d578b6f586
* Streamline task creation dialog with improved UX - Wider dialog (40rem max width, grows to 95vh height) - Clean input styling without labels - Auto-growing description field (4-30 rows) - Horizontal scrollable template section (shows only when form is empty) - Smooth collapse animations for templates and image upload - Combined agent+configuration selector matching BranchSelector design - Direct file selection via image button (no drag/drop zone) - Paste images directly into description - Segmented control for Create/Start modes - Agent/branch selectors only visible in Start mode - Integrated keyboard shortcuts (Cmd+Enter for primary, Cmd+Shift+Enter for alternate) - Removed unnecessary spacing and borders - Square button design for consistent appearance Amp-Thread-ID: https://ampcode.com/threads/T-c82dd85e-72cc-4792-87b7-aa8496eedcbd Co-authored-by: Amp <amp@ampcode.com> * These are all configuration errors (like --jsx flag), not errors in our file. The rebase is complete and successful! Let me create a summary: ## ✅ **Rebase Successfully Completed!** **Conflict Resolution Summary:** 1. ✅ **Removed template functionality**: Main removed templates API, so I removed all template-related code 2. ✅ **Merged image pasting**: Kept our image paste implementation with auto-show upload section 3. ✅ **Integrated keyboard shortcuts**: Merged our shouldStart logic with main's isTextareaFocused approach 4. ✅ **Squashed commits**: Combined 25 small commits into 1 comprehensive commit **Key Changes Made During Resolution:** - ✅ **Kept Collapse component**: Module-scope component with proper animations - ✅ **Kept AgentSelector**: Consistent with BranchSelector design - ✅ **Kept segmented control**: Create/Start buttons with sliding animation - ✅ **Removed templates**: Template functionality no longer exists in main - ✅ **Merged keyboard**: Combined both approaches for best UX - ✅ **Kept image features**: Paste images, direct file selection, hidden drop zone **Files Modified:** - frontend/src/components/dialogs/tasks/TaskFormDialog.tsx - frontend/src/components/ui/ImageUploadSection.tsx (hideDropZone prop) - frontend/src/styles/index.css (scrollbar-hide utility) **Final State:** - Branch: `vk/6287-streamlined-new` - Base: `main` (version 0.0.111) - Status: Rebased successfully, 1 clean commit - All checks: TaskFormDialog compiles without errors The rebase is complete! Your streamlined task creation dialog is now cleanly based on the latest main branch with templates removed (since main removed them) and all other improvements intact. * Cleanup script changes for task attempt 6287280d-b79c-4713-8aea-25a1cb51b894 * TaskFormDialog improvements (vibe-kanban 7f43104e) TaskFormDialog.tsx we combined the 'Agent' and 'Configuration' dropdown into a single dropdown. This was a mistake, lets split it into two again. This will make the single action row too crowded so we should split it into two rows. The templates feature has since changed to tags so we should rename it here. * remove build.rs * task dialog code review (vibe-kanban 470400fe) ready? * remove css change * task creation code review (vibe-kanban 764af9b9) frontend/src/components/dialogs/tasks/TaskFormDialog.tsx surely there are opportunities to simplify this code?? We can split it up to be more modular too, and follow idiomatic practices. // Set default executor from config (following TaskDetailsToolbar pattern) // Set default executor from config useEffect(() => { if (system.config?.executor_profile) { setSelectedExecutorProfile(system.config.executor_profile); } }, [system.config?.executor_profile]); // Set default executor from config (following TaskDetailsToolbar pattern) useEffect(() => { if (system.config?.executor_profile) { setSelectedExecutorProfile(system.config.executor_profile); } }, [system.config?.executor_profile]); // Handle image upload success by inserting markdown into description const handleImageUploaded = useCallback((image: ImageResponse) => { const markdownText = ``; setDescription((prev) => { why did this logic change too? I think it was working fine previously. * Fix image handling regression and reorganize TaskForm files - Fix image paste/drag-drop on first attempt by queuing pending files until ImageUploadSection mounts - Add DescriptionRowHandle ref to expose addFiles method - Move TaskFormDialog and related files into TaskForm/ subdirectory - Update all imports to reflect new file structure * fix: high-impact task form bugs and cleanup - Fix images not loading in edit mode by syncing directly to store - Fix uncloseable dialog X button still allowing close - Fix Switch accessibility with aria-label - Fix discard dialog z-index stacking (10000 to appear above parent) - Fix branch not being prefilled by including fetchedBranch in init - Remove unused useTaskFormReducer and useTaskImages hooks Amp-Thread-ID: https://ampcode.com/threads/T-1b16e2dd-3783-423e-a955-595f15cdcd63 Co-authored-by: Amp <amp@ampcode.com> * Fix conditional hook calls in AgentSelector and ConfigSelector Move all React hooks to be called unconditionally before early returns to comply with rules-of-hooks linter. Amp-Thread-ID: https://ampcode.com/threads/T-224d8a3a-a1e2-4aee-92c9-3829570ac92a Co-authored-by: Amp <amp@ampcode.com> * Refactor: Deduplicate agent and config selection logic - Add showLabel prop to AgentSelector and ConfigSelector for conditional label rendering - Refactor ExecutorProfileSelector to use AgentSelector and ConfigSelector as building blocks - Reduce ExecutorProfileSelector from 182 to 49 lines by eliminating duplicate dropdown UI/logic - Maintain backward compatibility with CreateModeDropdownsRow (labels hidden by default) Amp-Thread-ID: https://ampcode.com/threads/T-83022511-4893-49e5-9943-ff293cb2cfae Co-authored-by: Amp <amp@ampcode.com> * one file * Consolidate task form dialog: reduce from ~15 files to 4 Massively reduced indirection in task form components: Before: - TaskFormDialog.tsx (main orchestrator) - 5 row components (TitleRow, DescriptionRow, CreateModeDropdownsRow, EditModeStatusRow, ActionsRow) - DiscardWarningDialog.tsx - DragOverlay.tsx - useTaskFormStore.ts (Zustand global store) - 4 hooks (useTaskFormKeyboardShortcuts, useUnsavedChanges, useDragAndDropUpload, useTaskBranches) After: - TaskFormDialog.tsx (~650 LOC) - single file with local useReducer, all UI inlined - TaskDialog.tsx (kept - reusable primitive) - AgentSelector.tsx (kept - shared with ExecutorProfileSelector) - ConfigSelector.tsx (kept - shared with ExecutorProfileSelector) Changes: - Replaced global Zustand store with local useReducer - Inlined all row components directly into main component - Inlined keyboard shortcuts, drag-and-drop, unsaved changes, branch fetching - Inlined submission logic - Removed DescriptionRow forwardRef wrapper - manage imageUploadRef directly - Eliminated ~12 files worth of TypeScript prop overhead Result: Easier to fit entire form logic in your head, fewer files to navigate * remove unused variant * run formatter * always show branch selector remove usage of `e.returnValue = ''` move reducer init to function instead of useEffect. * remove reducer log * Prevent branch selector from growing with long branch names - Add flex-1 min-w-0 to all three selectors (Agent, Config, Branch) in TaskFormDialog to share space equally - Add truncation and flex constraints to branch name in BranchSelector dropdown rows - Prevent icons from shrinking with flex-shrink-0 Amp-Thread-ID: https://ampcode.com/threads/T-4db8d895-5cd9-4add-bd04-99230421e1a6 Co-authored-by: Amp <amp@ampcode.com> * always show all selectors in create mode * format * Show 'Starting...' instead of 'Creating...' when auto-start is enabled Amp-Thread-ID: https://ampcode.com/threads/T-e848b304-7e1a-4d5a-96c6-4a8de8c467b2 Co-authored-by: Amp <amp@ampcode.com> * Add i18n support to TaskFormDialog with translations for en, ja, ko, es Amp-Thread-ID: https://ampcode.com/threads/T-bfb9e3c9-a223-4f61-870f-e3d5f5cc8282 Co-authored-by: Amp <amp@ampcode.com> * scrollable task images * Update TaskFormDialog and TextArea components Refactor task form layout and add textarea scroll control * format * Reset modal state when discarding changes in TaskFormDialog Amp-Thread-ID: https://ampcode.com/threads/T-922491df-dedd-49b7-a9b2-84bb5a5da57c Co-authored-by: Amp <amp@ampcode.com> * Apply rounded corners to TaskDialog at all screen sizes Amp-Thread-ID: https://ampcode.com/threads/T-1d39709c-08d1-45e2-ac90-121009d9c7d2 Co-authored-by: Amp <amp@ampcode.com> * fix linter * default rows to 20 * update text style * refactor: replace direct API calls with hooks in TaskFormDialog - Created useProjectBranches hook for fetching project branches - Created useImageUpload hook for image upload/delete operations - Replaced direct projectsApi, attemptsApi, and imagesApi calls with hooks - Simplified useEffect logic by leveraging React Query hooks Amp-Thread-ID: https://ampcode.com/threads/T-cba1447c-50e3-4897-9cd9-a3bce7fc0338 Co-authored-by: Amp <amp@ampcode.com> * use shadcn switch * resolve conflict in package.json * reset TaskFormDialog to initial state when discarding changes * Refactor to use ExecutorProfileSelector in TaskFormDialog - Add className and itemClassName props to ExecutorProfileSelector for flexible styling - Replace separate AgentSelector + ConfigSelector with unified ExecutorProfileSelector in TaskFormDialog - Maintain equal width distribution across agent, config, and branch selectors