...
However the cache is not used for Snaps with binary output. Any binary output view stores only 8K of encrypted preview data in cloud and continues to stream remaining data until the downstream document Snap gets enough data to output 50 documents for its preview. Because the size of data needed to generate 50 documents can be more than just 8K, the preview data for the binary output Snap from previous run is not useful. In other words, the 8K data from the cache might not necessarily generate 50 documents in the downstream document output Snap. To handle this case, the caching is enabled only for document Snaps. If there is a binary output Snap preceding the document output Snap and the document output Snap is changed, then the cache is invalidated for the upstream binary output Snap as well.
The runtime info map shows the whether cache is used for a particular suggest run. The use_cache parameter in flags is True when cache is used at pipeline level. The Snaps that use cache have use_cache turned True in their runtime info.
Following are the scenarios when the suggest run does not use cache:
...