LLM Observability
Evaluation-first monitoring — worst responses surface first
⚠ Regression detected
—
Total Calls (24 h)
—
Avg Quality (24 h)
—
Total Cost (24 h)
—
Bad Response Rate
—
Quality Trend (last 24 h)
Bad Response Categories
Cost ↔ Quality by Model
Root-Cause: Worst Prompts
Worst Responses
Quality
Hall.
Flags
Model
Output preview
Why