@pai/test-utils

runTool, createTestRuntime, runAgentEval.

Package: @pai/test-utils

pnpm add -D @pai/test-utils

Callable	Guide
`runTool()`	Tool tests
`createTestUtils()`	Runtime tests
`createTestRuntime()`	Runtime tests
`runAgentEval()`	Evals

For provider-specific model fixtures, use rawStreamResponse(parts) or createTestUtils().rawStream(parts). Both emit only the supplied stream events; include a finish event when testing normal completion.

Also: stubTools, waitForTool, waitForPending. React UI: @pai/react-test-utils.

officeSkills

Word, Excel, and PowerPoint skills for agents with a sandbox.

runTool()

Execute one tool with no model, no HTTP, no thread runtime.