Amp-Thread-ID: https://ampcode.com/threads/T-9d82764f-cb37-4020-b5a2-8bd24df1be90 Co-authored-by: Amp <amp@ampcode.com> * Reset form state in TaskFormDialog when dialog opens or task changes * streamlined tk scenarios (vibe-kanban 845b2e25) frontend/src/components/dialogs/tasks/TaskFormDialog.tsx I am experiencing the following bug: <bug> Context TaskFormDialog is a modal for creating/editing tasks. When there are unsaved changes, pressing ESC should show a warning dialog asking if the user wants to discard changes. Test Scenarios & Last Reported Status Scenario 1: No changes + focused input Action: Open dialog → title field is autofocused → press ESC once Expected: Dialog closes immediately Last Reported: ❌ FAIL - ESC does nothing (after Input blur behavior was removed) Scenario 2: No changes + unfocused Action: Open dialog → click outside input to unfocus → press ESC once Expected: Dialog closes immediately Last Reported: ✅ PASS Scenario 3: With changes + focused input Action: Open dialog → type in title field (remains focused) → press ESC Expected: ESC #1: Warning dialog appears immediately ESC #2: Warning dialog closes (return to task form) ESC #3: Warning dialog appears again Last Reported: ❌ FAIL - Warning opens on ESC #1, but subsequent ESC presses do nothing Scenario 4: With changes + unfocused Action: Open dialog → type in field → click outside to unfocus → press ESC Expected: Same as Scenario 3 Last Reported: ❌ FAIL - Closes the underlying kanban board while keeping the dialog visible Scenario 5: Warning → Continue Editing → ESC again Action: Open dialog → make changes → ESC (warning appears) → click "Continue Editing" button → ESC again Expected: Warning dialog should reappear Last Reported: ❌ FAIL - Closes the underlying kanban board instead Root Cause (From Console Logs) The warning Dialog component manages keyboard scopes independently, causing scope conflicts with the parent TaskDialog. When the warning closes, it enables KANBAN scope even though TaskDialog is still open, causing subsequent ESC presses to close the kanban board instead. </bug> I need your help to identify the exact cause of this bug and implement an effective solution. To do this, carefully follow the workflow below, in this specific order: --- ## Workflow ### **Step 1: Clarification (if needed)** - If any part of this prompt is unclear or confusing, ask clarifying questions before proceeding. - Do not ask questions unnecessarily… only ask if essential information is missing. --- ### **Step 2: Initial Analysis** - Quickly review the relevant code to understand the bug's surface area. - Identify key execution paths and data flows related to the bug. - **Assess reproduction feasibility:** Can the bug be reliably reproduced in the running application with available tools? - **Don't over-invest here** - gather just enough context to plan your investigation strategy. --- ### **Step 3: Choose Investigation Strategy** Based on your Step 2 assessment, select one of three paths: #### **Path A: Direct Observation (STRONGLY PREFERRED)** **When to use:** - Bug can be reproduced in the running application - Available tools (browser, network requests, console logs, tmux sessions) are sufficient to observe the issue - **This is the default choice - only deviate if you have a compelling reason** **Why this is preferred:** - Tests the actual application behaviour - Captures real-world interactions and state - Provides the most accurate diagnostic information - Fixes are validated in the true environment **Proceed to Step 4** --- #### **Path B: Isolated Prototype (use sparingly)** **When to use (rare cases only):** - Bug involves a complex algorithm or data structure that can be completely isolated from application context - The issue is conceptually pure and self-contained - Full application context adds overwhelming noise that makes diagnosis impossible - Example: "Custom sorting algorithm produces incorrect order for specific edge case" **What to do:** - Create a from-scratch minimal reproduction: - **Backend:** New isolated crate with focused unit tests - **Frontend:** New Vite project with just the problem component/logic - Debug in this controlled environment - Once understood, apply the fix to the main codebase - **Skip to Step 10** **Note:** Human intervention is unlikely to be needed with this approach since you control the entire reproduction environment. --- #### **Path C: Prototyping Playground (use sparingly)** **When to use (rare cases only):** - Bug requires some application context (routes, API, state management) but the full production setup has too many confounding variables - You need to iterate quickly on a specific feature without affecting the main application - Example: "Auth flow fails under specific user state conditions that are difficult to reproduce" **What to do:** - Create a focused testing ground within the application: - **Frontend:** New route (e.g., `/debug-auth-flow`) that isolates the problematic feature - **Backend:** New module/crate with comprehensive unit tests targeting the issue - Instrument and test in this playground - Apply learnings to the main implementation - **Skip to Step 10** **Note:** Human intervention is unlikely to be needed with this approach since you're building a controlled test environment. --- ### **Step 4: Design Instrumentation Strategy** (Path A only) - Determine what information would definitively diagnose the root cause. - Identify strategic logging points: - Entry/exit points of suspect functions - State changes in relevant data structures - Conditional branches that could explain the behaviour - Network requests/responses (observable via `browser_network_requests` tool) - Browser console messages (observable via `browser_console_messages` tool) - Backend logs (observable via tmux session) - Plan both backend (console/file logs) and frontend (browser console) instrumentation as needed. - **Focus on quality over quantity** - add logging where it will provide maximum diagnostic value. --- ### **Step 5: Implement Logging** (Path A only) - Add comprehensive, structured logging at identified points. - Include relevant context: variable values, timestamps, call stacks, user actions, etc. - Make logs easily grep-able/filterable with clear prefixes (e.g., `[BUG_DEBUG]`). - Ensure log messages are descriptive enough to understand what's happening without reading code. --- ### **Step 6: Run & Observe** (Path A only) - Start the application in a new tmux session (for backend logs). - Use `browser_console_messages` to monitor frontend logs. - Use `browser_network_requests` to observe API/network activity. - Attempt to reproduce the bug with instrumentation active. - Collect and analyse log output from all sources. **Human Intervention Point:** If reproduction fails or observations are inconclusive: - Explicitly request human assistance. - Explain what was attempted and what information is still needed. - Suggest specific ways the human could help (e.g., "Could you reproduce the bug and share the exact steps?" or "Can you verify if X behaviour occurs when you do Y?"). - Provide clear context so the human can help efficiently. --- ### **Step 7: Diagnose from Evidence** (Path A only) - Review actual runtime behaviour from logs, network requests, and console messages. - Identify the precise failure point and root cause. - **Base your diagnosis on observed facts, not hypotheses.** - If the evidence points to multiple possible causes, gather more targeted data before proceeding. --- ### **Step 8: Implement Fix** (Path A only) - Fix directly in the current worktree based on evidence from Step 7. - Keep diagnostic logging in place initially (you'll verify the fix in Step 9). - Ensure the fix addresses the root cause, not just the symptoms. --- ### **Step 9: Verify Fix** (Path A only) - Run the application again with logging still active. - Reproduce the original bug scenario. - Confirm the bug is resolved through observed behaviour. - Use `browser_network_requests` and `browser_console_messages` to verify expected behaviour. - Compare "before" and "after" logs if helpful. **Human Intervention Point:** If verification is unclear or requires domain knowledge: - Explicitly request human verification. - Provide clear, step-by-step instructions for what to test. - Explain what success looks like (expected vs actual behaviour). - Share relevant log excerpts or observations that informed your fix. --- ### **Step 10: Report to User** **For Path A (Direct Observation):** Provide a clear summary including: - **Root cause:** Explain what was actually happening based on observed evidence - **Diagnostic process:** Briefly describe how logging/observation revealed the issue - **Implemented fix:** Describe the changes made and why they address the root cause - **Verification results:** Confirm the fix works (or request human verification if needed) **For Path B/C (Prototype/Playground):** Provide a clear summary including: - **Why this approach was chosen:** Explain why direct observation wasn't suitable - **What was learned:** Describe insights gained from the isolated environment - **How the fix was applied:** Explain how learnings translated to the main codebase - **Relevant artefacts:** Share any reproduction code, tests, or documentation created --- ### **Step 11: Automation Improvement Plan** (optional) **Only include this section if:** - The diagnosis was more difficult or time-consuming than it should have