Keenable SELECT listing history

Keenable SELECT answers a research question with one query instead of ten page reads.

Ordinary web search hands the model a list of links and leaves the reading, the extraction and the de-duplication to it. SELECT does that work inside the search. The model writes one read-only SQL statement in a DuckDB dialect; the server runs the searches in parallel, fetches the pages, pulls out the fields the query asks for with a per-row model call, removes duplicates, and returns a table.

One query can cover more than a thousand pages, filter them with ordinary SQL at no extra cost, and group them into the answer. The rows arrive as a table, not as a wall of page text, so the answer is grounded in a lot more sources than the context could hold.

Every result is saved as a result set for 30 days, so the next question can be asked of the same pages without searching again. A finished table can be published as a single HTML report, charts and all, at a link you can hand to someone.

Use it for a survey, a comparison or a ranking across many sources. It is not the tool for a quick look at one page: a query runs from seconds to minutes, because it really does search, fetch and read.

Connect your Keenable account once to enable it. Searching and fetching bill that account.

Description from the ChatGPT Plugins listing.