3 ms·

Show HN: I built a site to instant-search 32 Million Songs in milliseconds

16 points by jabo 6y ago

jabo 6y ago
Why? - A friend and I are working on an open source alternative to Algolia called Typesense [1]. I kept getting asked how large a dataset Typesense can handle. So I built a demo with the largest open structured dataset I could find. You get instant search-as-you-type results in as little as 40ms from (did I mention) 32 Million records! Here's the source code: https://github.com/typesense/showcase-songs-search https://github.com/typesense/showcase-songs-search Some details about the tech stack: - The search backend is powered by Typesense Server v0.17.0 running on a geo-distributed cluster (Oregon, Frankfurt, Mumbai) on Typesense Cloud: https://cloud.typesense.org/ https://cloud.typesense.org/ - The 32M songs dataset is from https://musicbrainz.org's https://musicbrainz.org's open library. Please contribute song metadata if you can. - The Search UI was built with https://github.com/typesense/typesense-instantsearch-adapter https://github.com/typesense/typesense-instantsearch-adapter - ParcelJS for an app bundler - Deployment: `git push` > Deploys to DigitalOcean's App Platform [1] https://github.com/typesense/typesense https://github.com/typesense/typesense
- hakkikonu 6y ago
  does song data on your db or you're requesting to musicbrainz instantly?
  - jabo 6y ago
    It's stored in a Typesense search server on my side.
purplecats 6y ago
Seems neat but the quality of the results didn't seem very high. "taylor swift style" -> no correct results lot of taylor swift searches seem to just surface "cardigan", whatever song that is of hers.
- jabo 6y ago
  The MusicBrainz dataset unfortunately does not have a popularity score and so I've only ordered results by their text_match_score and release_date. So songs that were more recently released are given higher weightage unfortunately. In a production-grade search setting, you'd typically want to have a popularity score and sort by that.

16 points by jabo 6y ago

jabo 6y ago

Why? - A friend and I are working on an open source alternative to Algolia called Typesense [1]. I kept getting asked how large a dataset Typesense can handle. So I built a demo with the largest open structured dataset I could find. You get instant search-as-you-type results in as little as 40ms from (did I mention) 32 Million records! Here's the source code: https://github.com/typesense/showcase-songs-search https://github.com/typesense/showcase-songs-search Some details about the tech stack: - The search backend is powered by Typesense Server v0.17.0 running on a geo-distributed cluster (Oregon, Frankfurt, Mumbai) on Typesense Cloud: https://cloud.typesense.org/ https://cloud.typesense.org/ - The 32M songs dataset is from https://musicbrainz.org's https://musicbrainz.org's open library. Please contribute song metadata if you can. - The Search UI was built with https://github.com/typesense/typesense-instantsearch-adapter https://github.com/typesense/typesense-instantsearch-adapter - ParcelJS for an app bundler - Deployment: `git push` > Deploys to DigitalOcean's App Platform [1] https://github.com/typesense/typesense https://github.com/typesense/typesense

hakkikonu 6y ago
does song data on your db or you're requesting to musicbrainz instantly?
- jabo 6y ago
  It's stored in a Typesense search server on my side.

purplecats 6y ago

Seems neat but the quality of the results didn't seem very high. "taylor swift style" -> no correct results lot of taylor swift searches seem to just surface "cardigan", whatever song that is of hers.

jabo 6y ago
The MusicBrainz dataset unfortunately does not have a popularity score and so I've only ordered results by their text_match_score and release_date. So songs that were more recently released are given higher weightage unfortunately. In a production-grade search setting, you'd typically want to have a popularity score and sort by that.