EvalView - AI Agent Testing

Version updated for https://github.com/hidai25/eval-view to version v0.1.8.

  • This action is used across all versions by ? repositories.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

EvalView is a regression testing framework for AI agents, designed to detect behavioral issues, such as tool changes, output differences, cost increases, or latency spikes, before deployment. It automates the process of comparing current agent performance against a golden baseline and integrates into CI pipelines to block problematic deployments. This ensures that agents maintain consistent performance and reliability while enabling fast iteration on prompts, models, and tools.

Release notes

What’s New

uv Support for Faster Development

  • Added uv as an optional package manager for 2x faster installs
  • Both pip and uv are fully supported - use whichever you prefer
  • New Makefile targets: make install (uv) and make pip-install (pip)

CI Hardening

  • Config file support for ci.fail_on and ci.warn_on
  • Exit code 2 for execution errors (network, timeout) vs exit code 1 for test failures
  • Model/provider validation to catch mismatches early (e.g., llama3.2 + openai)

LangGraph Example

  • Added complete LangGraph example agent at examples/langgraph/agent/
  • Includes ReAct agent with search and calculator tools

UX Improvements

  • Diverse quickstart test cases (division, weather, multi-tool, multiplication)
  • Demo agent now supports division operations
  • GitHub star CTA always shows after quickstart
  • evalview connect shows actual paths instead of placeholders

Bug Fixes

  • Fixed list builtin being shadowed by CLI command
  • Fixed demo agent division support

Installation

pip install evalview==0.1.8

Full Changelog

https://github.com/hidai25/eval-view/compare/v0.1.7...v0.1.8