Chunked uploads through the API now carry an upload ID

Anthias used to keep a file that was still uploading under a name based on its file name. Two uploads of files with the same name, at the same time or after an earlier one was interrupted, could end up mixed into one corrupt file. From Anthias 2026.07.2, every upload gets its own temporary file, whatever it's called.

Uploads from the web interface need nothing from you. The change is for integrations that upload through /api/v1/file_asset or /api/v2/file_asset in chunks with a Content-Range header. Anthias can no longer tell from the file name which chunks belong together, so it gives each upload an ID instead:

• The response to every upload now includes upload_id.
• Send that value as an X-Upload-Id header with every later chunk of the same file, and the chunks are put together as before.
• A chunk without X-Upload-Id starts a new upload. A malformed X-Upload-Id gets a 400 response.

If you upload in chunks through the API: read upload_id from the response to the first chunk and send it as X-Upload-Id with each chunk after it. Without it, each chunk is stored on its own and the asset you create from the last chunk's uri comes out broken.