Introduce workspace-level goal state persistence, goal prompt injection, and after-turn review handling so an active task can continue until the configured completion conditions are met.
Add a dedicated goal review agent with readonly and active evidence modes, configurable model settings, review prompt, token/turn boundaries, idle-no-tool protection, and progress/completed/stopped events.
Wire goal_mode through task creation, task restoration, compression handoff, runtime user messages, API message sanitization, and tool-call ordering so goal continuations survive long-running tasks and deep compression.
Add Vue UI for arming goal mode from the quick menu, showing running/completed banners, displaying progress metrics, restoring running goal state, and exposing personalization settings for review mode and stop limits.
Include goal mode research notes and default goal review configuration.
Use a two-phase delete flow for conversation list items.
Instead of immediately removing the deleted conversation from the array and relying only on transition-group leave hooks, mark the target conversation as pending deletion first so it can play the left-slide animation, then remove it after the animation completes.
Also keep the list container mounted during delete mode so empty/loading branches do not interrupt the exit transition.
Validation: npm run build --silent 2>&1 | tail -n 20
Implement host-mode multi-workspace concurrency while preserving one active task per workspace. Host terminals are scoped by workspace id, host workspace switching no longer globally refreshes or blocks on tasks, and task APIs now expose workspace/conversation metadata for frontend coordination.
Update conversation navigation so creating/loading conversations during an active workspace task is view-only and does not mutate backend terminal context. This prevents running task output from being pushed into the currently viewed conversation after switching or creating a new conversation.
Add running task awareness to the frontend sidebar, composer, and polling flow. Running tasks can be shown across workspaces, current-workspace active conversations render inline loaders, completed unviewed tasks persist as check indicators across refresh, and completed tasks clear once viewed.
Restore running conversations correctly after switching workspace/conversation or refreshing during an active task. REST polling is rebound only for the matching conversation/task, stale poll responses are ignored, empty assistant placeholders are restored before first content, and completion clears loader/stop states without requiring page refresh.
Adjust host workspace UX so switching workspaces loads the active conversation if one is running, otherwise opens the new-conversation state. Other conversations in a busy workspace are view-only with disabled input instead of blocked navigation.
Validation: python3 -m py_compile server/auth.py server/context.py server/conversation.py server/status.py server/tasks.py; python3 -m unittest test.test_server_refactor_smoke; npm run build --silent 2>&1 | tail -n 12.
This commit introduces a substantial security and runtime architecture update for host mode, with three major goals: (1) enforce OS-level sandboxing by default, (2) support a controlled temporary direct-execution escape hatch for admin users, and (3) eliminate silent fallback behavior that could hide risk.
Backend/runtime changes:\n- Added a unified host sandbox runner (macOS sandbox-exec / Linux bubblewrap+seccomp / Windows WSL2).\n- Converted host terminal creation to sandboxed interactive shells by default.\n- Kept run_command/run_python on the same execution policy and added host execution mode plumbing (sandbox|direct).\n- Added session-scoped direct mode with TTL auto-revert (default 10 minutes), configurable via environment.\n- Added execution mode APIs (GET/POST /api/execution-mode), host+admin gating, status propagation, and rate limiting.\n- Added runtime refresh hooks so tool calls honor TTL expiration and mode transitions consistently.\n- Removed docker->host silent fallback paths: docker startup/runtime failures now fail fast instead of degrading silently.\n- Added host sandbox prompt injection (host-only) to guide agent behavior under permission constraints.
Configuration and policy changes:\n- Added HOST_EXECUTION_MODE_DEFAULT and HOST_EXECUTION_DIRECT_TTL_SECONDS.\n- Added HOST_SANDBOX_MACOS_WRITABLE_PATHS to support user-defined writable path allowlists.\n- Updated macOS sandbox profile generation to use minimal defaults (workspace + tmp + /dev/null) plus explicit allowlist extensions.\n- Updated .env.example documentation for new execution/sandbox controls.
Frontend changes:\n- Extended permission popover to a two-column model (Permission + Execution Environment) for host admin sessions.\n- Added execution mode state/options to app state, fetching, and status synchronization flows.\n- Added execution mode switching action and user feedback toasts.\n- Kept warning emphasis via text styling only (removed warning border per UX request).
Validation:\n- Python syntax checks passed for modified backend modules.\n- Existing smoke tests passed: python3 -m unittest test.test_server_refactor_smoke\n- Frontend production build passed: npm run build
- Prompt user to confirm when switching/creating conversation with active task
- Automatically stop running task after user confirmation
- Filter out events from previous conversation to prevent cross-talk
- Show toast notification after task is stopped
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
