v0.0.8

Added

• Added TERMINALS_IDLE_CLEANUP_TIMEOUT_SECONDS so deployments can choose how long idle cleanup waits for one terminal stop or reset before trying again later.

Changed

• After a Terminals restart, recovered Docker and Kubernetes terminals now keep their policy settings instead of falling back to the global defaults.
• After a Terminals restart, recovered terminals keep their last recorded activity time instead of being treated as newly active.
• New Kubernetes terminal pods now save the real policy id, so recovered pods are matched to the right policy even when the policy label is shortened.

Fixed

• Fixed idle cleanup getting stuck behind one slow terminal stop or reset. Timed-out cleanup is left tracked and retried on the next sweep.

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.

v0.0.6

Added

• Added a policy rollout action in the admin UI. It refreshes idle terminals for one policy and reports exactly how many were refreshed and how many active terminals were skipped.

Changed

• Reworked the admin UI into a cleaner, denser light workspace with fewer borders, tighter rows, clearer policy actions, and a policy editor that keeps Save visible while scrolling.
• Made busy terminal traffic lighter to serve by avoiding repeated health checks, repeated Open WebUI login checks, and repeated Kubernetes status writes during normal proxy use.
• Split proxy connection pools by terminal instance so heavy traffic to one terminal does not slow down unrelated terminal sessions.
• Turned off access logs and WebSocket compression by default to reduce noisy runtime overhead.
• Documented the new runtime settings for status caching, token caching, WebSocket compression, and access logs.

Fixed

• Fixed packaged installs and Docker images so the built admin UI is included in the Python package. Visiting the root page now loads the UI instead of falling through to the proxy and returning 401 Missing Authorization header.
• Fixed policy save and delete flows so cached policy and tool details are cleared immediately after a change.
• Fixed HTTP and WebSocket retries so, when a terminal is starting or has been replaced, the proxy re-checks the current terminal before retrying instead of retrying an old address.
• Fixed the active WebSocket connection count so failed connection attempts no longer leave the admin status count permanently too high.
• Fixed retry failure handling so HTTP and WebSocket requests return a clean terminal-unreachable error when a terminal cannot be resolved during retry.

v0.0.5

Added

• Added Kubernetes node selector and toleration overrides for terminal and reset pods.

v0.0.4

Added

• Added a minimal admin UI for viewing terminal status, active sessions, and policies.
• Added policy lifecycle support, including scheduled resets and lifecycle state tracking.
• Added OpenShift-focused security context controls and deployment documentation.
• Added frontend build packaging to the server Docker image.
• Added terminal environment propagation for system prompts and resource metadata.
• Added configurable server and operator log levels.

Fixed

• Fixed Docker backend storage limit handling with a best-effort fallback when the host driver cannot enforce quotas.
• Fixed stale proxy connection handling by retrying once after keep-alive failures.
• Fixed Kubernetes and operator provisioning paths to pass effective policy environment values consistently.

v0.0.3

Update pyproject.toml

v0.0.2

Merge pull request #19 from westbrook-ai/fix-orchestrator-dns-names

fix: ensure container names don't exceed 63-character DNS label limit

v0.0.1

Added

• Multi-tenant terminal orchestrator with Docker and Kubernetes backends.
• Kubernetes operator for terminal custom resource management.
• CLI interface for managing terminals.
• Docker build workflows for orchestrator and operator images (multi-arch: amd64/arm64).