Franck Pachot
speaker
138 appearances
1 recordings
1 series
first heard Jul 2026
last heard 8 Jul
Franck Pachot’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Jul 2026 with 1.
Appearances
The first one is being able to store vectors and embeddings.
And that's already possible with Postgres PG vector extension.
So of course it is available there in HorizonDB, like it is available in other Postgres managed services.
And then you want to index those embeddings.
So PG vector comes with already with vector indexes, but they have some limitations.
In HorizonDB, you can use the PG vector indexes and you have, in addition to that,
the disk ANN extensions there that you can run on Horizon DB with a different index that provides better performance, but also a very interesting feature like pre-filtering.
Because usually you want to do similarity search on embeddings, but you don't want to look at all embeddings.
For example, if you are in multi-tenant, you want to focus only on that tenant.
If you
If you look at the past month data, you want to focus on it.
And the indexes that come with PG vector, they cannot combine the filtering with the index traversal.
So either they use the index, they have too many candidates, and then they have to filter them.
With ANN, one big advantage is that it is filtered during the traversal of the index.
But it's a choice because this one is available only on HorizonDB, but the syntax, everything is the same from the application point of view.
You just create a different index that gives you better performance, but still the application that runs queries on pgVector
The query planner will pick the disk and then index if it's better.
You'll see the difference in the execution plan.
If you look at the execution plan, then you have some clues about how it is executed.
But the query itself is the same, yeah.
Showing 41–60 of 138 · page 3 of 7
← Previous
Next →