Repository navigation
# 🚀 OpenRouter Fixed v1.1.1 - Major Update with Vision Support
Pre-release🚀 OpenRouter Fixed v1.1.0 - Major Update with Vision Support
We're excited to announce the initial release of OpenRouter Fixed, a comprehensive solution that not only resolves all known issues with Home Assistant's official OpenRouter integration but also introduces powerful new vision capabilities for image analysis!
🎯 What This Release Solves
This integration addresses the critical compatibility and functionality issues that have been affecting Home Assistant users with the official OpenRouter integration since version 2025.8.0.
✨ NEW FEATURES
📷 Complete Image Analysis Support
The headline feature everyone has been waiting for!
- Camera Snapshot Analysis: Analyze doorbell cameras, security feeds, parking lots, and more
- Vision Model Integration: Full support for Claude-3, GPT-4 Vision, and Gemini Pro Vision
- Multi-Format Support: JPEG, PNG, WebP, and other common image formats
- Base64 Encoding: Automatic conversion for OpenRouter API compatibility
- Structured Data Output: Get JSON responses for object detection, counting, classification
🤖 Vision Models Supported
anthropic/claude-3-haiku- Fast & cost-effective for camera analysisanthropic/claude-3-sonnet- Best balance of quality and performanceanthropic/claude-3-opus- Highest quality analysisopenai/gpt-4-vision-preview- OpenAI's vision modelgoogle/gemini-pro-vision- Google's multimodal AI
🔧 Smart Error Handling
- Vision model validation with helpful error messages
- Graceful fallback when images can't be processed
- Clear guidance for choosing compatible models
🐛 CRITICAL FIXES RESOLVED
1. OpenAI Library Compatibility (#150165)
Issue: ImportError: cannot import name 'ResponseFormatJSONSchema'
- Root Cause: Version conflicts with Extended OpenAI Conversation and other integrations
- Solution: Flexible version requirements + fallback implementations
- Result: Works alongside any OpenAI-based integration
2. Model Parsing Errors (#149905)
Issue: 'audio' is not a valid Modality and architecture parsing failures
- Root Cause: New OpenRouter model types not supported by python-open-router library
- Solution: Safe parsing with direct API fallback + individual model skipping
- Result: All OpenRouter models load successfully
3. AI Tasks API Failures (#150294)
Issue: "Error talking to API" for all AI task operations
- Root Cause: Multiple compatibility and parsing issues combined
- Solution: Comprehensive error handling + better API communication
- Result: AI tasks work reliably with all supported models
🏗️ TECHNICAL IMPROVEMENTS
Robust Architecture
- Flexible Dependencies:
openai>=1.99.5,<2.0.0prevents version conflicts - Safe Model Loading: Direct API calls when library parsing fails
- Non-blocking Errors: Individual model failures don't crash the integration
- Enhanced Logging: Better debugging information for troubleshooting
Vision Implementation
- Multi-part Messages: Text + image content in single API calls
- Image Processing Pipeline: Automatic base64 conversion and validation
- Format Detection: Smart handling of different image sources
- Error Recovery: Graceful degradation when image processing fails
🚀 REAL-WORLD USE CASES
Security & Monitoring
# Doorbell visitor identification
- service: ai_task.generate_data
data:
prompt: "Who is at the door? Person, delivery, package, or other?"
attachments: ["/tmp/doorbell_snapshot.jpg"]
# Security camera alerts
- service: ai_task.generate_data
data:
prompt: "Analyze this security footage for unusual activity"
attachments: ["/config/www/security_cam.jpg"]
structure:
person_detected: bool
activity_type: str
confidence_level: strSmart Home Automation
# Parking spot availability
- service: ai_task.generate_data
data:
prompt: "Count available parking spaces"
attachments: ["/tmp/parking_lot.jpg"]
# Garden monitoring
- service: ai_task.generate_data
data:
prompt: "Check plant health and watering needs"
attachments: ["/config/www/garden_cam.jpg"]📦 INSTALLATION & SETUP
Via HACS (Recommended)
- Add custom repository:
https://github.com/brouwerict/open_router_fixed - Install "OpenRouter Fixed" integration
- Restart Home Assistant
- Add integration with your OpenRouter API key
- Configure AI tasks with vision models for image analysis
Model Selection Tips
- Quick camera checks: Use Claude-3-Haiku (fast, cheap)
- Detailed analysis: Use Claude-3-Sonnet (best balance)
- Critical applications: Use Claude-3-Opus (highest accuracy)
🔄 MIGRATION FROM OFFICIAL INTEGRATION
This integration runs alongside the official OpenRouter integration without conflicts. You can:
- Keep your existing setup running
- Install OpenRouter Fixed as additional integration
- Test with vision features
- Migrate automations when ready
- Remove official integration if desired
No data loss or configuration migration required!
🤝 COMMUNITY IMPACT
This release addresses the most requested features and critical issues reported by the Home Assistant community:
- 846+ active installations affected by the original issues
- Multiple GitHub issues resolved with single integration
- New vision capabilities unlock countless automation possibilities
- Backward compatible - existing automations continue working
🛠️ TECHNICAL SPECIFICATIONS
- Home Assistant: 2025.8.0+ required
- Dependencies:
openai>=1.99.5,<2.0.0,python-open-router>=0.3.1 - Platforms: AI Task, Conversation
- Image Formats: JPEG, PNG, WebP, GIF
- Max Image Size: Limited by OpenRouter API (typically 20MB)
- Supported Features: Text generation, Structured output, Vision analysis, Tool calling
🔮 WHAT'S NEXT
We're actively monitoring for:
- New OpenRouter model releases
- Additional vision capabilities
- Performance optimizations
- Community feature requests
💡 GET STARTED TODAY
Ready to supercharge your Home Assistant with AI vision?
- Install via HACS:
https://github.com/brouwerict/open_router_fixed - Get API key at OpenRouter.ai
- Choose vision model (recommend Claude-3-Haiku for testing)
- Try camera snapshot with your first AI task
- Share your automations with the community!
Questions or issues? Open a GitHub issue - we're here to help!
Enjoy your new AI vision superpowers! 🤖👁️