Pydantic

Pydantic Logfire introduced “managed variables” / feature-flag style controls for AI engineers to change models, prompts, and tool policies at runtime and connect those changes back to traces and outcomes via observability.

Pydantic AI added native support for Crusoe as a model provider, enabling agent code to use Crusoe models via a Pydantic AI model/provider string while integrating with Logfire tracing.

Pydantic AI Harness added StackOne as a capability so agents can access StackOne’s connected SaaS actions through linked accounts (e.g., Workday, Salesforce, Zendesk).

Pydantic AI added a native Snowflake provider (SnowflakeProvider / SnowflakeModel) to run governed, secure AI agents within Snowflake’s environment.

Pydantic described how Logfire’s AI Gateway provides a single-key provider routing endpoint with guardrails (e.g., secrets/PII scanning), routing groups, and hard spend caps tied to traces.

Pydantic published a “Agents Week” set of articles focused on productionizing agents, including the role of traces, AI gateway controls, annotations, and managed variables for shipping and iterating safely.

Pydantic welcomed Bill Easton as Head of Product, citing his prior product leadership in observability/security (including work related to OpenTelemetry).

Pydantic announced Gregor Stewart joining as an advisor, with a focus on production AI and enterprise adoption experience.