Weir - unit test for your AI agent - Deterministic tests for AI agents — no LLM required

Weir turns OpenTelemetry GenAI traces into deterministic, machine-checkable evidence of what an AI agent actually did. Measure how observable your agent is, trace data flow across steps, and produce witness paths for CI and security testing - with no LLM in the analysis loop. Built for engineers working on agent evals, observability, security, and CI. check it out! https://github.com/IdoGol24/weir