926df2cfaf
- Complete architectural overhaul of useUnifiedCompletion hook - Unified state management: 8 separate states → single CompletionState interface - Simplified core logic: getWordAtCursor 194 lines → 42 lines (78% reduction) - Fixed infinite React update loops with ref-based input tracking - Smart triggering mechanism replacing aggressive auto-completion - Integrated @agent and @file mention system with system reminders - Added comprehensive agent loading and mention processing - Enhanced Tab/Arrow/Enter key handling with clean event management - Maintained 100% functional compatibility across all completion types Key improvements: • File path completion (relative, absolute, ~expansion, @references) • Slash command completion (/help, /model, etc.) • Agent completion (@agent-xxx with intelligent descriptions) • System command completion (PATH scanning with fallback) • Terminal-style Tab cycling, Enter confirmation, Escape cancellation • Preview mode with boundary calculation • History navigation compatibility • Empty directory handling with user feedback Architecture: Event-driven @mention detection → system reminder injection → LLM tool usage Performance: Eliminated 7-layer nested conditionals, reduced state synchronization issues Reliability: Fixed maximum update depth exceeded warnings, stable state management
2.1 KiB
2.1 KiB
System Reminder Solution for @mentions
Design Principles
- Keep it Simple: No complex async operations
- Don't Modify Input: Preserve original user message
- Append Only: Add system-reminders at the end
- Synchronous: Use require() instead of import()
Implementation
// Simple pattern matching
const mentions = input.match(/@[\w\-\.\/]+/g) || []
// Generate reminders without modifying input
for (const mention of mentions) {
// Simple agent detection
if (isLikelyAgent(name)) {
reminders.push(agentReminder)
} else if (existsSync(filePath)) {
reminders.push(fileReminder)
}
}
// Append reminders to message
processedInput = processedInput + '\n\n' + reminders.join('\n')
How It Works
For @file mentions:
- User types:
@improvements_summary.md - System adds reminder: "You should read the file at: /path/to/improvements_summary.md"
- AI sees the reminder and uses Read tool
- Agent loop continues normally
For @agent mentions:
- User types:
@code-reviewer - System adds reminder: "Consider using the Task tool with subagent_type='code-reviewer'"
- AI sees the reminder and may invoke the agent
- Agent loop continues normally
Why This Works
- No Flow Interruption: Synchronous execution preserves message flow
- Original Input Intact: User's message isn't modified
- Simple Logic: No complex async imports or checks
- Agent Loop Safe: Messages flow through without breaking
Difference from Previous Approach
Previous (Broken):
- Complex async operations
- Modified user input (removed @mentions)
- Multiple async imports
- 70+ lines of complex logic
- Broke agent loop
Current (Working):
- Simple synchronous checks
- Preserves user input
- Uses require() not import()
- ~40 lines of simple logic
- Agent loop works properly
Testing
You can test with:
@improvements_summary.md summarize this- should read file@code-reviewer check my code- should suggest agent@src/commands.ts explain this- should handle paths
The agent should continue executing multiple rounds without interruption.