Jev API search endpoint
TypeSafe’s line-by-line search and re-ranking cookbooks still assume you have candidates, then ask Choice/Noul/Score questions. There is no documented /v1/search that indexes your corpus.
This page is for builders who searched an extra “endpoint” name while wiring Jev / System One. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. We do not sell or issue keys. Confirm first-party routes on docs.typesafe.ai/api.
The public API reference documents one evaluation route — POST https://api.typesafe.ai/v1/systemone — plus GET /v1/models. This title is not a separate first-party URL in that reference. Treat the capability as something you implement around the official call.
Steps
- Retrieve candidates with your search / vector store.
- Filter to the passages the question needs.
- Ask Jev which passage is relevant or whether a claim is supported.
- Do not send the entire index as state.
Map onto the official POST
Put named passages in state (or structured instructions with backtick paths). Use a Choice among candidate ids (unordered, max 255) or a Noul per passage (“does this support the claim?”). Score is for ordered quality rungs, not a substitute for BM25. Question ids are not sent to the model; they only key answers.
Choice/Score return probabilities plus confidence. Noul returns noul only. Gate citation emit on those fields in your code. Schema-safe “supported” is not the same as true in the world.
jev-latest and jev-preview currently resolve to jev-1.13.0 — confirm. Pin the versioned id before you tune a hold floor. Vendor-claimed context (64k state+all questions; 32k state+longest question) is why you must not dump the index. Unused tokens still meter as input (vendor claim: $0.042 / million input, output free).
When not to use this pattern
Do not use Jev as Elasticsearch, Solr, or a crawler. It evaluates text you already have. It does not browse the web. If you need offsets and URLs, store them beside the evaluation — Jev will not invent a bibliography. Pagination of hits is yours.
Pitfalls
- Treating Jev as Elasticsearch.
- Inventing query-string APIs on
api.typesafe.ai. - Sending the top-200 hits “just in case” and wondering why
jev-1.13looks distracted. - Asking Jev to generate a citation string. Generation is the wrong product shape.
Documented HTTP errors remain 401, 422, 429, 529. Official SDKs retry 429/529. A 200 with a flat Choice is a retrieval-or-wording problem, not a missing search route.
See also
- API hub — first-party POST versus pattern titles
- Evidence fields — what you persist beside the judgment
- State — filter before you evaluate
- RAG decision workflow — retrieve, then ask
FAQ
Is /v1/search documented?
No.
Vector DB vs Jev?
Retrieval vs judgment. Cover both; do not clone a rival IA.
Pagination of hits?
Yours — pagination.
Can I POST a query string to api.typesafe.ai?
Not in the public reference we used. Hydrate state, then evaluate.
What this page does not claim
- Official TypeSafe status, logos, or support.
- That jev.pro sells, issues, or proxies API keys.
- Extra first-party HTTP routes beyond
POST /v1/systemoneandGET /v1/modelsunless the API reference lists them. - Invented latency SLOs, search volume, or unofficial prices. List price and rate limits are vendor claims — confirm on models.
- That a schema-constrained answer is automatically factually correct.
Disclaimer
This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.