Omid Saffari Labs OSS
Trace-to-Eval Builder
Paste an agent trace and get replayable eval cases, scorers, and instrumentation gaps.
Input
Trace or incident note
Model
OpenAI Responses
Output
Eval pack
Your key is held in sessionStorage and sent only to this app's same-origin run route.
The key stays masked; output streams back as plain text.
Failure modes
Replay eval cases
Scorers
JSONL seed cases
Instrumentation gaps
Eval pack
Failure analysis, replay cases, scoring checks, and JSONL seeds.
Run a trace to generate an eval pack.