What problem does it solve? Calling LLMs for embeddings, completions, or structured extraction normally requires external Python pipelines and glue code. This Skill lets you invoke OpenAI, Azure OpenAI, VertexAI, and Amazon Bedrock models directly inside Cypher via the Neo4j GenAI Plugin's ai.text.* functions, keeping GraphRAG workflows entirely in the database. ## Core Features & Use Cases - In-Cypher Embeddings: Generate single or batch vector embeddings with ai.text.embed() and ai.text.embedBatch(), storing results directly on nodes. - Completions & Structured Output: Run ai.text.completion(), aggregate summaries over grouped rows, and extract JSON Schema-validated maps with ai.text.structuredCompletion(). - Chat & Tokenization: Maintain stateful chat sessions via chatId and chunk text by token limits without external dependencies. - Use Case: Build a pure-Cypher GraphRAG query that embeds a user question, runs vector search, traverses related articles, and returns an LLM-generated answer in a single statement. ## Quick Start Ask your assistant to write a Cypher query using ai.text.embed to generate embeddings for Chunk nodes missing an embedding property, using the OpenAI provider with the API key passed as a parameter.
