I am trying to run the llm-graph-builder, GitHub - neo4j-labs/llm-graph-builder: Neo4j graph construction from unstructured data using LLMs , locally with Ollama language models. I have the project set up with the frontend, backend and local neo4j database working.
Now i want to define a schema in the “Graph Enhancement” tab, but there is no explicit option for importing a manually made schema - I tried building one with the schema builder mentioned here User Guide: Knowledge Graph Builder — neo4j-graphrag-python documentation . If there is a way to import such a schema any guidance would be appreciated. Extracting a schema from text with my Ollama model fails with the exception in the backend Exception: with_structured_output is not implemented for this model.I tried using llama3.3:70b and llama4:scout with 109B, but it seems they were not even loaded by ollama yet.
So it seems i can circumvent this problem by using Neo4j Aura (DataServices/Import section in aura) to draw up a schema, download the model of that schema and import it in the llm-graph-builder under GraphEnhancement/AddSchemafrom → choose Data Importer JSON.
1 Like