abuse above a high bar, archive, or send to human review.
Human review covers low confidence, health or injury mentions, criticism, and leads or collaborations that also look like spam, because real sponsor outreach often reads like spam and hiding one costs money; health-related messages never get an automated reply. The README estimates about $0.00003 per message, cheap enough to read every comment and save the LLM for the few worth replying to. A confidence slider re-routes the whole queue in the browser from the existing probabilities without new API calls.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What does this person want? | — |
| Score | How much is a personal reply worth? | — |
| Yes / no | Is this message spam / abusive / mentioning a health issue? | — |
How it works
- Each message gets one Jev call with five questions
- Routing rules map answers to notify, draft, like, hide, archive or human review
- Safety rules force health mentions and spammy leads to humans
- Slider re-routes the queue locally without new calls
Strengths
- README estimates about $0.00003 per message
- Deliberate safety rules for health and lead messages
- Instant re-routing from cached probabilities
Caveats
- Demo app; not connected to real platforms by default
- No accuracy evaluation reported
Getting started
npm install && npm run dev, open http://localhost:3100 and paste a TypeSafe key.At a glance
- Type
- Project
- Status
- prototype
- Language
- TypeScript
- Runtime
- Node web app
- Has code
- yes
Use cases and integrations
How curated lists describe it
Confidence-gated creator inbox triage on TypeSafe.
wh000wh000/awesome-jev-live
Frequently asked questions
What does each message cost?
About $0.00003 per message, according to the README.
Will it auto-reply to health messages?
No. Messages mentioning injury or health always go to human review.
Why not hide spammy sponsor messages?
Real sponsor outreach often looks like spam, so those go to a human instead.
Related projects
Google Apps Script that labels, archives or trashes Gmail conversations based on Jev answers to your plain-English yes/no questions.
ProjectbetaJavaScriptChrome/Brave extension that uses Jev to classify, label and safely bulk-clean a Gmail inbox with undo.
PluginbetaJavaScriptDry-run-first Proton Mail Bridge triage on macOS where Jev answers whether each message is spam or phishing.
ProjectbetaPythonGoogle Apps Script Gmail classifier that applies your own plain-English labels with Jev and archives only high-confidence unwanted mail.
ServicebetaJavaScriptRead-only Chrome extension that badges each Gmail message with Jev's category, priority, spam % and reply %.
Browser extensionprototypeJavaScriptRead-only, local Gmail triage that sorts each message into five trays with an urgency score using Jev via Vercel AI Gateway.
ServicebetaTypeScript