Spring-AI embedding, cannot search result by method...

bake_betty

I'm using Spring ai to build an app, the stack I chose is jdk 21, spring-boot 3.3.11 ,spring-ai 1.0.0M6. And for both embedding and chat models, I chose to use Ollama, thus the spring-ai-ollama-spring-boot-starter is implemented, and for vector saving, I chose to use neo4j. The chat and embedding text and save the result to neo4j modules are working finely, while I had some problem of search content via method vectorStore.similaritySearch which...

Originally asked by xiong on Stack Exchange (stackoverflow) (CC BY-SA 4.0).

uphilltech

Great question — answering as the UphillTech team:

Check the data type of your vector column. Dimensions must match with the embedding model you are using otherwise search will simply not work. VectorStore defaults to 1536 dimensions as per official docs. You need to set a specific property in vector store configs to set the dimensions to 768. Update: Seen your code and this property is causing the issue: spring.ai.vectorstore.neo4j.dimensions=1536 // change to 768

Full answer by saran3h: https://stackoverflow.stackexchange.com/a/79575828 (CC BY-SA 4.0).

Need it done for your business? UphillTech.