Support tooling
Triage runs
Every scored run of the triage pipeline against the labeled dataset. Runs with the same prompt hash used the same system prompt, so their numbers compare fairly.
Compare two runs →| Started | Provider | Model | Split | Category | Severity | MAE | Failed | Tokens | Latency | Prompt |
|---|---|---|---|---|---|---|---|---|---|---|
| 2026-10-04 17:23 UTC | openai | gpt-5.5 | dev | 45% | 73% | 0.27 | 0/22 | 46.3K / 13.1K | 8.0s | ef0017a |
| 2026-10-04 17:20 UTC | anthropic | claude-sonnet-4-6 | dev | 68% | 77% | 0.23 | 0/22 | 54.6K / 10.6K | 11.9s | ef0017a |