Explains what the Conductor Profile is, how to configure it, what to
include, and how agents use it. Previously only had brief parenthetical
mentions without dedicated documentation.
Display "Ctrl+Enter" on Windows/Linux instead of macOS-specific
"Command+Enter" and "⌘ + Enter" symbols. Uses isMacOS() helper from
shortcutFormatter.ts to detect platform and render appropriate text.
- Update ToastContext to compute willTriggerCustomNotification before logging
- audioNotification.enabled now only true when notification WILL be sent
- Add reason field to explain why notification was skipped (disabled, no-command, opted-out, no-content)
- Bump notification skip log from debug to info level in IPC handler
- Update test to match new audioNotification format with reason field
Add Task Recall documentation to both wizard prompts (wizard-system.md
and wizard-inline-system.md), enabling AI to access session history for
context about previously completed work. Extended documented fields to
include fullResponse, elapsedTimeMs, and contextUsage for detailed
context retrieval.
Updated wizardPrompts.ts and inlineWizardConversation.ts to pass
historyFilePath through template context. Modified useInlineWizard.ts
to fetch history file path during wizard initialization and store in
state for use across conversation session creation points.
Introduces an "About Me" field in Settings → General that allows users
to describe their background, preferences, and communication style.
This profile is exposed as {{CONDUCTOR_PROFILE}} template variable and
included in AI chat, group chat, and wizard prompts so agents can
tailor their responses to the user.
- Rename wizard header from "Create a Maestro Agent" to "New Agent Wizard"
- Add info tip box explaining in-tab wizard alternative (/wizard command, wand button)
- Only show tip box when user has existing agents (first-time users won't see it)
- Update note copy: "The new agent wizard captures application inputs until complete"
- Add final sentence: "The in-tab wizard runs alongside your other work"
- Move "Select the provider" text directly above agent tiles
- Center note box vertically with equal spacing
- Update test mocks with sessions.getAll and sshRemote.getConfigs
When a tab is renamed (manually or via auto-generation), its width may
change. Added activeTabName as a dependency to the scroll-into-view
effect so the tab bar auto-scrolls to keep the active tab (including
close button) fully visible after name changes.
- Daily backups created on app startup (stats.db.daily.YYYY-MM-DD)
- Automatic rotation removes backups older than 7 days
- On corruption, iterate through backups to find valid one to restore
- Add getAvailableBackups() and restoreFromBackup() public methods
- Backups validated with integrity_check before restoration
- Legacy timestamp backups also recognized for restore candidates
When using stream-json output mode on Windows, the prompt was being added
both as a CLI argument AND sent via stdin, causing the command line to
exceed cmd.exe's ~8191 character limit and resulting in immediate exit
code 1.
Now detects when stream-json is in the args and ensures the prompt is
sent only via stdin, avoiding the command line length limit.
Fixes#301
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Added IPC handler to fetch earliest stats timestamp for smarter ranges 🕰️
- Implemented StatsDB query to compute oldest event across all tables 🧮
- Introduced IPC stubs for stats initialization/reset notifications groundwork 🧱
- Added “New Agent” action to hamburger menu with proper shortcut hint ➕
- Reordered menu to separate New Agent, Wizard, and Tour actions clearly 🧭
- Upgraded menu links: dedicated Website and Documentation entries with icons 🌐
- Polished external-link UX by showing an explicit external-link indicator 🔗
- Restored “Check for Updates” as a focused menu action at the bottom 📥
When Codex reads prompts from stdin, it outputs "Reading prompt from
stdin..." followed by the actual response to stderr. This caused the
response to appear in red STDERR blocks in the UI.
Fix: Detect and strip the "Reading prompt from stdin..." prefix for
Codex sessions, then emit the actual response content as regular data
instead of stderr.
Added StderrHandler tests covering:
- SSH informational message filtering
- Codex stdin response extraction
- Buffer accumulation
Shows a modal on first launch for Windows users explaining that Windows
support is actively being improved. Features:
- Inline toggle to enable beta updates for latest bug fixes
- Link to report issues on GitHub (with note that vetted PRs welcome)
- Link to join Discord Windows-specific channel for community support
- Create Debug Package button for easy bug reporting
- Checkbox to suppress the modal for future sessions
- Console debug function: window.__showWindowsWarningModal()
- Add shrink-0 to AI Tab and FileTab containers to prevent flex
compression, ensuring tabs maintain natural width for all content
including extension badge and close button
- Use double requestAnimationFrame for scroll-into-view to ensure
DOM has fully rendered (including conditional close button) before
measuring tab width for scrolling
- Replace scrollIntoView with manual scrollLeft calculation to ensure
the ENTIRE tab (including right edge with close button) is visible,
not just enough to show some portion of the tab
- Update scroll behavior tests to match new implementation
The AgentSelectionScreen now uses Info and Wand2 icons from lucide-react
(added for the inline wizard tip box). Update the test mock to include
these icons so tests don't fail with missing export errors.
When pasting content, automatically trim leading and trailing whitespace
to reduce manual cleanup. The trimming only intercepts paste when there
is actual whitespace to remove, otherwise native paste behavior is used.
