Audit eleven hundred Cloudinary assets in one sitting

Search the whole Cloudinary library at once

An expression like `folder:campaigns/2023/* AND resource_type:image` returns the set, newest first, five hundred at a time with a cursor for the rest. For each hit the get tool returns dimensions, format, bytes, tags, context and the derived transformations generated from it.

A person does that by scrolling a grid and opening assets one at a time. A companion does it for the whole account and hands back a sheet somebody can act on, which is ordinary data extraction.

How do you find the crops nothing requests?

The transformation list.

Named and ad hoc transformations come back with usage counts, so “which of these forty crops does anything still request” has an answer rather than an opinion. That is where the storage on a mature account has actually gone.

Pair it with the account usage read, which returns plan, credits consumed, transformation count, storage bytes, bandwidth and request counts. So “what is large and unused” comes back as a list of public ids rather than a hunch.

What does deleting an asset take with it?

The derivatives.

Deleting by public id also purges the derived assets, so a crop some page still requests stops existing alongside the original. There is no recycle bin here and no undo call, which is why the tool is flagged destructive and confirmed before it runs.

The guidance sits in the tool’s own description: scope by public id rather than by tag. A tag can match assets nobody in the room remembers applying it to, and a companion that has just enumerated four hundred matches is exactly the moment a bulk delete looks reasonable. Deletion belongs after an audit a human read.

Get the audit into a ticket and a sheet

The marketing site that references the files. The sheet the team works from. The ticket that has to name specific public ids before anyone acts. A companion opens the pages in the session you are already signed into, matches public ids against what it found, and files all three in the same sitting.

Work on the pixels stays in Cloudinary, where generative editing, background removal and automatic tagging produce new derivatives from an original. Those need an upload or a transformation write, and this surface is the question-answering half beside them.

Frequently asked questions