Metrics for request durations currently have a hardcoded histogram bucket size, which is currently tailored towards eigenda latencies).
Would be nice to let users set them as needed, such that different configurations of the proxy server could use different buckets sizes (memstore, s3, and eigenda backends have very different latencies).
See https://github.com/Layr-Labs/eigenda-proxy/pull/94#discussion_r1723900596
Metrics for request durations currently have a hardcoded histogram bucket size, which is currently tailored towards eigenda latencies). Would be nice to let users set them as needed, such that different configurations of the proxy server could use different buckets sizes (memstore, s3, and eigenda backends have very different latencies).