neo4j-genai-plugin-skill

To generate vector embeddings inside a Cypher query, use the `ai.text.embed()` function. This enables direct embedding generation and batch ingestion on nodes during data ingestion without leaving Neo4j.

Yes, you can get structured JSON output from an LLM within Neo4j using `ai.text.structuredCompletion()`. It validates and returns schema-constrained JSON maps directly in your query results.

Yes, calling LLM functions from Cypher requires Neo4j 2025.12 or later with the GenAI Plugin enabled. Queries must also use the CYPHER 25 syntax via a per-query prefix or an ALTER DATABASE default.

The best way to build a pure-Cypher GraphRAG pipeline is using the GenAI Plugin's `ai.text.*` functions. It supports embedding generation, text completion, stateful chat, and token-based chunking entirely within your database workflows.

You configure LLM providers in Neo4j by passing parameterized provider configuration maps to the `ai.text.*` functions. Use lowercase provider identifiers and include your API token and model via query parameters.

Yes, Neo4j supports stateful chat completions with OpenAI or Azure using the `ai.text.chat()` function. This allows for conversational LLM interactions directly within your database queries.