Check out this short video in which Sid Probstein, creator of Swirl Metasearch, explains the main features of Swirl.
This video goes by very quickly. Sid takes time to highlight a couple of key points:
Metasearch can query “anything with a search API” and make sense of the results; it doesn’t have to ingest and analyze the content with all the answers first.
Swirl includes query and result processing pipelines; these allow developers and administrators to customize the way Swirl integrates with and normalizes information from disparate sources without necessarily writing any code.
Subscribing to a search is one of the more powerful features; it literally offers the ability to never run an important search again. Instead, users can subscribe and be updated whenever new results are detected – systematically and exhaustively!
Here are a few things Sid left out:
Swirl highlights evidence of the user’s query in each result item to support rapid visual scanning of results
Swirl re-ranks results using large language models – the same technology underlying modern AI including ChatGPT and other Generative AI. But Swirl is not generative. You can trust the results you get from Swirl as much as the sources that provide the responses. It doesn’t alter results, or invent… anything.
Swirl also detects and removes duplicated content; although this is still a work in progress, this helps keeps result lists clean for scanning. Users can drill into any single silo, if they want to see all results including duplicates.