Vibeleaderboard
← All Intel
Intel / video

Agent Engineering with Pydantic + Graphs — with Samuel Colvin, CEO of Pydantic Logfire

Source
youtube.com
Author
Latent Space
Date

Context

Pydantic's founder Samuel Colvin explains in a Latent Space interview why a Python validation library became convenient for AI tooling: it can generate JSON from type-annotated Python classes, the format model APIs use to describe inputs and structured outputs. Colvin says Pydantic is best described as a validation library that also converts data (a string reading 123 becomes the integer 123 by default), started in 2017, before language models were common. JSON Schema support was added for tooling. Colvin does not know why OpenAI picked JSON Schema, but says it let one definition serve both validation and structured outputs.

Colvin also explains why the company built its own agent framework, Pydantic AI: many agent frameworks, Colvin says, fall far below the engineering standards of the wider Python ecosystem, and every documentation example in Pydantic AI is run as a test. These are the views of a framework builder.

Terms in this piece · Glossary
  • Schema — A written description of a data format: which fields exist, what kind of value each holds and which are required.
  • Tool — An operation that the software around a model makes available for the model to request, with a name, expected inputs and a kind of result.
  • HTTP and JSON API — A common way for software to request data or actions from another service.
  • Large language model — A model trained on large amounts of data to process and generate language.
Read the source www.youtube.com
More from Latent Space
Recommended reads
Comments

Checking sign-in…

Loading comments…