Input areas updated:
- Main AI/terminal input (App.tsx)
- Group chat input (GroupChatInput.tsx)
- Prompt composer modal (PromptComposerModal.tsx)
- Auto Run document editor (useAutoRunImageHandling.ts)
- Wizard document editor (DocumentEditor.tsx)
- Inline wizard document generation (DocumentGenerationView.tsx)
- Move "Select the provider" text directly above agent tiles for better flow
- Add info box explaining keyboard capture and alternatives (/wizard command, wand button)
- Display inline Wand2 icon for visual clarity
- Center note box vertically with equal spacing above/below
- Update test mocks to include Info, Wand2, and AlertTriangle icons
- Move "Select the provider" text directly above agent tiles for better flow
- Add info box explaining keyboard capture and alternatives (/wizard command, wand button)
- Display inline Wand2 icon for visual clarity
- Update test mocks to include Info, Wand2, and AlertTriangle icons
Add new Custom Configuration section documenting custom CLI arguments and
environment variables with screenshot. Rename navigation group to "Providers
& CLI" and reorder to list provider-notes first.
Enable stream-json mode for SSH stdin script execution. When SSH wraps
agent commands, the args being checked (SSH args like -o BatchMode=yes)
don't contain 'stream-json', causing JSON output to bypass the parser.
Added sshStdinScript to isStreamJsonMode detection so remote agent
output is properly parsed through the JSON output parser.
- Add addBreadcrumb() utility to track user actions before crashes
- Add startMemoryMonitoring() with configurable threshold and interval
- Record breadcrumbs at agent spawn/kill operations
- Auto-start memory monitoring when Sentry initializes (500MB threshold, 1min interval)
This provides context in crash reports to help identify patterns
leading to renderer crashes (MAESTRO-5A/4Y).
Relates to MAESTRO-5A, MAESTRO-4Y
Add try-catch wrapper and isDestroyed() check to prevent "reply was
never sent" errors when the window closes during dialog operations.
Fixes MAESTRO-58
Wrap console output in try-catch to gracefully handle EPIPE errors
that occur when a parent process consuming output dies unexpectedly.
Fixes MAESTRO-5C
- Symphony modal now receives sessions list for richer contribution context 🧩
- Active contributions show clickable session name with new Terminal icon 🖥️
- Added “navigate to session” flow directly from contribution cards 🧭
- Selecting a session updates active session and auto-closes modal ⚡
- Session lookup wires contributions to matching session IDs seamlessly 🔗
- Renamed empty-state action from “Select Folder” to “Change Folder” for clarity 🧭
- Refreshed empty-state helper text to match the new folder-change flow 📝
- Cleaned up `EditGroupChatModal` icon imports by dropping the unused `X` 🎛️
- Symphony IPC now validates active contributions against stored sessions 🧩
- Orphaned contributions auto-filtered when sessions disappear, keeping UI clean 🧹
- `symphony:getState` returns only session-backed active items for accuracy 🎯
- `symphony:getActive` now excludes contributions tied to missing sessions 🚫
- Added reusable `filterOrphanedContributions` helper with detailed logging 🪵
- Wired `sessionsStore` dependency through main + handler registration flow 🔌
- Integration tests now mock sessions store for realistic state scenarios 🧪
- Expanded handler tests to cover missing-session filtering behavior 🛡️
Merges OpenCode fixes and SSH stdin improvements:
- Use stdin passthrough for all SSH prompts (simplifies escaping)
- Add question:deny to OpenCode permission block for robust tool disabling
- Simplify stdin prompt delivery by appending after exec (no heredoc needed)
Per OpenCode GitHub issue workaround, add "question": "deny" to the
permission block in addition to the existing "tools":{"question":false}.
This ensures the question tool is disabled via both configuration methods,
preventing stdin hangs in batch mode.
Config now: {"permission":{"*":"allow","external_directory":"allow","question":"deny"},"tools":{"question":false}}
Replace heredoc-based prompt delivery with simpler stdin passthrough.
How it works:
1. Bash script is sent via stdin to /bin/bash on the remote
2. Script sets up PATH, cd, env vars, then calls `exec <agent>`
3. The `exec` replaces bash with the agent process
4. The agent inherits stdin and reads the remaining content (the prompt)
Benefits:
- No heredoc syntax needed
- No delimiter collision detection
- No prompt escaping required - prompt is never parsed by any shell
- Works with any prompt content (quotes, newlines, $, backticks, etc.)
- Simpler, more maintainable code
Changes:
- Remove heredoc logic from buildSshCommandWithStdin()
- Update process.ts to use stdin passthrough for ALL agents over SSH
(not just OpenCode - all agents benefit from this approach)
- Update tests to verify stdin passthrough behavior
Verified locally that both OpenCode and Claude Code read prompts from stdin.
When Claude Code performs multi-tool turns (many internal API calls),
accumulated token values cause estimateContextUsage to return null,
freezing the context gauge. This adds estimateAccumulatedGrowth which
provides conservative 1-3% per-turn growth estimates so the gauge
keeps moving during tool-heavy sessions.
Safety: App.tsx caps all estimates at yellowThreshold - 5, guaranteeing
that estimates can never trigger compact warnings — only real
measurements from non-accumulated turns can.
Update SummaryCards to accept sessions prop and filter out terminal-only
sessions when calculating agent count for consistent metrics display.
This ensures queries-per-session and total sessions accurately reflect
AI agent sessions rather than including terminal sessions.
Replace process.platform with navigator.platform for Windows detection
in the inline wizard conversation service, as process.platform is not
available in the browser/renderer context.
