Open jonnor opened 9 years ago
Processing data could be kept in a "job" type structure, see #25 The request -> result mapping should be a separate datastructure.
request -> result mapping should contain at least:
request id | input url | input content hash | params |
With this info, we could possibly replace our Redis frontend for S3 cache in favor of just checking DB. If something goes out of cache we would delete the line in DB and remove blob from S3.
Right now request are mapped to a certain hash for the output image. However, there is no way to get the request id from the output. This could be useful to analyze which images are being requested/used the most. Processing time, and which API key was used for it can be interesting to track for quotas.