Studio
/
Evals
/
New
— left
·
Add
Composer
Save
Test
Description
Evaluator
Provider
OpenAI
Anthropic
Model
Rubric
Pass the conversation if the assistant follows its system prompt, stays on-topic, and answers the user accurately. Fail if it hallucinates, ignores instructions, or refuses inappropriately.
Mode
assistant
squad
Conversation Turns
0 turns
Assistant
Squad
Loading assistants...
No turns yet — add the first user message to start the conversation.
Add User
Add Assistant
Add Tool Response
Generate conversation with AI