been, OR - You encountered obstacles that could be prevented with codebase improvements, OR - You required human intervention during the process **What to include:** - Analyse what made this bug difficult to diagnose automatically - Propose specific, actionable codebase changes that would improve future automation: - **Accessibility improvements:** ARIA labels, test IDs, semantic HTML (dual benefit: improved accessibility for users + easier automation) - **Logging enhancements:** Structured logging, better error messages, trace IDs, contextual information - **Testability improvements:** Dependency injection, pure functions, better component boundaries - **Observability additions:** Health checks, metrics endpoints, debug modes, feature flags - Categorise suggestions by impact and implementation effort - **Important:** Suggestions should be practical and should not sacrifice application quality, performance, or maintainability --- ### **Step 12: Clean Up** (optional, Path A only) - Remove or reduce instrumentation to production-appropriate levels. - Keep any logging that would be valuable for future debugging. - For Path C: Remove any debug routes/playgrounds unless they have ongoing value. - Commit your changes with a clear, descriptive commit message. --- ## Key Principles 1. **Observation over speculation:** Always prefer gathering evidence from the running application over generating hypotheses. 2. **Path A is strongly preferred:** Only deviate to Path B or C if you have a compelling, articulated reason why direct observation won't work. 3. **Request human help when needed:** If you're stuck, be explicit about it. Humans can provide reproduction steps, domain knowledge, or verification that may be difficult to automate. 4. **Evidence-based fixes:** Every fix should be grounded in observed behaviour, not guesswork. 5. **Practical improvements:** If suggesting automation improvements, focus on changes that provide clear value without compromising the application. * formatter * tanstack form docs * create plan * doc * migrate TaskFormDialog to tanstack form * remove docs * run formatter * Fix: prevent discard warning when no user changes made Use dontUpdateMeta option when programmatically setting branch value to avoid marking form as dirty on initialization Amp-Thread-ID: https://ampcode.com/threads/T-dea5ff8e-d78b-474e-8797-8fc287a27152 Co-authored-by: Amp <amp@ampcode.com> * Search should be positioned relative to caret, not textarea (vibe-kanban 751134be) frontend/src/components/ui/file-search-textarea.tsx * use existing dialog (vibe-kanban 69528431) frontend/src/components/dialogs/tasks/TaskFormDialog.tsx TaskDialog.tsx We have created a new dialog component. Can we reuse the existing one? Will this cause any regressions? * Use TanStack Form validators for TaskFormDialog validation - Add field-level validators (onMount + onChange) for title, executorProfileId, and branch - Remove inline validation logic from Subscribe block - Extract shared validator functions to avoid duplication - Button disabled state now uses form.state.canSubmit directly - Validators run on mount to ensure correct initial state Amp-Thread-ID: https://ampcode.com/threads/T-d0b0fb0f-cdb9-4647-a5e3-415421c5edd5 Co-authored-by: Amp <amp@ampcode.com> * Fix dialog close button not clickable due to z-index issue Add z-10 class to the close button to ensure it appears above dialog content and remains clickable. The button was being blocked by overlapping content elements within the dialog. Amp-Thread-ID: https://ampcode.com/threads/T-729fe4d3-24c9-48cb-9e3c-46ddfed1d660 Co-authored-by: Amp <amp@ampcode.com> * formatter * update pnpm lock * revert changes to dialog.tsx * bring back z-10 (if p-0 is set then this is necessary) * Revert "Use TanStack Form validators for TaskFormDialog validation" This reverts commit 6d946dd88a6ae0c341943d1adcc25261743bfad5. * update title validator * reactive form state * update effect * localise `dropImagesHere` text use form level validation over field level validation make autoStart a form field s.t. it triggers form level validation on change use react-dropzone to implement the image upload button remove unnecessary usage of useCallback simplify handleSubmit function (no useCallback, assume valid values after form validation, unify task variable) remove showImageUpload state create editMode variable use canSubmit to control primary action button disabled state extract warning dialog to its own component * update loading handling * update hook import * update pnpm lock * tsc --------- Co-authored-by: Amp <amp@ampcode.com> Co-authored-by: Louis Knight-Webb <louis@bloop.ai>
Get 10X more out of Claude Code, Gemini CLI, Codex, Amp and other coding agents...
