A conference recap from PyAI covering key talks on AI evaluation pitfalls, writing agent-friendly Python code, open source maintainer challenges with AI-generated PRs, MCP server context engineering, a minimal Python runtime for agents called Monty, FastAPI updates including streaming JSON lines support, and Redis semantic caching for generative AI apps.
Table of contents
AI Evals PitfallsBuild Reasonable SoftwarePanel: Open Source in the Age of AIDo developer tools matter to agents?Context Engineering for MCP ServersControlling the wild: from tool calling to computer useWhat's new in FastAPI for AIContext Engineering 2.0: MCP, Agentic RAG & MemorySort: