Share files between sandbox sessions

Beta

Agent Sandboxes is in beta on Logfire Cloud. If you do not see Agent Sandboxes in your project, see Enable Agent Sandboxes.

Keep files between sandbox sessions and share them with sessions that use the same profile, a named session configuration. A volume stores files separately from Python memory. A mount gives a profile access to a volume at a path such as /data.

Before you start

You need access to Agent Sandboxes and permission to manage the project’s volumes and profiles. Create a profile such as Analysis before you add a mount.

Create and mount a volume

1. Open AI Runtime > Agent Sandboxes > Volumes.
2. Select Create volume, enter a Volume name such as Shared reports, then select Create volume.
3. Open Profiles and select Mounts beside Analysis.
4. Choose Shared reports under Volume and enter /data under Mount path.
5. Choose Read and write under Access, then select Mount volume.

Use Read only when a session should read files without changing them. Use Search volumes and the sort control to find a volume by name in a longer list.

Write and read a shared file

Close the mount dialog and select Use profile beside Analysis. The Workbench is the browser editor for sandbox code.

Volume files are stored by Logfire

File contents written to a hosted volume leave your machine and remain available to sessions with that volume mounted. Keep secrets and sensitive data out of hosted volumes.

In the Workbench, run:

from pathlib import Path

Path('/data/shared.txt').write_text('shared volume bytes')

The result is 19, the number of characters written. Then run:

from pathlib import Path

Path('/data/shared.txt').read_text()

The result is 'shared volume bytes'. Other sessions that use this profile can read the same file.

Snapshots do not roll back files

A snapshot saves Python state. Volume files keep their current contents when you load that snapshot. Use separate files or an application backup when you need an earlier version of a file.

Change or remove access

Open Profiles > Mounts again. Select Edit mount, change the path or access, then select Save changes. Changes apply when the profile next accesses a volume. Select Unmount to remove the profile’s access without deleting the volume’s files.

Volume deletion is permanent

Deleting a volume removes its files. A Python snapshot cannot recover them. Copy any files you need before you delete the volume.

To delete the stored files, open the volume’s … menu in Volumes and select Delete volume. Check the name in the confirmation before you continue.

Verify

Check that the profile’s mount list shows /data with Read and write access. Run the write and read examples, then read the file from a new session with the same profile. It should return 'shared volume bytes'.

Troubleshooting

• Volumes or Mounts is missing. Check that you use Logfire Cloud. Self-hosted Logfire deployments do not include hosted volumes.
• A write fails. Check that the session uses the correct profile and that the mount allows Read and write access.
• A file is missing. Check the mount path. An unmounted or deleted volume is unavailable to the session.

Next steps

Manage profiles to control which sessions share files, or manage saved snapshots to restore Python state.