feat: fix Gemini embedding config and expose Postgres index tuning #1050
Conversation
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
|
Update: fixed the rust formatting error in Someone kindly run the checks once again for this. |
|
Thanks a lot for contributing! |
georgeh0
reviewed
Sep 25, 2025
|
@georgeh0 thankyou for the review. I pushed a change addressing the comments. Updates:
|
georgeh0
reviewed
Sep 25, 2025
|
Great PR, thanks a lot @banrovegrie for your contribution! |
|
@badmonster0 @georgeh0 thankyou. We have been using cocoindex for our startup and it has been a pleasure! |
0xTnxl
added a commit
to 0xTnxl/cocoindex
that referenced
this pull request
Oct 2, 2025
Implements HNSW vector index support for Kuzu following the same pattern as the Postgres implementation in PR cocoindex-io#1050. Changes: - Remove blanket "Vector indexes are not supported for Kuzu yet" error - Add validation to accept HNSW and reject IVFFlat with clear error message - Implement CREATE_VECTOR_INDEX and DROP_VECTOR_INDEX Cypher generation - Map cocoindex HNSW parameters to Kuzu format (m→mu/ml, ef_construction→efc) - Add vector index lifecycle management (create, update, delete) - Install Kuzu vector extension automatically when needed - Support all similarity metrics (cosine, l2, dotproduct) Technical details: - Add VectorIndexState struct to track index configuration - Update SetupState and GraphElementDataSetupChange for index tracking - Implement diff_setup_states logic for index change computation - Add vector index compatibility checking in check_state_compatibility - Integrate vector index operations in apply_setup_changes Fixes cocoindex-io#1055 Related to cocoindex-io#1051 Follows pattern from cocoindex-io#1050
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
This PR attempts at rolling in two related improvements. It addresses both #1049 and #1043.
First, Gemini embeddings now mirror
outputDimensionalityinside theconfigblock, sogemini-embedding-*honors custom dimensions as requested by AurumnPegasus.Second, we expose vector index tuning for Postgres: callers can choose between HNSW and IVFFlat (with parameters) from Python, and the engine creates the matching index with pgvector options and updated naming.
Changes
src/llm/gemini.rs: build the embed payload via a helper so Gemini gets both top-level andconfigoutputDimensionality.src/base/spec.rs: addVectorIndexMethod(HNSW/IVFFlat with optional params), default existing indexes to HNSW, extendVectorIndexDefformatting.src/ops/targets/postgres.rs: honor the selected method when generatingCREATE INDEX, emitWITHoptions (m,ef_construction,lists), suffix non-default methods in the index name.python/cocoindex/index.py,python/cocoindex/__init__.py: expose matching Python dataclasses (HnswVectorIndexMethod,IvfFlatVectorIndexMethod) so flows can configure index methods/options.docs/docs/core/flow_def.mdx,docs/docs/examples/examples/simple_vector_index.md: document the newmethodfield and show an IVFFlat example.P.S. most of the changes are schema/SQL plumbing but I couldn't find existing tests covering the vector-index SQL paths (not sure if I missed anything).