Pydantic
@pydantic
The Pydantic Stack: Pydantic Validation, Pydantic AI, Pydantic Logfire, Pydantic Evals, and Pydantic AI Gateway
168 Following 19.7K Followers
Bumped the dspy-monty-interpreter to take advantage of the 0.0.19 @pydantic Monty release, for all your dspy.RLM needs:
Pydantic AI version 2.21.0 is out! 🎉
Pydantic AI Harness v0.14.0 is out! 🎉
Official skills for Pydantic, Pydantic AI, and Logfire, maintained in-repo.
Works with Claude Code, Cursor, Codex, Gemini CLI, and anything on the
Running AI agents in financial services? Sampling production conversations works until a customer complains about the guidance they got.
scores every one, using LLM judges from their Pydantic Evals suite. Learn how they built agents their compliance reviewers can trust:
Show more
Pydantic AI Harness v0.13.0 is out! 🎉
Pydantic AI version 2.20.0 is out! 🎉
Models are pretty good at orchestrating more of themselves. DynamicWorkflows gives it a script for its agents. Hand over a catalog, it writes the Python.
Now on Pydantic AI:
uv add "pydantic-ai-harness[dynamic-workflow]"
Show more
Tomorrow: @samuelcolvin demos Pydantic Monty, a minimal Python interpreter for running the code agents write. For most agents, it beats a full sandbox on complexity, cost, operational overhead, and latency.
Here's your chance to register for the live, online session:
Show more
Pydantic AI Harness v0.12.0 is out! 🎉
Pydantic AI version 2.19.0 is out! 🎉
new post on how to test SQL agents with @pydantic AI.
I'm very proud of that we're building here.
Subprocess isolation plus websocket execution for a Python sandbox runtime is the unglamorous infra work nobody hypes but every agent framework quietly ends up depending on. This is the layer that actually determines if agents are safe to run.
Show more
Pydantic AI Harness v0.11.0 is out! 🎉
Pydantic AI version 2.18.0 is out! 🎉
You'd never let a junior engineer learn Terraform on prod.
Your agent is faster than the junior. In both directions.
Give it a fake AWS it can break: apply, wipe, rebuild until green. Harness Week finale, with @LocalStack.
Show more
Your AI tools spend just outgrew a senior engineer's salary and nobody owns it.
Our latest blogpost makes the case that "cost control" is really an observability problem in a CFO hat.
The fix is a trace, not a cap.
🔗
Show more
