Repository navigation
Keenable as a free, keyless web search source for boards without a portal skill #811
Replies: 1 comment
|
Hi Ilya, thank you for bringing this here before building it, and for the disclosure. The pricing part clears the bar from #250 cleanly: public endpoints, no account, no per-call charge. The proposed shape also reads like someone who has studied the contract (stderr JSON errors, We are still going to say no to it upstream, and the reason is fit rather than quality. A portal skill earns its place by giving If you want it to exist anyway, the right home is a standalone skill repo under your own account, with a short post in the community index at #78 the way other standalone skills have done. People who want a web-search fallback can copy the folder in, and you keep ownership of it. If Keenable ever exposes structured job results (title, company, location, date), that would be a different conversation, and a portal-shaped one. Thanks again for the careful proposal, and good luck with Keenable. |
Uh oh!
There was an error while loading. Please reload this page.
Hi Mads, a proposal before anything goes upstream, following your note on #250 about free-at-point-of-use services.
#250 (firecrawl-search) was declined because every search was metered. Keenable's web search and page fetch have public endpoints that need no account or API key and charge nothing per call (rate-limited to 10 requests/s and 1,000/hour per IP; an optional token only lifts the limits). That would cover the fallback need #250 identified (boards without a portal skill, a broken portal, reading a posting URL) at the same zero-signup bar as linkedin-search.
Proposed shape:
.agents/skills/keenable-search/under the portal-skill contract (search -q ... [--site <domain>] [--jobage N],detail <url>, zero runtime deps, stderr JSON errors, backoff on 429/5xx,stripInvisibleon all output, network-free bun tests), plus its settings.json and security_guards.py entries. It would shipenabled: false: web hits have no structured company/location/date, so /scrape's 14-day window,candidatesand the health check can't read them, and it is meant for direct use.Dependency, stated plainly: Keenable is a hosted commercial service with no self-hosted version; when it is unreachable the CLI exits 1 with a clear message. Only the query and filters (search) or the posting URL (detail) are sent, nothing from the profile. I have a draft that follows the contract and will adjust it to whatever you decide, including not upstream at all.
Disclosure: I work at Keenable.
All reactions