Repository navigation
feat(search): search IVF indexes over generated embeddings by text - #331
Merged
Merged
Conversation
An IVF index on a text column with an embedding provider can be searched by text: `hotdata search` names the table's columns rather than `*`, which the index cannot serve, and orders by vector_distance on the source text column. IVF over a vector column keeps its refusal and hint. `search show` and `search list` name the generated embedding column for any embedding-backed index.
anoop-narang
requested review from
eddietejeda
and removed request for
a team
October 12, 2026 14:02
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
An IVF vector index can now be created on a text column with an embedding provider (
--algorithm ivf --provider <id>); the server generates the embeddings and builds the IVF index over them. This adds the CLI side.hotdata search "<text>" --index <ivf index>works when the index has a provider. The query names the table's columns (excluding the generated embedding column) and orders byvector_distance(<source column>, '<text>'). ASELECT *or* EXCLUDE (...)projection can't be served by an IVF index and would fall back to a slow exact scan.--selectis still honoured. An IVF index over a column that already holds vectors keeps its existing refusal and hint.search create --algorithmhelp now saysivfworks on a vector column, or on a text column with--provider, with thel2orcosinemetric.search showandsearch listname the generated embedding column for any embedding-backed index (HNSW too):embedding_columnin show and its JSON/YAML, andbody_embedding (from body)in the list.Tests cover the named projection (no
*orEXCLUDE), quoting and escaping,--select, the unreadable-columns error, the kept refusal, the show/list output, and the help text.cargo test: 650 passed.scripts/check-skills.shpasses.