Search the live web in plain language
TypeSafe’s Jev reads your request, picks sources and a time window, then ranks what Search1API returns. You get links, snippets, and relevance scores. Nothing is written as an answer.
Clone — bashgit clone https://github.com/superagents-lab/jev-search.git
A search UI, not a chatbot. Each result keeps its source, score, and original snippet.
Jev turns a typed question into a query, sources, and a time range you can still override.
Google, DuckDuckGo, Yandex, and vertical engines run concurrently through Search1API.
Results merge by URL and stream as each lane finishes. Lower scores sit in a separate group.
Try asking
TypeSafe Jev API documentation and examples / Jev discussions on Hacker News this week / Videos about TypeSafe Jev this month
What Jev Search is
Jev Search is an open-source search application built by Search1API. TypeSafe’s Jev model chooses sources, time ranges, and search terms, then scores the results Search1API returns. The hosted demo is at jev.s1.dev. You can install it as an app from the browser; searches still need a network, and the installed app only caches an offline page. This is an independent project, not an official TypeSafe product.
How a search runs
The application streams newline-delimited JSON from POST /api/ask. Each engine has a 15-second deadline inside a 30-second request deadline.
Understand
Jev answers typed questions about your request. Those judgments pick a query, sources, and a time range. You can still change the source and time chips.
Search
Google, DuckDuckGo, and Yandex cover the open web. Hacker News, Reddit, and GitHub each combine a Google site-restricted search with their dedicated engine. X, arXiv, YouTube, Wikipedia, IMDb, and WeChat use vertical engines. One failed engine does not discard another’s results.
Rank
Jev scores each result for relevance. Hits merge by URL, ordered by relevance, engine agreement, and original rank, and stream as each lane finishes. A failed source shows a warning rather than a zero-result count.
Sources it can pick
Default open-web sources are Google, DuckDuckGo, and Yandex. The rest turn on when Jev decides they fit the request — or when you override the chips.
DuckDuckGo
Yandex
Hacker News
GitHub
X
arXiv
YouTube
Wikipedia
IMDb
Worth knowing
There is no generated answer. The page shows links, snippets, and model-judged relevance scores — not verified accuracy.
TypeSafe also receives result titles and snippets for scoring. Search1API queries the selected engines.
Application code is MIT licensed. TypeSafe and Jev names and brand assets belong to their owners and are not covered by that license.
Run it yourself
Local development needs Node.js 22.12+, pnpm, a Search1API key, and a TypeSafe key. The README covers Cloudflare Workers deploy, cache, and rate limits.
Questions
Is this an official TypeSafe product?
No. It is built by Search1API and uses TypeSafe’s Jev model. TypeSafe names and brand assets belong to TypeSafe.
Does it write an answer?
No. You get ranked links and snippets with visible scores. Selecting and ranking existing results does not verify their claims.
Do I need an API key to try the demo?
The hosted demo at jev.s1.dev runs on Search1API’s keys, with a per-IP rate limit. Self-hosting needs your own Search1API and TypeSafe keys; each search can make several billable provider calls.
Can I install it as an app?
Yes. Add to Home Screen on iOS, or Install app on Chrome and Edge. Searches still need a network connection. The installed app caches only an offline page, not results.
Explore more
Try a live search
Open the hosted demo, or clone the repo and run it with your own keys.