Building Reliable Support Agents Using the Effect Typescript Library
summary
Claims production-ready support agents built with the Effect TypeScript library using a React front end, an internal effect RPC server, effect HTTP, and effects SQL talking to a postgress database and vector store. Agents run a custom DSL for planners, duplicate token streams (one to user, one to storage), and use stateful retry policies to fail over from GPT4 mini to GD flash 2.0. Includes a live demo, architecture diagrams, and concrete testing patterns via dependency injection.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:21Our architecture
- 2:27Our agents
- 3:13Our language
- 3:31Reliability
- 4:23Developer Experience
- 5:30Lessons Learned
- 6:35Conclusion
related talks
27mEffect: Production-Grade TypeScript
01h 43mVibe Engineering Effect Apps
46mEffect: A Functional Foundation For TypeScript
48mAI Coding Agents and How to Code Them
33mAgent Experience
52mDesigning Production-Ready Agentic Systems: Lessons from the Field
48mAI Coding Agents and how to code them
23mBuilding AI Agents in TypeScript
