May 13
Loading video player...
Sebastián Ramírez (tiangolo), creator of FastAPI, covers the latest updates to FastAPI built specifically for AI and streaming use cases. Recorded live at PyAI Conference 2026 in San Francisco. FastAPI now has 95K GitHub stars and ~280 million monthly downloads. In this talk, Sebastian walks through what's new for developers building AI-powered APIs today. What's covered: 00:00 Intro and FastAPI by the numbers 01:41 pyproject.toml support for app configuration 03:57 The new VS Code / Cursor extension (endpoint explorer, code lens) 05:15 LLM expectations vs. reality 07:57 High-performance JSON serialization with Pydantic (Rust-side, 2x speedup) 10:07 Streaming data with yield and JSON lines 12:01 Streaming pure bytes 13:33 Server-sent events (SSE) with EventSourceResponse 17:29 FastAPI's bundled skill file for AI agents 19:44 Live demo: streaming LLM responses with Pydantic AI and server-sent events 21:30 FastAPI Cloud waiting list Key topics: FastAPI streaming, server-sent events in Python, JSON lines API, Pydantic serialization performance, FastAPI VS Code extension, FastAPI for LLMs, Pydantic AI agent demo Follow Sebastián: https://x.com/tiangolo FastAPI Conf: https://fastapiconf.com/ PyAI Conference is organized by Pydantic and Prefect. Learn more at https://pyai.events