What counts as an execution?
- Running a prompt
- Running a step in a flow
- Running an offline eval in Datasets
- Running a dynamic column or experiment in Datasets (each generated cell counts as a separate execution)
Every execution uses 1 credit, irrespective of the number of tokens
processed, or where the execution is run.
Example: if you run a dynamic column on a dataset with 50 rows, it will
use 50 credits.
What does not count as an execution?
- Logs
- Online evals
- Annotations