When SSHing into a VM with network latency, terminal responses to startup
queries (kitty protocol detection) can arrive after the 200ms timeout expires.
The original code immediately restored raw mode, causing late responses to
leak through as visible text.
This fix adds two layers of defense:
1. Drain handler in kittyProtocolDetector: Adds a 100ms window after timeout
to silently consume late-arriving responses
2. Regex filter in KeypressContext: Catches any terminal response patterns
that make it past the detection phase, regardless of timing
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Add null/undefined guard in formatBtwError to avoid "null"/"undefined" strings
- Add type guard for btw property in HistoryItemDisplay to prevent crash
- Extract isBtwCommand regex to module-level constant and simplify with [/?]
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This unifies the parameter naming convention across file-related tools
(edit, write_file, read_file) to consistently use file_path.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Add btwItem state management independent from pendingItem
- Add cancelBtw functionality to abort in-flight BTW API calls
- Allow /btw commands to execute concurrently with main responses
- Add isBtwCommand utility function
- Update BtwMessage UI with cleaner styling (remove spinner)
- Add tests for concurrent /btw execution scenarios
- Update layouts to render BTW messages in fixed bottom area
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Simplify export data by removing filesRead stat, keep only written files count and paths
- Restore lines-related statistics (linesAdded and linesRemoved)
- Update HTML display to show only file operation stats instead of total files count
- Change 'Written' label to 'Files modified'
- Remove distinction between requestId and sessionId, always display sessionId
- Remove Session ID and Export Time from Header (already shown in MetadataSidebar)
- Display Project field with raw value and support multiline display
- Fix filesWritten calculation to count unique files instead of operations
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Fix four accuracy bugs in export metadata/sidebar feature:
1. File read counting: Now properly counts read_file operations by checking
functionResponse.name and args.absolute_path, instead of relying on
resultDisplay which returns string for reads.
2. Unique file tracking: Uses full file path from args.file_path or
args.absolute_path instead of basename-only fileName, preventing
collision between same-named files in different directories.
3. TaskTool token aggregation: Includes tokens from TaskTool executionSummary
in total token count, fixing under-reporting when subagents are used.
4. Context window display: Removes hardcoded '128k' fallback in HTML sidebar,
now only displays context usage when contextWindowSize is actually defined.
Also fixes lint errors (Array<T> type annotations) and applies formatting.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Added candidatesTokens prop to LoadingIndicator for displaying token counts.
- Updated formatting to show elapsed time and token counts inline.
- Refactored tests to validate new token display functionality and formatting changes.
- Introduced formatTokenCount utility for consistent token count representation.
This improves user feedback during loading states by providing clearer information on token usage.
- Extract shared BaseTextInput component with readline keyboard handling
- Add AgentComposer and AgentFooter components for agent interaction
- Add useAgentStreamingState hook for managing agent streaming state
- Refactor InputPrompt to use BaseTextInput with agent tab bar focus support
- Move calculatePromptWidths to shared layoutUtils
- Disable auto-accept indicator on agent tabs (agents handle their own)
This enables a dedicated input experience for agent tabs with proper
focus management and keyboard navigation between main input and agent tabs.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Introduce AgentStatus.IDLE for agents that finished work but can accept follow-up messages
- Add isSettledStatus() helper to check if agent is settled (IDLE or terminal)
- Update ArenaManager to transition to IDLE after agents finish initial task
- Keep agent tabs visible when session is IDLE so users can continue interacting
- Fix listener cleanup to not detach on IDLE (agents remain alive)
- Update tests to expect 'idle' status after successful completion
This enables the arena collaboration feature where agents can receive
additional tasks after completing their initial work.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Remove acp.ts and schema.ts in favor of SDK types
- Refactor acpAgent.ts to leverage SDK client
- Update session management types and implementations
- Adjust all test cases for new SDK-based architecture
- Update integration tests and export utilities
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
The colorizeCode function signature was changed to have tabWidth as the
last parameter with a default value. Update all call sites to match.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Moves export-html and insight templates from cli/assets to a new
dedicated web-templates package. Updates Dockerfile and build scripts
to use consolidated bundle/prepare:package/pack workflow.
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
Add InProcessBackend to run subagents in-process rather than via subprocess,
enabling faster initialization and better resource management for agent
collaboration arenas.
Key changes:
- Add InProcessBackend with sandboxed in-process agent execution
- Refactor agent runtime into headless vs interactive modes
- Add AsyncMessageQueue utility for agent message passing
- Update ArenaManager with backend selection (in-process vs subprocess)
- Refactor subagent types/exports; consolidate in subagents/types
- Remove deprecated agent-hooks.ts (functionality merged into runtime)
- Update task tool to support new agent lifecycle
Breaking: Subagent type exports restructured; import from subagents/types
Introduces a new Arena system for running multiple AI agents in parallel
terminal sessions with support for iTerm and Tmux backends.
Core:
- Add ArenaManager and ArenaAgentClient for orchestrating multi-agent sessions
- Add terminal backends (ITermBackend, TmuxBackend) with feature detection
- Add git worktree service for isolated agent workspaces
- Add arena event system for real-time status updates
CLI:
- Add /arena command with start, stop, status, and select subcommands
- Add Arena dialogs (Select, Start, Status, Stop)
- Add ArenaCards component for displaying parallel agent outputs
- Consolidate message components into StatusMessages and ConversationMessages
- Add MultiSelect component for agent selection
Config:
- Add arena-related settings to schema and config
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Restore plan update handling in collect.ts, formatting todo data as markdown checklist
- Skip todo_write tool_result processing in normalize.ts to avoid duplicates
- Add getMessageTimestamp() and getMessageUuid() methods to maintain message order
- Fix version parsing to support @latest tags
- Fix Windows spawn EINVAL error (CVE-2024-27980)
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
- Replace ConsolePatcher with centralized debugLogger utility
- Refactor errorReporting to use debugLogger instead of file-based reporting
- Remove user-facing console message components:
- Delete ConsolePatcher.ts, useConsoleMessages.ts/hook
- Delete ConsoleSummaryDisplay.tsx, DetailedMessagesDisplay.tsx
- Update all tests in packages/core and packages/cli:
- Mock debugLogger where needed
- Remove assertions for console output on non-critical errors
- Keep debugLogger assertions for fatal/network errors
- Use HOME directory mocking for hermetic file system tests
Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>
