Neo4j 2026.02.0 introduces a new `ai.text.structuredCompletion` Cypher procedure that returns AI-generated output conforming to a user-defined JSON schema. The post demonstrates this with a practical dinner planner app: a structured recipe is generated via OpenAI using user preferences, allergies, and recent meals, then stored as a graph with ingredient nodes, vector embeddings, and relationships. The resulting graph enables weekly dinner planning, shopping list generation, and even vector-similarity-based user matching. The post also covers the new CYPHER 25 vector search syntax and notes limitations like unit normalization in shopping lists.

16m read timeFrom medium.com
Post cover image
Table of contents
Predictable randomnessNew Cypher AI proceduresAlea iacta estRandom NumberA recipe generatorGet Christoffer Bergman’s stories in your inboxGitHub - ChristofferBergman/recipe-viewerCypher 25Disclaimer/Warning

Sort: