Playground
Chat & judge
Free-form conversation with one candidate, then a multi-judge panel that classifies the transcript and scores it with the matching category rubric. Reopen recent chats to inspect transcripts and judging.
New session
Chat freely with one candidate, then run 3–5 judges. Category locks after the first judging round.
Candidate
No candidate selected.
Judges (0/5)
Choose 3–5 models. Structured outputs preferred; others use the JSON repair path.