Lightning Talk: Search and Execute with Code Mode Backed by the...
summary
WunderGraph shows a working "code mode" that indexes GraphQL schemas and turns natural language into executable queries: a prompt-to-query pipeline (Yoko/Yuko) that returns JavaScript functions which run in a secure sandbox and perform deterministic ops like map, reduce, and count. Demo uses a getCountry tool and a WonderGPT agent; the team reports the GitHub schema (≈1.5 MB) trims to kilobytes for models, enabling fast search and reliable cron-run workflows that avoid exploding MCP tool chains.
AI-written summary · about
chapters (7)
- 0:00Introduction and icebreaker
- 3:24GraphQL and AI future
- 6:35Code mode and tool registry
- 9:26Why GraphQL for AI tools
- 13:19New workflows with Yuko
- 14:42AI-native GraphQL APIs
- 17:08Schema size and search quality
related talks
40mLightning Talks
31mAccelerating GraphQL API Development with Schema Storming and Code-first Implementation
33mThe Future of Efficiency Is Here: Add Planning to Your Schema!
27mSupergraph Saving Lives
47mLightning Talks (Wednesday) with Andrew Mshar
23mGradually Adopt GraphQL Without Writing any Backend Code
18mUnderstand Your Codebase to Innovate Faster
25mBuild Advanced Search Experiences with a Custom Query Language