You could say a channel is similar to a repository.
In filefilego, storage providers can store data which can be then retrieved by other peers using the file hash. This part of the system is completely off-chain meaning that file hashes are not stored anywhere.
Now here comes the concept of Channels. Let's assume that I am selling my own ML models which are 2TB of data, I could create a channel called "MLModels XYZ" and create different subchannels and categories. In those subchannels you can create entries which are basically posts, or upload files/folders.
Channels are public and if some files are uploaded to a channel, they are first uploaded to a storage provider and then the hash is included in the channels posts/files this way they point back to the file hash (the data will be on-chain)