MastraM
Mastra4mo ago
moret

Verbose scorer

Hi! I've been trying to use scorers now that the new version is out. I noticed though that there are two aspects on them which generate a lot of content. First, at least in development locally - I did not double check the production build yet - each scorer logs a large "row is" followed by the entire JSON of data the scorer received and produced. Second, when I run multiple scorers each stores the entire input/output/runtimeContext etc. values, which include remembered messages - which is several times larger than the current output - even though they share the same runId.

Could you help me find out how to make the logging less verbose? Do you think it makes sense to eventually reformat
mastra_scorers
table to reduce the repetition of columns which are shared between runs?
Screenshot_2025-09-10_at_10.53.23.png
Was this page helpful?