🤖 feat: AI-generated workspace creation on first message #500
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
ammario
requested changes
Nov 6, 2025
- Add workspaceTitleGenerator service using Vercel AI SDK - Generate human-readable titles and git-safe branch names - Add sendFirstMessage IPC handler that creates workspace automatically - Add ProjectSelector and FirstMessageInput components for empty state - Update App.tsx to show project selector when no workspace exists - Gracefully fall back to timestamp names if AI generation fails Users can now type a message and hit send to create a workspace with an AI-generated title and branch name based on their message. Generated with `cmux`
The 'name' field is used throughout the backend to compute filesystem paths (e.g., runtime.getWorkspacePath uses metadata.name as the directory). Storing the AI-generated title there would break path resolution. Solution: - Add optional 'displayName' field to WorkspaceMetadata - Store git-safe branch name in 'name' (used for paths) - Store human-readable title in 'displayName' (used for display) - UI prefers displayName over name when rendering This maintains backward compatibility - existing workspaces without displayName continue to work, showing the branch name as before. Generated with `cmux`
FirstMessageInput was using useSendMessageOptions which internally uses useThinking hook, causing 'useThinking must be used within a ThinkingProvider' error in E2E tests when empty state was rendered. Instead, directly use useModelLRU to get the most recent model without requiring context providers. Generated with `cmux`
Instead of showing a project selector dropdown when no workspace is selected, we now: - If only one project exists: Show FirstMessageInput immediately - If multiple projects exist: Show welcome message asking user to select/create workspace via sidebar This is cleaner UX - users already interact with projects via the left sidebar, no need for duplicate selection UI. Generated with `cmux`
1. Merge sendFirstMessage into sendMessage - make workspaceId optional to reduce duplication in IPC layer. When null, automatically creates workspace with AI-generated title/branch. 2. Use standard useSendMessageOptions in FirstMessageInput instead of manually managing model selection. 3. Use the same model as provided in send message options for title generation instead of hardcoding cheap models. This respects custom BaseURL configurations and avoids manual updates when models change. 4. Remove redundant WORKSPACE_SEND_FIRST_MESSAGE IPC channel and handler. Generated with `cmux`
useSendMessageOptions internally uses useThinking which requires ThinkingProvider context. Since FirstMessageInput is rendered outside of context providers, manually build send options instead. Generated with `cmux`
Fixes floating promise lint error in sendMessage when creating workspaces without explicit workspaceId. editConfig() returns a promise that must be awaited before continuing to retrieve the complete metadata.
Changes workspace creation UX to show FirstMessageInput instead of NewWorkspaceModal when user clicks the '+ Add Workspace' button. User flow: 1. Click '+ Add Workspace' in sidebar 2. See FirstMessageInput (same as single-project empty state) 3. Type message and hit send 4. Workspace created with AI-generated title/branch 5. Message sent to new workspace Added cancel functionality: - Escape key cancels workspace creation - Cancel button shown when in 'add workspace' mode - Returns user to empty state on cancel NewWorkspaceModal still exists for advanced users (can be triggered via command palette or other flows in the future).
Redesigned FirstMessageInput to use the same visual style as ChatInput for a unified user experience. Changes: - Replaced custom textarea with VimTextArea component (matches ChatInput) - Added ModelSelector component for choosing AI model - Added runtime config selector (Local vs SSH) with SSH host input - Styled with same CSS classes as ChatInput (bg-separator, border-border-light, etc.) - Options displayed inline below input (Model | Runtime | Cancel/Send) - Added helpful tooltip for runtime configuration All workspace creation options now visible and accessible without needing to open a separate modal.
Changes: - Removed Send and Cancel buttons (use Cmd+Enter to send like ChatInput) - Input now takes full width (removed button constraints) - Added project title in center of empty view - Shows trunk branch below project name (auto-detected, informational) - Removed manual trunk branch selector (backend auto-detects it) - Cleaner, more unified UX matching ChatInput behavior
- Import useMode hook from ModeContext - Add MODE_OPTIONS and ModeHelpTooltip (matching ChatInput) - Add mode selector UI with responsive layouts (full and compact) - VimTextArea now uses mode state instead of hardcoded 'exec' - Includes tooltip explaining Exec vs Plan mode Related to #500
…ut backend - Import useThinkingLevel, use1MContext hooks and policy utilities - Build SendMessageOptions with all required fields (mode, toolPolicy, thinkingLevel, providerOptions) - Add ThinkingSliderComponent and Context1MCheckbox UI (matching ChatInput layout) - Mode selector now positioned with ml-auto (right-aligned like ChatInput) - All message options (mode, thinking, 1M context) now properly passed to backend This ensures the first message respects the same settings as regular messages. Related to #500
FirstMessageInput doesn't have a workspace yet, so it can't use the ModeProvider context (which requires a workspaceId). Instead, use usePersistedState directly with a project-scoped mode key. This matches the pattern already used for the model preference and ensures the mode persists per-project before workspace creation. Fixes: "useMode must be used inside ModeProvider" error Related to #500
…eInput Similar to the mode fix, thinking level also requires a provider with workspaceId. Use usePersistedState directly with a project-scoped key to avoid the "useThinking must be used within a ThinkingProvider" error. All three preferences now use the same pattern: - Mode: project-scoped (persists per-project) - Thinking level: project-scoped (persists per-project) - 1M context: global (already uses usePersistedState directly) Related to #500
