Where the bytes live

The file is recorded before the upload finishes. A rename does not ask you to upload it again.

The file row is the record. ContextDrive stores the bytes. You do not choose a store.

Creating an organization names a location: us_east, us_west, eu, or apac. v1 offers us_east. A location this deployment does not offer is refused. A shared drive uses the organization’s location.

The path you see is the directory path and the file name. A rename or a move changes that path. You do not upload the file again.

Upload

The file is recorded before the bytes are confirmed. Until then the inspector would say Pending, and the file stays off the list. When the size matches, Status says Ready. Search may still say Indexing.

Bytes can be written only for a file that already has a row. Upload and download is the task in Drive. The grant is on the HTTP reference.

Download

Download uses a short-lived link, issued only after you are allowed to see the file. The saved name is the name on the file, including the extension. A rename changes the name on the next download. The record of the decision does not include that link.

Search reads one index for the drive. The index is written after a file is Ready, and again when a note on a file or a directory changes. A drive whose index has not been written yet returns no hits. There is no download for the index.

The scan keeps the files and the directory notes you may see. A hidden file is absent. Notes on a hidden directory are absent. The path on a citation is the live path, read when the result is shown. A rename, a move, or a tag change leaves the index where it is.

A new note, an edit, or a delete joins the next index. Until that index is written, search can still return the previous text of something you may see.

A document is split into passages. Notes on that file are additional text beside those passages. Notes on a directory are how search finds the directory. A table stays the bytes SQL reads. Search can still find it from the file name, the columns, and one passage per row. Measures stay out of that row text. CSV, Parquet, and a top-level flat JSON array are the table files.

Deleting a file hides it and drops its place in the next index. The tombstone is written only after the bytes are gone. The index can be rebuilt from the file and from the notes.