ParqDB – Vector search in the browser from Parquet over HTTP"
Runs vector search in your browser, reading Parquet index files via HTTP range requests.
ParqDB embeds your search query locally in the browser with a MiniLM model, then finds matches by reading only the needed byte ranges from immutable Parquet files over plain HTTP — no vector database or query server involved. The public demo searches 100,000 Wikipedia articles using an IVF-LVQ8 index published straight to object storage. Because the query is embedded client-side, the text never leaves the browser tab.
What holds up
- +Query text is embedded in the browser and never sent to a server — there's no backend to log it.
- +No vector database to run — the index lives as immutable Parquet files in plain object storage.
- +Demo searches 100,000 Wikipedia articles live, showing real query times and byte-range requests.
Mind the limits
- −First query loads a MiniLM transformer into the browser — a cold start before results appear.
- −Index is immutable Parquet in object storage — updating the dataset means republishing it, not live writes.
Featured here? Take the badge
Put it on your site — it links back to this review. Free for every listed product, always.
<a href="https://stillworks.dev/products/p/parqdb-vector-search-in-the-browser-from-parquet-over-http/"><img src="https://stillworks.dev/badge/parqdb-vector-search-in-the-browser-from-parquet-over-http.svg" alt="Picked by StillWorks" width="250" height="54"></a>