EvalView - AI Agent Testing

Version updated for https://github.com/hidai25/eval-view to version v0.1.4.

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

Go to the GitHub Marketplace to find the latest changes.

Release notes

What’s New

Ollama Support (Free Local Evaluation)

  • Ollama as LLM-as-judge - Run evaluations locally with zero API costs
  • Auto-detection - Automatically detects Ollama running on localhost:11434
  • New adapter - Test LangGraph agents powered by local Llama models
# Free local evaluation
evalview run --judge-provider ollama --judge-model llama3.2

Improved Hallucination Detection

- Reduced false positives for local models
- Unit conversions and formatting no longer flagged as hallucinations
- Confidence threshold: 90% for Ollama, 70% for cloud providers

README Updates

- Added "Who is EvalView for?" section
- Added LangSmith/Langfuse complement positioning
- New Ollama example in /examples/ollama/

Fixes

- Fixed mypy type annotation error
- Fixed action.yml description length for Marketplace

Full Changelog: https://github.com/hidai25/eval-view/compare/v0.1.3...v0.1.4