Start with Jev.
Then make it local.

Turn your Jev-usage into small, task-specific classifiers. Train with one command. Run locally in your browser or Node.js.

A few lines, then it’s yours.

Bring saved answers, or have Jev label your inputs.

Train

npx jimothy train \
  --task task.json \
  --data emails.jsonl \
  --out models/email

# Set AI_GATEWAY_API_KEY in your shell
npx jimothy train \
  --task task.json --inputs emails.jsonl \
  --teacher typesafe-ai/jev \
  --out models/email

Run

import { loadClassifier } from 'jimothy/browser';

const model = await loadClassifier('/models/email/');
const { answer } = await model.predict({
  from: 'maya@example.com',
  subject: 'Lunch on Friday?',
  body: 'Are you free around twelve?'
});

console.log(answer);
await model.dispose();

import { loadClassifier } from 'jimothy';

const model = await loadClassifier('./models/email');
const { answer } = await model.predict({
  from: 'maya@example.com',
  subject: 'Lunch on Friday?',
  body: 'Are you free around twelve?'
});

console.log(answer);
await model.dispose();

Browser setup & full API

Route support tickets
Send billing, delivery, and technical questions to the right team.
Organize your content
Tag notes, documents, and saved links as they arrive.
Make sense of feedback
Sort feature requests, bug reports, and customer sentiment.