Skip to content

Jev directory

Jev email intent workflow

An async LangGraph workflow that sends a mocked email to Jev, receives a typed Choice between invoice and general, and routes it to a demo handler, with confidence and both label probabilities printed for inspection.
Stars
11
Language
Python
Reading time
3 min
Published by
Community
Author
GiesN
Last verified
Sep 2026

About this project

A small async LangGraph workflow that sends a mocked email to TypeSafe's Jev model, receives a typed Choice (invoice or general), and routes to a demo handler. The handlers only set a destination in graph state; they do not send email or make payments.

Quoted from the project's README. Read the full README.

Highlights

  • A LangGraph node sends the mocked email to Jev and uses the typed invoice-or-general Choice to branch to accounts payable or a general inbox.
  • The ten labelled mock emails are a smoke check, not an accuracy benchmark; expected labels are never sent to Jev.
  • Output includes intent, confidence, both label probabilities, destination, model, and comparison with the expected label.
  • Defaults to the jev-1.12 model, overridable through TYPESAFE_DEFAULT_MODEL, and API failures stop execution instead of inventing an intent.
  • Tests mock the SDK response and run the actual graph without credentials or network.

Quickstart

bash
uv sync
uv run python main.py
uv run python main.py --email-id email-09
uv run python -m unittest discover -s tests -v

Watch out

No license file, so reuse terms are unclear. Needs Python 3.10+, uv, and a TypeSafe API key; classification makes one real API call per email, and the handlers only set a destination in graph state rather than sending email or making payments.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai