Pydantic

Pydantic AI Gateway: Open Beta Now Live 🚀 https://lnkd.in/emyhdCd9 Introducing Pydantic AI Gateway (PAIG) — AI model plumbing, the Pydantic way. One key, access to all models. Zero translation, requests flow through in the provider's native format. When OpenAI, Anthropic, or Google ship something new, you use it immediately. What you get: - One key for OpenAI, Anthropic, Google Vertex, Groq, and AWS Bedrock - Cost limits at project, user, and API key levels (daily, weekly, monthly caps) - Built-in observability through Logfire or any OpenTelemetry backend - Automatic failover around provider outages - Open source core (AGPL-3.0) with self-hosting option - Enterprise ready: SSO, granular permissions, flexible deployment Free during open beta. No waitlist. Sign up at https://lnkd.in/ev8kr4Fg Blogpost with announcement: https://lnkd.in/echee3Ja #pydanticaigateway #aigateway #pydantic

Join us for a LIVE recording of Insecure Agents next Thursday after AI Engineer Code Summit! With special guests.... Samuel Colvin, CEO of Pydantic Ian Livingstone, CEO of Keycard Kerem Proulx, CEO of Pensar Leonard Tang, CEO of Haize Labs Luma link in the comments. RSVP now to save your spot!

Picture this👇 Your AI agent completes three LLM calls and is running its final task. Suddenly, there's a network timeout. Now it starts over from scratch, re-running every step you already paid for. Alexander S., one of our Staff Software Engineers, just wrote about how to fix this with the Prefect x Pydantic AI integration. The solution is durable execution. When workflows fail, they resume from the failure point instead of starting over. Successful steps get cached, so retries skip completed work. Alex covers: 👉 How caching actually works under the hood 👉 Configuring different retry policies for LLM calls vs tool invocations 👉 Using structured outputs to compose agents into larger workflows 👉 Patterns for deploying agents as production services If you're building with agents and tired of burning API credits on redundant retries, Alex has the fix. 🛠️ Learn how to build a fault-tolerant agent: https://lnkd.in/dxs853iM #pydanticai #AI #LLM

🚀 PyUtrecht — 3rd Edition at Channable, Utrecht 🚀 Join us for an evening of Python, ideas & community — and we’re looking for speakers! 📅 Date & Time: November 4th, 18:30 📍 Venue: Channable, Utrecht 🎤 Want to speak? You have 20 minutes of Python content, a demo, or a topic you’d like to explore? If you are interested, send me a message or comment below. Please share this post — hope to see many of you there! 👩💻👨💻 #Python #PyUtrecht #TechMeetup #UtrechtCommunity

❗️❗️ATTENTION!! This year we’re rolling out a workshop series at AI By the Bay! Pydantic Kicking off with Pydantic's “Building Durable Deep Research Agents — Using PydanticAI, Temporal, and Pydantic Logfire,” led by Samuel Colvin. This session covers production-grade, failure-resilient AI agents. Participants will build a multi-agent “20 Questions” research system using PydanticAI for orchestration, Temporal for durable execution and recovery, and Pydantic Logfire + Pydantic AI Evals for observability and model evaluation. You’ll leave with: • Working code for a fault-tolerant multi-agent agent • A clear model for when to use durable execution • Techniques for observability, debugging, failure recovery without recompute, and model evaluation Requirements: Python 3.10+ (3.12+ recommended), Git, Code editor (VS Code / PyCharm / Zed), and a Pydantic Logfire account (free tier). No LLM API keys needed — inference is provided via the Pydantic AI Gateway. Come prepared: Python fundamentals - async/await, type hints, decorators, context managers; Pydantic Validation basics - models, validation, field types, serialization; LLM concepts - prompts, function/tool calling, structured outputs; Basic async patterns - task groups, concurrent execution, error handling. RSVP 👇