This message was deleted.

Hello. There are two scenarios: 1. If you want to log the meta-information of an image, such as size, color distribution, etc., you can use the Bentoml.monitor API. Bentoml.monitor supports recording numerical and text types. You can manually extract the desired information and then record it using the API. 2. If you want to log the binary file of the image itself, Bentoml currently does not have a dedicated optimization for this, but there are many mature solutions. A common solution is to import boto3 in your service.py and use the https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/s3.html#S3.Client.upload_fileobj API to upload it to an S3 storage, and then put it into a data lake or other ultimate destination through subsequent pipelines. There are some optimization points for using this solution: 2.1 You can deploy an S3 service, such as MinIO, in the local network where you deploy the Bento service. This can greatly reduce the burden of network transmission. 2.2

boto3.upload_fileobj

is a synchronous call, which may affect the response time of your online service. If you care about response time, you can put the call to this API in a thread pool.