v0.0.7

Added

• Added TERMINALS_WORKERS and --workers so deployments can run multiple Terminals server processes.
• Added TERMINALS_REPLAY_BODY_LIMIT. Leave it unset for unlimited retry buffering, or set a byte limit when large uploads should stream instead of staying in server memory.

Changed

• Docker mode now adopts an already-running user terminal when another worker finds it, instead of deleting that terminal and starting over.
• Docker mode can now quiet hosted terminal container logs when orchestrator logging is set to WARNING, ERROR, or CRITICAL.
• Worker processes now share terminal activity heartbeats, so active sessions are not mistaken for idle sessions just because another worker served the request.
• Scheduled policy resets now refresh matching running terminals, so a browser tab left open all day does not block scheduled file cleanup.
• Kubernetes operator deployments now group activity updates instead of sending one for every request, which lowers cluster API load while keeping active sessions protected.

Fixed

• Fixed retry behavior for ordinary proxied requests so requests can be replayed after a terminal is replaced. Chunked uploads still stream once because they cannot be replayed safely.
• Fixed Docker multi-worker startup conflicts where two workers could fight over the same user's terminal container.
• Fixed stale activity records being left behind after refreshes, resets, missing terminals, and Docker restart recovery.