Work Items Workflow

What Is a Work Item?

• Description: Your request explaining what to generate
• Context: Files, knowledge, media, and meetings that inform the AI
• Artifact: The generated output
• Iterations: Multiple rounds of generation with history preserved

What Can Work Items Generate?

• Source Code: Code files, configuration files, and other text-based project files
• Knowledge: Structured documentation, requirements, specifications, and technical references
• Personalities: AI behavior configurations for different roles and contexts

Creating a Work Item

From the Sidebar

1. Click New Work Item in the sidebar
2. The new Work Item page opens with the Chat Panel ready for input

Quick Start

1. Type your description in the message input
2. Click Create to generate code immediately

Writing Effective Prompts

What You Want

Create a React hook called useDebounce that delays updating
a value until the user stops typing for a specified duration.

Technical Requirements

Create a REST API endpoint for user authentication using
Express.js with JWT tokens. Include input validation with Zod
and rate limiting with express-rate-limit.

Context References

Add a delete button to the UserCard component that matches
the style of existing buttons in the codebase.

Generation Modes

Plan Mode

• The task is complex or unclear
• You want to review the approach before implementation
• Multiple solutions exist and you want AI recommendation

Create Mode

• The task is straightforward
• You have a clear understanding of what you want
• You want to iterate quickly

Providing Context

Selecting Source Files

1. Open the Context Panel on the right
2. Browse your project file tree
3. Check files relevant to your task:
   • Files the AI should read for reference
   • Files the AI might modify
   • Related components or utilities

Tip: Select 5-15 relevant files. Too few files means missing context; too many files dilutes focus.

Attaching Knowledge

1. Expand the Knowledge section
2. Select relevant Knowledge articles:
   • Coding standards
   • API documentation
   • Architecture guidelines

Adding Images

1. Expand the Media section
2. Select images that help explain your request:
   • UI mockups
   • Error screenshots
   • Design references

Including Meetings

1. Expand the Meetings section
2. Select meeting transcripts with relevant discussions

Submitting for Generation

1. Verify your prompt is complete
2. Check that context files are selected
3. Choose your preferred AI model
4. Optionally select a Personality
5. Click Create (or Plan for planning mode)

Reviewing Generated Code

Browse Files

1. Open the Changes Panel
2. Click files to view their contents
3. Files are organized by type (Code, Docs, etc.)

Use Diff View

1. Toggle Diff View in the Code Viewer toolbar
2. Compare generated code against existing files
3. Review additions (green) and deletions (red)

Check the Plan

Editing Before Applying

1. Select a file in the Changes Panel
2. Click Edit in the Code Viewer toolbar
3. Make your changes
4. Click Save to update the iteration

Applying Changes

Apply All Files

Apply Individual Files

1. Select a file in the Changes Panel
2. Click Apply for that specific file

What Happens

• New files are created in your project
• Modified files overwrite existing versions
• Deleted files are removed from your project

Warning: Applied changes overwrite existing files. Review diffs carefully before applying.

Iterating on Results

Follow-Up Prompts

1. Type a follow-up message explaining what to change
2. Click Create to generate a new iteration
3. New iteration builds on previous context

• "Add error handling to the API endpoint"
• "Use async/await instead of promises"
• "Include unit tests for the new component"

Viewing Iteration History

• Switch between iterations
• Compare different versions
• Understand how code evolved

Reverting to Previous Iterations

1. Select an earlier iteration
2. Review its contents
3. Apply that iteration instead

Example Workflows

Generating Source Code

1. Describe the feature or change you need
2. Select relevant source files as context
3. Attach coding standards from Knowledge Base
4. Choose a developer-focused Personality
5. Click Create to generate code
6. Review and apply to your project

Generating Requirements from Meetings

1. Add meeting transcript to the Meetings section
2. Create a Work Item referencing that meeting
3. Select a "Requirements Analyst" or "Project Manager" Personality
4. Prompt: "Analyze the attached meeting and create structured requirements documentation"
5. AI generates Knowledge articles with organized requirements
6. Review and save to your Knowledge Base

Generating Personalities

1. Create a Work Item describing the role you need
2. Prompt: "Create a Personality for a senior React developer who follows our coding standards"
3. Attach your coding standards from Knowledge Base as context
4. AI generates a Personality with appropriate instructions
5. Review and save for team use

Tips for Better Results

Be Specific

Provide Examples

Iterate Incrementally

1. Generate basic structure
2. Add error handling
3. Include validation
4. Add tests

Use Personalities

• Senior Developer for architecture decisions
• Code Reviewer for quality improvements
• Documentation Writer for comments and docs
• Requirements Analyst for structuring requirements from meetings

Review Before Applying

What's Next?

• Knowledge Base - Create reusable context
• Personalities - Tune AI behavior
• Meetings - Add meeting context for requirements
• The Workbench - Interface reference