Dynamic, Safe Ops Using AI to Generate Trusted Operations from Text Prompts
summary
Demonstrates turning a huge GraphQL schema into searchable fragments and using RAG to generate executable operations from plain text prompts. Uses LangChain with vector embeddings (olama / nomic or OpenAI), stores vectors in PGVector or MongoDB, and slices the schema by definitions so a 300k+ token GitHub schema becomes relevant 10-document contexts. Shows a working demo that embeds a schema in ~65 lines, generates queries that run in ~5–7 seconds, and notes hallucination risk mitigated by confidence scores and schema comments.
AI-written summary · about
chapters (14)
- 0:00Introduction to talk
- 1:03Developer productivity with APIs
- 2:31GraphQL in platform APIs
- 3:55GitHub API evolution
- 5:19AI and GraphQL challenges
- 6:42AI for developer queries
- 8:51RAG for GraphQL context
- 10:37Demo overview
- 12:43Schema design for AI
- 15:50Embedding GitHub schema
- 18:19Query generation demo
- 21:53Ideal developer experience
- 23:16Safe AI integration strategy
- 23:59Q&A session
related talks
31mGraphQL Embeddings: AI-Powered Dynamic Operations From Schema To IDE
33mOperationalizing GenAI Applications with DevOps Practices
48mGoing Beyond Prompt Engineering
31mRethinking DevOps and DevSecOps with Generative AI: Insights
49mRAG: from dumb implementation to serious results
01h 13mBuilding Production-Ready AI Systems in 90 Minutes
50mRetrieval Augmented Generation Beyond Toy Demos
31mBuild an Agentic RAG system using Langchain, Ollama and Milvus