feat: Neo4j vector index method support #1111
Conversation
This commit introduces support for specifying vector index methods (HNSW)
for Neo4j targets.
- Modified `src/base/spec.rs` to derive `Eq` for `VectorIndexMethod`.
- Modified `src/ops/targets/neo4j.rs` to:
- Allow `VectorIndexMethod` to be passed to `IndexDef::from_vector_index_def`.
- Store `VectorIndexMethod` in `IndexDef::VectorIndex`.
- Implement error handling for unsupported `VectorIndexMethod` (IVFFlat).
- Update `SetupComponentOperator::describe_state` to display the method.
- Update `SetupComponentOperator::create` to include HNSW parameters in the Cypher query.
|
@georgeh0 kindly review the PR when you have time. |
|
Thanks @belloibrahv just to check - have you get a chance to bring up neo4j and test it end to end with an example? Here is an example - https://cocoindex.io/docs/examples/knowledge-graph-for-docs#query-and-test-your-index |
Hi @badmonster0, Thanks for the suggestion! I've successfully brought up a Neo4j instance using Docker and identified the |
|
got it and make sense, thanks a lot for setting it up @belloibrahv ! https://cocoindex.io/docs/ai/llm#llm-api-types there's different kinds of api that we support, ollama is completely on-prems free if you have any question to wire it, please let us know! |
@badmonster0, Following your valuable feedback, I've successfully configured the During execution, I encountered a This indicates that the cocoindex library, when generating the Cypher query for Neo4j HNSW vector index creation, is using hnsw.ef_construction and hnsw.m as parameter keys. However, Neo4j explicitly expects these parameters to be prefixed with vector., i.e., vector.hnsw.ef_construction and vector.hnsw.m. This suggests that the internal mapping within the Could you please confirm the intended parameter naming convention for Neo4j HNSW indexes within Once this adjustment is made within the Thank you for your guidance. |
|
@belloibrahv Thanks for testing and debugging! These property names (like |
- Updated docs_to_knowledge_graph example to use Ollama instead of OpenAI - Added alternative configuration comments for OpenAI usage - Enables end-to-end testing without requiring OpenAI API key - Supports the Neo4j vector index method implementation
|
Hi @badmonster0 @georgeh0,
|
src/ops/targets/neo4j.rs
Outdated
| if parts.is_empty() { | ||
| "".to_string() | ||
| } else { | ||
| format!(", {}", parts.join(", ")) | ||
| } |
There was a problem hiding this comment.
This logic is a little bit fragile. A slightly simpler way is to put all index config fields into parts (i.e. for existing ones like "vector.dimensions: {vector_size}" also put into parts. Then we only need a simple parts.join(", ") at last, which will make our logic simpler and more clear.
There was a problem hiding this comment.
@georgeh0 Refactored vector config logic using parts.join as requested feedback. kindly help review.
This commit addresses feedback on the docs-to-knowledge-graph example and Neo4j vector index implementation. - Refactors the docs-to-knowledge-graph example to use Ollama by default, removing the explicit address parameter. - Updates the corresponding README to reflect the change to Ollama. - Removes the unused `_method` parameter from the `from_vector_index_def` function in the Neo4j target.
|
@georgeh0 , I have added the request changes, please help review it |
This PR introduces support for specifying the HNSW VectorIndexMethod when creating vector indexes in Neo4j.
Key changes include:
Eqtrait forVectorIndexMethodinsrc/base/spec.rs.src/ops/targets/neo4j.rs) to:VectorIndexMethodparameters.VectorIndexMethodin Neo4j #1053 (comment).Resolves #1053.