[ Use case · Search ]
Can Jev power semantic search?
Partly. Jev can't search a catalogue on its own, but it works well as a re-ranker: once a conventional search returns 50 or so candidates, Jev scores how well each one matches the query so the best rise to the top. One builder used this to search property listings from a plain-English brief.
Is it a fit?
Use Jev when
- Queries are written in natural language, not keywords.
- A conventional search already produces a shortlist of candidates.
- Each result has a short text description Jev can read.
Skip it when
- You need to search millions of items directly. Jev evaluates one item at a time, so narrow the list first.
- Results are mostly images.
The questions
One request per result, 2 questions. Jev answers them in parallel.
- Score
How well does this listing match what the person asked for?
Scale: Not a match → Partly matches → Good match → Exactly what they asked for
- Noul (yes/no)
Does this listing break a hard requirement in the search, such as price, pets or number of bedrooms?
Options: Yes · No
Getting started
[ 1 · Test it in the playground ]
No code needed. Paste a real result into TypeSafe's playground and ask: “How well does this listing match what the person asked for?” with the options Not a match, Partly matches, Good match, Exactly what they asked for. You'll see the answer and its probability.
Open the playground (opens in a new tab)[ 2 · Use an existing tool ]
We haven't found a ready-made tool for this workload yet. These directories track what's been built on Jev:
Made with Jev (opens in a new tab)Jev.Store (opens in a new tab)JevHunt (opens in a new tab)
[ 3 · Integrate it ]
To run it on every result automatically, call Jev from the system where your results live. The developer details below include the full request.
What people have reported
Natural-language property search
self-reported1 search for 18¢, under 20 seconds
Justine Moore · Made with Jev (opens in a new tab)
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 re-rank search results with Jev?
- About 1 cent per 1,000 results scored at list price. At 50 results per query, 1,000 queries cost around 50 cents.
- Can Jev search my whole catalogue by itself?
- No. It evaluates one item at a time. Use a conventional search to produce a shortlist, then let Jev order it.
[ For developers ]Request body, wiring and pitfalls
The request
One request per result with all 2 questions batched over the same state, so Jev reads the state once. Keys below are the names you'll see in the response.
{
"model": "jev-latest",
"state": "Search: \"Quiet 2-bedroom near a park, under $3,000, allows a dog\"\nListing: 2 bed / 1 bath apartment, $2,850 a month. Two blocks from Riverside Park. Top-floor unit, no shared walls. Cats only.",
"questions": {
"match": {
"type": "score",
"instructions": "How well does this listing match what the person asked for?",
"criteria": [
"Not a match",
"Partly matches",
"Good match",
"Exactly what they asked for"
]
},
"breaks_requirement": {
"type": "noul",
"instructions": "Does this listing break a hard requirement in the search, such as price, pets or number of bedrooms?"
}
}
}How to wire it up
- 1.Run your normal search first (keyword or vector search) and keep the top 50–100 results.
- 2.Send Jev the person's search plus one result at a time, and ask both questions in the same request.
- 3.Drop anything with breaks_requirement ≥ 0.5, then sort the rest by the match score.
- 4.Show the top results. Cache answers for popular searches so you don't pay twice.
Watch out for
- Costs multiply: every search checks many results. Price it in the calculator with results per month, not searches per month.
- Two-word searches give Jev little to judge. For short searches, a normal search may do just as well.
New to writing Jev questions? Read the question design guide.
Related use cases
Can Jev make decisions inside an AI agent?
Pick the next action in an agent loop in milliseconds.
Can Jev score and route inbound leads?
Rank inbound leads by fit and intent before a human looks at them.
Can Jev triage an inbox?
Sort mail into categories, rate urgency and flag what needs a reply.