was wondering whether its good for timeseries data?

It depends on how you plan to use/query the timeseries data… Could you expand on your use-case?

ha

Oh haha! Hey Josh!

I’m founder of Bundlr a blockchain based startup (immutable dataset) context on what we’re trying to achieve: - build an index/query service for bundlr txs (can be in the billions in next year) - each tx has key-value metadata attached to it (like S3) - each tx has a unique timestamp attached to it (i.e. not per block) - building an api which lets you query in time order with filters on metadata + sender address graphql API we’re building

query {
 transactions(
  owners: ["0x..."], 
  tags: [{ name: ..., key: ... }],
  order: "ASC"
 ) {
  edges {
   node {
    id
    address
    receipt {
     timestamp
    }
   }
  }
 }
}

so multivariat timeseries

Yup, that query pattern should be fine with Typesense, you’re essentially filtering on numeric values + string values from Typesense’s perspective. Now in terms of dataset size, Typesense is an in-memory datastore (optimized for performance), and it typically takes 2x-3x RAM to index a dataset of size X. So you’d have to weigh the cost-benefit of putting 1B 10B documents in memory vs the performance gain / UX improvement you get for that

ah

Hahaha!

…yet

Sadly, I think ES might be your best bet for that scale

yeah i figured

There’s also Zinc search, but they’re tackling the log-search use-case primarily