The File Object
Your workflows make files from very different places: an email attachment, a generated image, synthesized speech, a parsed document. Inside the workflow, every one of them travels as the same thing: a file object.
What you will learn
Different payloads, identical package
A PDF, an image, and an audio clip arrive from different blocks, each delivered exactly the same way.
The fields never change
Every file object carries an I.D., a name, a URL where the bytes live, a size, a type, and the content, no matter which block made it.
Producers are swappable
Because every producer speaks the same format, you can change which one feeds an agent and the agent side never changes.
Consumers compose
The same object rides to any consumer, into an agent, out as an email attachment, or straight into storage.
The common shape
Catch a file mid-run and look inside: an I.D., a name, a URL where the bytes live, a size, a type, and the content itself. A couple more fields exist under the hood, but the shape never changes: which is exactly what lets a Gmail attachment, a generated image, and a text-to-speech clip all feed the same agent without special handling.
The video closes on this machine: one file object riding the whole chain:
Why one shape matters
Swap which producer feeds the agent, and nothing on the agent side changes. Point the same object at a different consumer, an email, cloud storage, another agent, and it just rides. In the closing machine, an invoice lands in Gmail, gets parsed, an agent extracts the fields, and the result settles in the database: one object carried the whole way through.