Add guardrail capabilities: InputGuardrail, OutputGuardrail, CostGuard, ToolGuard#182

Open

DouweM wants to merge 13 commits into

mainfrom

capability/guardrails

Commits on Apr 2, 2026

Add guardrail capabilities: InputGuardrail, OutputGuardrail, CostGuard, ToolGuard

DouweM
and
claude
committed
Fix rebase onto main: resolve conflicts, fix trio test failures, achieve 100% coverage

DouweM
and
claude
committed
Add AsyncGuardrail capability with concurrent/blocking/monitoring modes

DouweM
and
claude
committed
Fix coverage: add tests for async context guard and model-finishes-first path

DouweM
and
claude
committed
Replace inspect.isawaitable checks with TypeGuard helpers

dsfaccini
and
claude
committed
Add InputGuardrail test cases: empty string, non-string prompt, multiple guardrails, async warn mode

dsfaccini
and
claude
committed
Add OutputGuardrail tests: multiple guardrails and async context_guard warn mode

dsfaccini
and
claude
committed
Extend test coverage: CostGuard boundaries, ToolGuard overlap, AsyncGuardrail context modes, exception propagation

dsfaccini
and
claude
committed
Add examples dependency group and directory structure

dsfaccini
and
claude
committed
Add guardrail example scripts and configure pyright/ruff for examples

dsfaccini
and
claude
committed
Fix examples: add logfire instrumentation, fix usage call, use direct OpenAI provider

dsfaccini
and
claude
committed
Add logfire.span labels to example scripts for trace identification

dsfaccini
and
claude
committed

Commits on May 11, 2026

Merge origin/main: port guardrails into renamed pydantic_ai_harness layout

dsfaccini
and
claude
committed