Скачать или смотреть Building AI agents with PydanticAI

In this livestream, we’ll explore how to create and launch AI agents using PydanticAI. Simple, practical, and hands-on — perfect for anyone curious about autonomous AI.

#AI #PydanticAI #Agents #FridayNightBuild

00:00:00 – Intro & welcome
00:00:55 – What are generative AI agents
00:02:10 – How agents think, use tools, and plan goals
00:02:36 – Introducing Pydantic and PydanticAI
00:03:30 – Data validation and structured outputs
00:04:50 – PydanticAI vs CrewAI: coding vs visual studio
00:06:10 – Introducing Logfire observability
00:07:30 – Supported models: OpenAI, Gemini, Perplexity, Anthropic, etc.
00:09:20 – GitHub star comparison: CrewAI, AgnoAI, PydanticAI
00:11:50 – Deep dive into Logfire: tracking AI actions
00:13:40 – Observability and telemetry for AI workflows
00:16:20 – Setting up VS Code for PydanticAI
00:17:30 – Building your first agent in Python
00:18:30 – Streaming vs static responses
00:20:20 – Installing dependencies & OpenAI key setup
00:21:40 – How authentication works (no login needed for PydanticAI)
00:22:10 – Running the first “Tell me about Tokyo” demo
00:23:40 – Creating structured outputs with BaseModel
00:24:30 – Adding “fun facts” automatically
00:25:40 – Connecting to the COTI MCP server
00:27:00 – Stateful vs stateless agents and context memory
00:28:10 – Creating a new COTI account via MCP
00:31:10 – How delegation between agents works
00:33:10 – Fun fact generation & schema guidance
00:34:00 – Using Claude Code to enhance prompts
00:36:10 – Creating APIs with FastAPI for production
00:38:20 – Comparing CrewAI, AgnoAI, and PydanticAI
00:39:30 – Cost optimization using Gemini models
00:41:10 – Hands-off workflow: Manager agent reviewing outputs
00:44:10 – Multi-agent collaboration and quality scoring
00:47:30 – Delegation: agents calling other agents
00:50:20 – Nested tools & structured output examples
00:53:00 – Using Pydantic models to format reports
00:55:00 – Implementing Logfire in Python (live demo)
00:58:10 – Real-time telemetry in Logfire dashboard
01:01:00 – Debugging OpenAI calls & inspecting token usage
01:04:10 – When to use Logfire (advanced debugging)
01:06:00 – Comparing RAG complexity: CrewAI vs PydanticAI
01:07:05 – Final ranking: CrewAI, AgnoAI, PydanticAI
01:08:00 – When to choose each tool
01:09:40 – Target audience: vibe coders vs pro developers
01:10:40 – Wrap-up and key takeaways
01:11:20 – Next week: AutoGen & LangChain deep dive
01:12:00 – Closing remarks