Repository navigation
feat(demos): add keyless Keenable MCP web search and screenshots - #191
ilya-bogin-keenable wants to merge 1 commit into
Conversation
|
@ilya-bogin-keenable is attempting to deploy a commit to the andylizf's projects Team on Vercel. A member of the Team first needs to authorize it. |
|
@ASuresh0524 and @andylizf, could one of you take a look at the code here when you have a moment? Thanks! |
|
Yes @ilya-bogin-keenable. Will take a look at this by EOD today. |
|
@ASuresh0524 @andylizf gentle ping on this one. The branch is still |
|
Status update — this is now green and mergeable, and the only thing left on it is a decision no amount of code review settles. What I changed: merged current What is left: the positioning question @andylizf raised — whether a managed/commercial backend demo belongs in this repo — has never been answered, and it is the only thing holding this. There is precedent in both directions: #175 merged a vendor demo into I have drafted one for @ASuresh0524 and @andylizf to settle, proposing the bar be about upkeep rather than commerce: the demo has to show something about PixelRAG rather than just the vendor, stay self-contained, pass lint, and carry a named owner and a You will get a yes or a no against that, not more silence. Apologies that it has taken this long to get to a straight answer. |
|
Thanks for merging main and fixing the lint. That bar works for us: I'll be the named owner, add a |
Sibling of demos/parallel_search: discover pages with Keenable's keyless MCP (search_web_pages, optional fetch_page_content), save a results.json manifest, and optionally render the selected pages with PixelRAG's render_urls. Skips localhost and loopback, private and link-local IP hosts before fetching or rendering, and tests both the structured-content and JSON-text tool result paths.
78ce1ad to
d14cad9
Compare
Adds an optional Keenable Search MCP demo next to
demos/parallel_search/, with the same shape: find pages on the open web, save source URLs and snippets, add--fetchfor page text, and opt into--renderfor PixelRAG screenshot tiles. Search and fetch work without an account or API key.What changed
demos/keenable_search/: isolated uv project (run.py,test_run.py,README.md,pyproject.toml,uv.lock,.gitignore). It callssearch_web_pagesandfetch_page_contentonhttps://api.keenable.ai/mcpover Streamable HTTP and hands selected URLs to the publicrender_urlsAPI.README.md: one line under "Render a page programmatically", after the Parallel demo link.What is sent to Keenable: the query, the selected URLs when
--fetchis used, and anX-Keenable-Title: PixelRAGheader. AnX-API-Keyheader is sent only if the user setsKEENABLE_API_KEY(for higher rate limits).Following the review notes on #175, results on
localhostor loopback, private and link-local IP hosts are skipped before fetching or rendering, and the tests run the MCP transport against both astructuredContentresult and a JSON text result.Tested
pytest demos/keenable_search/test_run.py: 6 passed (Python 3.12, mcp 1.30.0, httpx 0.28.1, macOS). The tests run the real MCP client overhttpx.MockTransport, check the exact tool arguments, the title header, no key header by default, the opt-in key, a failed fetch, URL filtering and the renderer handoff.demos/parallel_searchtests still pass (3).ruff checkandruff format --checkondemos/keenable_searchpass (demos/is excluded from the root lint and test jobs, so these were run by hand).uv lock --checkpasses. The lock is the feat(demos): add free Parallel MCP web search and screenshots #175 lock with only the project name changed (samemcp,httpxandpixelrag 0.4.0entries).run.search_pages: 3 queries returned 10 results each; 0.8 to 1.2 s end to end per run including MCP initialize, and 1.7 to 1.9 s with--fetchof 3 pages.run.py "Find explanations of retrieval augmented generation" --limit 2 --fetch --renderon macOS with Google Chrome: 2 arXiv pages selected and fetched, 3 screenshot tiles written, manifest saved.Disclosure: I work at Keenable, which runs this search API.