Python data validation library that uses type annotations to parse input, enforce schemas, serialize structured objects, and report actionable errors.

Open-source evaluation framework for defining test datasets, running AI application experiments, and scoring model behavior with custom or model-based evaluators.

Unified gateway for accessing model providers with shared credentials, spending controls, and observability, integrated into the Logfire platform.

Typed Python agent framework for developing model-independent AI applications with structured outputs, tools, dependency injection, and production instrumentation.

OpenTelemetry-based observability platform connecting agent traces, application logs, metrics, and evaluations so developers can investigate behavior across the full software stack.

Experimental, pre-v1 Python interpreter written in Rust for sandboxed execution, exposing approved host functions and enforcing resource limits for agent workloads.