Jev Mail Classifier
Turn plain-language inbox categories into thresholded tag, move, flag, read-state, and webhook actions.
What Jev decides
The independent probability that each category applies to the current message.
Each configured category becomes an independent Jev yes-or-no question for one email; local policy compares every probability with its threshold and runs the actions attached to matching categories.
- Jev role
- Estimate whether an email belongs to every configured plain-language category.
- Input
- The email subject and body plus the configured descriptions of all categories.
- Output
- One Noul probability per configured category.
- Decision frequency
- One Jev request for each unprocessed email.
- Fallback
- Use dry-run for inspection, cap each run, and leave actions to local thresholds; provider or IMAP errors are surfaced rather than converted into category matches.
How it works
Overview
Jev Mail Classifier connects to an IMAP mailbox, reads unprocessed messages, and classifies each subject and body against user-defined categories. One Jev request contains one independent Noul per category, so a message may match more than one. Code then applies each category threshold and can tag, move, flag, change read state, or send a bounded webhook payload.
A private IMAP keyword marks processed messages, and a per-run limit protects the first run from consuming an unexpected backlog. Dry-run mode shows proposed actions without touching the mailbox. The repository reports that its OpenRouter adapter was live-tested while the direct TypeSafe and Vercel Gateway adapters were implemented from documentation but had not been live-tested at review time.
- Language
- Python
- License
- MIT
- Author
- Parth Kp
Sources
Related projects
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.
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.
Kill My Idea
Turn a startup pitch into a compact set of scores and a blunt Kill, Fix, or Ship verdict.
What Jev decides
Dimension scores, idea category, whether the pitch is understandable, and supporting diagnostic signals.