Uploaded files and AI processing

When you submit a file for analysis, its content may be sent to our AI processing provider, Google Gemini. Upload only material you are permitted to share with that service. This page describes file handling in the application.

Who can access saved originals?

Saved uploads are available through authenticated links to the uploading account. Staff need explicit private-file access permission; administrative downloads are logged. Server operators also have administrative access. These files are not end-to-end encrypted; hosting, backup protection and provider data practices also affect privacy.

What is saved?

Successfully submitted Research, Animation, Simulation, Learning Path, Exam and Deep Analysis originals are retained in private file storage. Successfully processed voice transcription uploads and camera OCR images are also retained. The database records file paths, ownership and size; it stores extracted study context separately, not the original binary contents. With the VPS storage configuration, originals are on the VPS disk; an operator may instead configure a private object-storage bucket.

File selection and study preview alone do not permanently upload an original. Files rejected before saving are not archived. Files saved only in the browser before this storage upgrade cannot be recovered from database text.

How long are files kept?

Saved originals remain available until you delete their associated content. Deleting a chat or animation immediately disables its file link. The application schedules the stored file for removal after 24 hours. Cleanup runs periodically and retries failures.

Temporary AI files are scheduled for deletion after the generation request. Research context files are scheduled for deletion within 24 hours; follow-up questions can create a fresh provider copy from your saved original. Provider deletion requests do not describe separate provider logs, backups or contractual retention.

Files selected in a browser may also be retained in that browser for draft recovery. Server cleanup does not clear local browser storage.

Voice and camera originals can be removed through the account's authenticated file-deletion endpoint or by contacting an administrator. Deleting an account schedules its saved originals for cleanup. Downloading a file creates a separate copy that application cleanup cannot remove.

Storage and file checks

Saved-file storage is limited to 1024 MB for Free accounts. Pro accounts have no per-account storage quota. Separate per-file, parser and total server capacity limits apply to everyone. Files awaiting deletion continue to count until cleanup succeeds.

Production uploads are scanned for malware and supported document/model parsing is isolated with resource limits. These checks reduce risk but cannot prove that every file is harmless.