File Watching & AI Comments

Byte automatically watches files in your project for changes. When it detects special AI comment markers, it can automatically add files to context and trigger actions.

AI Comment Markers

Byte recognizes four types of AI comment markers, each triggering different behavior:

Editable marker (AI:)

Used for standard code modification requests. Files are added to context as editable, allowing the AI to propose SEARCH/REPLACE blocks.

Marker Quick Reference:

AI: - Request code modifications (editable mode) AI @ - Provide reference context (read-only mode) AI ? - Ask questions about code (read-only mode, no changes) AI ! - Mark urgent/critical issues (editable mode, highest priority) Read-only marker (AI@)

Used to provide reference context without allowing modifications. Files are added as read-only.

AI@ Reference this pattern for other implementations

class BaseHandler: def handle(self): pass

Behavior:

• File added to context in read-only mode • AI can reference the code but cannot propose changes • Useful for providing examples or patterns to follow

───────────────────────────────────────────────────────────────────────────────────────────────────── Question marker (AI?)

Used to ask questions about code. The AI provides answers and recommendations without making changes.

Before (problematic code):

AI? Should this use a context manager instead?

file = open("data.txt") data = file.read() file.close()

Behavior:

• File added to context in read-only mode • AI analyzes the code and provides recommendations • No code changes are proposed • Response includes best practices and alternatives

Example AI Response:

Yes, this should definitely use a context manager. The current code has several issues:

1. If an exception occurs between open() and close(), the file remains open
2. Manual resource management is error-prone
3. Not following Python best practices

Recommended approach:

# Improved version using context manager
with open("data.txt") as file:
    data = file.read()
# File automatically closed when exiting the with block


Benefits:

 • Automatic cleanup even if exceptions occur
 • More Pythonic and readable
 • Follows RAII pattern and prevents resource leaks



---

### Urgent task marker (`AI!`)

Used for critical issues requiring immediate attention. Takes priority over other markers.

```python
# AI! Fix this security vulnerability immediately
password = request.args.get('password')  # Plain text password in GET request!


Behavior:

 • File added to context in editable mode
 • Takes highest priority if multiple AI comments exist
 • AI proposes immediate fixes using SEARCH/REPLACE blocks
 • Clear indication this is a critical issue

─────────────────────────────────────────────────────────────────────────────────────────────────────

How It Works

Detection Flow

 1 File change detected: Byte's watcher service detects when you save a file
 2 AI comment scan: File content is scanned for AI:, AI@, AI?, or AI! markers
 3 Auto-add to context: File is automatically added with appropriate mode:
    • AI: or AI! → Editable mode
    • AI@ or AI? → Read-only mode
 4 Prompt interruption: Current prompt is interrupted to process the AI comment
 5 AI response: AI processes the comment based on marker type:
    • : or ! → Proposes code changes (CoderAgent)
    • ? → Provides analysis and answers (AskAgent)
    • @ → References the code as context

Workflow steps:

1. You add an AI comment in your editor
2. Save the file
3. Byte automatically:
   - Detects the AI comment marker
   - Adds the file to context (editable for `AI:` and `AI!`, read-only for `AI@`)
   - Interrupts the current prompt
   - Shows you the AI comment
   - Asks the AI to implement the requested changes
4. AI responds with SEARCH/REPLACE blocks showing proposed changes
5. You review and approve or reject the changes
6. If approved, changes are applied and linters run automatically (if enabled)

**Multiple comments:**

If multiple AI comments are detected, they're processed together based on priority:

1. `AI!` (urgent) takes precedence
2. Then `AI?` (questions)
3. Then `AI:` (regular tasks)