Assevra
What it is
Assevra is a Python library and CLI tool that generates signed, statistically defensible reliability scorecards for LLM agents. It scores agent traces across nine dimensions (including grounding, safety, and PII) using deterministic checks and confidence intervals, producing portable artifacts for release gating.
The problem it solves
Lack of reproducible, statistically rigorous, and portable evidence for LLM agent reliability, which often relies on subjective or non-deterministic evaluations that are hard to verify over time.
What you could use it for
- Gating CI/CD pipelines based on agent reliability metrics
- Auditing agent outputs for PII leakage and safety violations
- Generating reproducible evidence of agent performance for security reviews
- Measuring tool-calling accuracy against defined specifications
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.