the sdk exposes a global serializer in BraintrstJsonMapper, but this is too heavy-handed. Different components may wish to configure different serialization options:
- user-supplied objects in eval datasets / remote-scorers
- integration specific mappers (e.g. ai vendor request/response bodies)
- otel serializers
- (TODO: investigate code base and look for other things here)
it's still fine to have a single "global" mapper, but it should be possible to configure different options for the above scenarios
the sdk exposes a global serializer in BraintrstJsonMapper, but this is too heavy-handed. Different components may wish to configure different serialization options:
it's still fine to have a single "global" mapper, but it should be possible to configure different options for the above scenarios