Skip to content

v0.5.1 — Optional LLM Semantic Judge + Dark Dashboard

Choose a tag to compare

View all tags
@VaradDurge VaradDurge released this 01 Jun 15:57
v0.5.1
84e03ed

What's New

Optional LLM Semantic Judge

  • New semantic_judge=True parameter on ArgusWatcher — enables GPT-4o to analyze every node's output for subtle quality issues (wrong tone, unhelpful responses, outdated info)
  • Runs after deterministic checks — so it only adds cost where structural checks can't decide
  • Configurable model via judge_model="gpt-4o" (default)
  • Requires OPENAI_API_KEY in environment
watcher = ArgusWatcher(
    semantic_judge=True,     # LLM quality checks
    record_http=True,        # deterministic reruns
    judge_model="gpt-4o",   # model selection
)

Dark Dashboard Theme

  • Full dark theme across all dashboard components
  • 30% reduced color saturation for a professional monitoring look
  • All 50+ components updated

Other Changes

  • Model name removed from AI investigation display (CLI + web)
  • Guide updated with combined optional parameters documentation
  • README updated with semantic judge section

Full Changelog: v0.5.0...v0.5.1

Assets 4
Loading