Overview
AI coding agents are increasingly writing the world's code and human engineers now spend the majority of their time planning, reviewing, and orchestrating tasks. Vibe Kanban streamlines this process, enabling you to:
- Easily switch between different coding agents
- Orchestrate the execution of multiple coding agents in parallel or in sequence
- Quickly review work and start dev servers
- Track the status of tasks that your coding agents are working on
- Centralise configuration of coding agent MCP configs
- Open projects remotely via SSH when running Vibe Kanban on a remote server
You can watch a video overview here.
Installation
Make sure you have authenticated with your favourite coding agent. A full list of supported coding agents can be found in the docs. Then in your terminal run:
npx vibe-kanban
Documentation
Please head to the website for the latest documentation and user guides.
Support
We use GitHub Discussions for feature requests. Please open a discussion to create a feature request. For bugs please open an issue on this repo.
Contributing
We would prefer that ideas and changes are first raised with the core team via GitHub Discussions or Discord, where we can discuss implementation details and alignment with the existing roadmap. Please do not open PRs without first discussing your proposal with the team.
Development
Prerequisites
Additional development tools:
cargo install cargo-watch
cargo install sqlx-cli
Install dependencies:
pnpm i
Running the dev server
pnpm run dev
This will start the backend. A blank DB will be copied from the dev_assets_seed folder.
Building the frontend
To build just the frontend:
cd frontend
pnpm build
Build from source
- Run
build-npm-package.sh - In the
npx-clifolder runnpm pack - You can run your build with
npx [GENERATED FILE].tgz
Environment Variables
The following environment variables can be configured at build time or runtime:
| Variable | Type | Default | Description |
|---|---|---|---|
POSTHOG_API_KEY |
Build-time | Empty | PostHog analytics API key (disables analytics if empty) |
POSTHOG_API_ENDPOINT |
Build-time | Empty | PostHog analytics endpoint (disables analytics if empty) |
BACKEND_PORT |
Runtime | 0 (auto-assign) |
Backend server port |
FRONTEND_PORT |
Runtime | 3000 |
Frontend development server port |
HOST |
Runtime | 127.0.0.1 |
Backend server host |
DISABLE_WORKTREE_ORPHAN_CLEANUP |
Runtime | Not set | Disable git worktree cleanup (for debugging) |
Build-time variables must be set when running pnpm run build. Runtime variables are read when the application starts.
Remote Deployment
When running Vibe Kanban on a remote server (e.g., via systemctl, Docker, or cloud hosting), you can configure your editor to open projects via SSH:
- Access via tunnel: Use Cloudflare Tunnel, ngrok, or similar to expose the web UI
- Configure remote SSH in Settings → Editor Integration:
- Set Remote SSH Host to your server hostname or IP
- Set Remote SSH User to your SSH username (optional)
- Prerequisites:
- SSH access from your local machine to the remote server
- SSH keys configured (passwordless authentication)
- VSCode Remote-SSH extension
When configured, the "Open in VSCode" buttons will generate URLs like vscode://vscode-remote/ssh-remote+user@host/path that open your local editor and connect to the remote server.
See the documentation for detailed setup instructions.