Skip to content
JevHub

[ Use case · Inbox ]

Can Jev triage an inbox?

[ Short answer ]Strong fit

Yes. Email triage is the most widely reproduced Jev workload: one request per email can assign a category, rate urgency and flag whether a human needs to see it. An open-source project reports 1,000 emails triaged in about a minute for around 3 cents.

Is it a fit?

Use Jev when

  • You or your team spend real time deciding what to open first.
  • Your categories are stable (for example: needs reply, updates, promotions, sales, spam).
  • A human stays in the loop for anything consequential.

Skip it when

  • You want replies drafted. Jev can't generate text; pair it with a language model for the emails that need one.
  • Your privacy obligations don't allow sending email content to a third-party API.

The questions

One request per email, 3 questions. Jev answers them in parallel.

  • Choice

    What kind of email is this?

    Options: Needs reply · Updates · Promotions · Sales · Spam

  • Score

    How time-sensitive is this email for us?

    Scale: Whenever → This week → Today → Within the hour

  • Noul (yes/no)

    Does this email involve money, legal risk or an upset customer?

    Options: Yes · No

Getting started

  1. [ 1 · Test it in the playground ]

    No code needed. Paste a real email into TypeSafe's playground and ask: “What kind of email is this?” with the options Needs reply, Updates, Promotions, Sales, Spam. You'll see the answer and its probability.

    Open the playground (opens in a new tab)

  2. [ 2 · Use an existing tool ]

    Made with Jev (opens in a new tab)Jev.Store (opens in a new tab)JevHunt (opens in a new tab)

  3. [ 3 · Integrate it ]

    To run it on every email automatically, call Jev from the system where your emails live. The developer details below include the full request, and the Recipe Pack ships a ready-to-run version.

What people have reported

Self-reported figures are the author's own; we haven't reproduced them. More on the costs page.

Common questions

How much does it cost to triage 1,000 emails with Jev?
Roughly 2 to 4 cents. Public runs report 3 to 3.5 cents per 500–1,000 emails, and the list-price estimate for a typical email with three questions is about 2.4 cents per 1,000.
Can Jev write the reply too?
No. Jev returns choices, scores and probabilities only. Use it to decide which emails need a reply, then hand those to a language model such as GPT or Claude.
[ For developers ]Request body, wiring and pitfalls

The request

One request per email with all 3 questions batched over the same state, so Jev reads the state once. Keys below are the names you'll see in the response.

POST api.typesafe.ai/v1/systemone
{
  "model": "jev-latest",
  "state": "From: Dana Lee <dana@acme.io>\nSubject: Invoice #4471 charged twice\nBody: Hi, our card was charged twice for March. Can you refund the duplicate today? Our finance close is tomorrow.",
  "questions": {
    "category": {
      "type": "choice",
      "instructions": "What kind of email is this?",
      "criteria": {
        "needs_reply": "A person expecting a response from us",
        "updates": "Notifications, receipts, product updates",
        "promotions": "Marketing and newsletters",
        "sales": "Someone selling to us",
        "spam": "Unwanted or suspicious mail"
      }
    },
    "urgency": {
      "type": "score",
      "instructions": "How time-sensitive is this email for us?",
      "criteria": [
        "Whenever",
        "This week",
        "Today",
        "Within the hour"
      ]
    },
    "needs_human": {
      "type": "noul",
      "instructions": "Does this email involve money, legal risk or an upset customer?"
    }
  }
}

How to wire it up

  1. 1.Send sender, subject and the first ~1,500 characters of the body as state. Long threads cost more and rarely change the answer.
  2. 2.Ask all questions in one request per email.
  3. 3.Apply labels automatically; only surface needs_reply mail with urgency ≥ 2 as a notification.
  4. 4.Anything with needs_human ≥ 0.5 goes to a person, whatever the category.

Watch out for

  • Newsletters that look personal are the most common misclassification. Add an example of one to the promotions criterion.
  • Test on a de-identified sample first and check the provider's data retention settings before running a live inbox.

New to writing Jev questions? Read the question design guide.

Related use cases

[ Support ]Strong fit

Can Jev route support tickets to the right team?

Send each ticket to the right queue and flag frustrated customers.

ChoiceScoreNoul (yes/no)
≈ 2.1¢ per 1,000
[ Trust & safety ]Good fit

Can Jev moderate user content and catch prompt injection?

Flag spam, abuse and prompt injection before content goes live.

Noul (yes/no)Score
≈ 1.1¢ per 1,000
[ Research ]Strong fit

Can Jev filter papers, articles and news for relevance?

Surface only the papers and headlines that match what you care about.

ScoreChoiceNoul (yes/no)
≈ 2.1¢ per 1,000