@jtgi

i was just winging it but made an e2e test suite run n times. I asserted key tool calls invoked with correct args and agent output (using another llm). Maybe @martinamps has more insight on proper evals.

0 reply

0 recast

1 reaction