👁️ Iris Vision System Release
讓 Iris 看懂你的螢幕，主動提供幫助
This release introduces a revolutionary visual understanding system that enables Iris to "see" your screen, understand your workflow, and proactively offer assistance based on visual context.
What's New
Vision Capture Module
- Full screen, active window, and region capture
- Automatic cleanup of old screenshots
- BrowserOS integration for browser-specific capture
- Statistics and monitoring tools
Vision Analyzer Engine
- Structured visual analysis with Claude's vision API
- Multiple modes: assistant, debug, automation, security
- Context management and change detection
- Workflow tracking and reporting
Vision Assistant System
- Proactive monitoring with customizable intervals
- Help detection intelligence (errors, patterns, idle time)
- Context memory system (10-item sliding window)
- Native macOS notifications
- Actionable assistance suggestions
Quick Start
```bash
cd vision
Capture screenshot
node vision-capture.js fullscreen
Analyze current screen
node vision-analyzer.js analyze
Start proactive assistant
node vision-assistant.js start
Debug mode (30s interval)
node vision-assistant.js start 30000 debug
```
Use Cases
Debug Assistant Mode
```bash
node vision-assistant.js start 30000 debug
```
Automatically detects errors, analyzes context, and provides solutions.
Automation Discovery Mode
```bash
node vision-assistant.js start 60000 automation
```
Identifies repetitive tasks and suggests automation opportunities.
Workflow Monitoring
```bash
node vision-analyzer.js monitor 60000
```
Tracks application usage, task switching, and error patterns.
Features
✅ Visual Understanding via Claude's vision
✅ Proactive assistance before you ask
✅ Automatic error detection
✅ Pattern learning and workflow understanding
✅ Smart macOS notifications
✅ Privacy-first local storage
✅ Multiple operational modes
✅ Auto-cleanup of old files
Technical Details
- 44KB of new code (~1,200 lines)
- Three modular components with CLI interfaces
- Integration with Claude Code's Read tool
- Local storage in ~/.iris-vision/
- Configurable retention policies (24-hour default)
Privacy & Security
- All screenshots stored locally
- No cloud uploads
- Automatic cleanup of old files
- Can be disabled at any time
- Configurable retention policies
Impact
Before v2.1.0: Iris relied on API integrations and scheduled tasks
After v2.1.0: Iris can "see" your screen and understand visual context
This enables proactive problem solving, workflow understanding, context-aware assistance, visual debugging, and automation discovery.
See CHANGELOG.md for complete details.
🤖 Generated with Claude Code
via Happy
Co-Authored-By: Claude noreply@anthropic.com
Co-Authored-By: Happy yesreply@happy.engineering