j·Jev ClassifyDOCUMENT INTELLIGENCE
Opening saved work…
YOUR WORKSPACE

Ask once. Read every document.

Prepare your documents and questions, then let Jev classify them together.

01   Add documents02   Prepare questions03   Run with Jev
01

Documents

Prepared locally in your browser

Add your documents

Drop documents hereor browse your computerPDF, Word, TXT, Markdown (.md, .markdown) or HTML
No documents added yet.
Extract text→No API key

Prepared document text

0 ready

This is the state Jev will read. One document per request.

Your document appears here

Add a file or paste text to inspect what Jev will receive.

Text extraction stays on this device.
02

Questions

Write new questions or reuse a saved set

What do you want to know?

Have answer choices? Paste the question and all its options into one box. Ordered rubrics are kept as choices unless you request a numeric score.

Study context & document context optional

When checked, document titles and the first 2,500 characters of the first document also go to the selected language model.

QUESTION CONVERSION MODELClaudeProvider key needed

Sends questions + study context + Jev format rules.

Question conversion connects directly to your chosen provider. Keep the app open while a request is running; completed work is saved on this device.

Uses its own provider key.
Convert→Your chosen LLM

Jev questions

0 ready

Approval is optional. All included questions with valid formatting can run. Incomplete questions are skipped.

Your prepared questions appear here

Convert English questions, load JSON, or .

Yes/no · Multiple choice · Ordered scale
03   DOCUMENTS + QUESTIONS

Ready to run with Jev?

jev-latest

Jev receives each document’s text and your prepared questions.

Jev connection key needed.

Install Jev Classify

Open Jev Classify from your home screen or desktop, in its own window.

Saved workspaces and the app open offline after installation finishes. AI requests need an internet connection. Workspaces are saved on this device; use a backup to move them to another device or browser.

Connections

Model connections

Keys are kept only in this browser (for this tab, or longer if you tick "remember"). Changes save automatically.

Service A Language-model key — rewrites your questions for Jev
Not set

Used in step 2, when the app rewrites your questions. Pick the company you have an account with. The key comes from that company's website, not from TypeSafe. The app remembers a separate key for each company.

The key is checked automatically a moment after you paste it.
Service B Jev connection — reads your documents
Not set

Advanced
Observed in September 2026: documents around 90,000 characters ran; 176,000 failed with max_tokens_exceeded. The real ceiling may move.
Off: keys are kept for this window’s session. On: keys stay on this device. Workspace backups never include keys. Translation connects directly to the selected provider; Jev uses the connection chosen above.

How it works

Documents are extracted locally into state. English questions and Jev format rules go to a selected language model with its provider key. Saved question JSON bypasses conversion and generated questions can be exported immediately. State and prepared questions are submitted together to Jev with a separate TypeSafe key, producing a results table.

For direct Jev access, use an OpenRouter key in place of the TypeSafe key shown in the diagram. The optional document-context setting also shares titles and an excerpt with the conversion model. Open full-size diagram ↗