…cies ThinkingSliderComponent and Context1MCheckbox both require context providers with workspaceId. Created simple inline versions for FirstMessageInput: - SimpleThinkingControl: Clickable button that cycles through thinking levels - Simple1MCheckbox: Standard checkbox with label These components accept state as props instead of using context hooks, eliminating the provider dependency while maintaining the same functionality. Fixes: "useThinking must be used within a ThinkingProvider" error Related to #500
- Made workspaceId optional in ModeProvider and ThinkingProvider - Falls back to global key (__global__) when workspaceId not provided - Wrapped FirstMessageInput in ModeProvider and ThinkingProvider in App.tsx - Removed inline SimpleThinkingControl and Simple1MCheckbox components - Now uses actual ThinkingSliderComponent and Context1MCheckbox - Uses context hooks (useMode, useThinkingLevel) instead of direct usePersistedState This ensures FirstMessageInput uses the exact same components as ChatInput, providing UI consistency and shared behavior. Related to #500
- Added trunk branch dropdown in options row (label: "From:") - Updates sendMessage IPC to accept optional trunkBranch parameter - Backend uses provided trunkBranch if specified, otherwise auto-detects - Trunk branch selector shows all available branches from the project - Matches styling of other controls (Runtime selector, etc.) This allows users to choose which branch to branch from when creating a new workspace, just like the NewWorkspaceModal did. Related to #500
- Removed 'Based on {trunkBranch}' text from center display
- Made trunk branch selector narrower (max-w-[120px])
- Center now just shows project name, all controls in options row
Related to #500
1. Workspace title generation now prioritizes fast/cheap models: - Haiku (Anthropic) first - GPT-4o-mini (OpenAI) second - Falls back to user's selected model if neither available This makes workspace creation much faster - no longer uses slow models like Sonnet for simple title/branch generation. 2. Fixed Enter key to send message (was requiring Shift+Enter): - Now uses matchesKeybind(KEYBINDS.SEND_MESSAGE) like ChatInput - Plain Enter sends, Shift+Enter adds newline - Matches ChatInput behavior exactly Related to #500
- Shows animated spinner in center when isSending is true - Displays "Creating workspace..." message below spinner - Replaces project title temporarily during creation - Uses accent color for spinner to match theme This provides visual feedback during the workspace creation process, preventing the UI from appearing frozen while AI generates the title and workspace is being created. Related to #500
Added descriptive text below project title explaining: - Describe what you want to build - Workspace created automatically with generated branch name - Runtime and model options configurable below Makes the first-time experience more clear and guides users on what to do. Related to #500
- Add projectPath prop to ModeProvider and ThinkingProvider - Use project-scoped storage keys when workspaceId not provided - Scoping priority: workspace > project > global - Matches existing pattern for model selection (__project__ prefix) - FirstMessageInput now persists mode/thinking per-project, not globally
Matches the standard naming convention used throughout codebase (slash commands, tests, tokenizer mappings)
- Created ModeSelector.tsx with MODE_OPTIONS and ModeHelpTooltip - Replaced duplicate mode selector code in ChatInput and FirstMessageInput - Ensures mode UI changes are reflected in both places - Includes responsive layouts (full/compact) and keyboard shortcut tooltip
- Remove unused cn import from ChatInput and FirstMessageInput - Remove unused UIMode import from ChatInput - Format ModeSelector.tsx with prettier
ammario
requested changes
Nov 13, 2025
- Use '/' delimiter in project scope IDs (e.g. __project__/path not __project__path) - Always use user's model for title generation (not hardcoded fast models) - Extract createInitLogger() helper to avoid duplication (3 occurrences) Addresses ammario's feedback on delimiter choice, model selection maintainability, and code duplication patterns.
- Regenerate missing title/branch on workspace load (WORKSPACE_GET_INFO handler) - Add Config.updateWorkspaceMetadata() for updating workspace fields - Extract createWorkspaceForFirstMessage() to reduce sendMessage handler complexity - Decomposed 140+ line function into focused helper method Addresses ammario's feedback on robustness and code decomposition.
- Await updateWorkspaceMetadata() call - Format ipcMain.ts with prettier
- Prioritize fast/cheap models for workspace title generation - Use claude-haiku-4-5 (Anthropic) instead of user's model - Use gpt-5-mini (OpenAI) as second choice - Fall back to user's model if neither provider available
ammar-agent
added a commit
that referenced
this pull request
Nov 13, 2025
Fixes two issues introduced in PR #500: 1. Runtime saving regression: The FirstMessageInput component wasn't persisting runtime selection to localStorage after workspace creation. Added runtime preference saving using updatePersistedState helper. 2. UI layout jank: Controls specific to new workspace creation (trunk branch selector and runtime selector) were mixed with general chat controls. Separated them into two rows for better visual organization. Also fixed: Use persistedState helpers (readPersistedState, updatePersistedState) instead of direct localStorage calls throughout FirstMessageInput for consistency and cross-component synchronization. Added note to AGENTS.md about using persistedState helpers and keeping edits minimal/token-efficient. _Generated with `cmux`_
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Changes the "New Workspace" modal to instead be inline in the chat form.
mux-titles-new-2025-11-13_09.43.32.mp4