OpenWebUI Adaptive Memory Plugin v4.0

A monolithic OpenWebUI filter function that provides persistent memory capabilities for AI conversations, enabling personalized responses based on user history and preferences.

What's New in v4.0

• User-Friendly Interface: Revolutionary valve system redesign with intuitive 4-field Quick Setup
• Smart Auto-Configuration: Automatically configures 90+ settings based on your preferences
• 36% Code Reduction: Optimized from 9,156 to 5,853 lines while maintaining all features
• Simple vs Advanced Modes: Choose between "just works" simplicity or full control
• Logical Field Grouping: Organized settings with emojis and clear descriptions
• Enhanced Validation: Helpful error messages guide you to correct configuration
• Security Fixes: Updated dependencies to patch CVE-2024-3772, CVE-2024-23334, CVE-2024-27306
• JSON Repair System: Added support for sub-3B models with improved JSON parsing
• Gemini API Fix: Corrected authentication method for Gemini API integration
• Enhanced Error Handling: Improved circuit breaker pattern for better fault tolerance

Installation

OpenWebUI functions are installed by copying the code directly into the OpenWebUI interface:

1. Open OpenWebUI Admin Panel

   • Navigate to your OpenWebUI instance
   • Login with admin credentials

2. Add the Function

   • Go to "Workspace" → "Functions"
   • Click the "+" button to add a new function
   • Copy the entire contents of adaptive_memory_v4.0.py (the only file you need)
   • Paste into the code editor
   • Click "Save"

3. Enable the Filter

   • Toggle the switch to enable the function
   • Go to "Workspace" → "Models"
   • Select the models you want to use with adaptive memory
   • Assign the filter to those models

4. Configure Settings (Optional)

   • Click on the function name to access configuration
   • Adjust settings like LLM provider, memory thresholds, etc.
   • Save your configuration

Testing the Installation

After installation, test with simple statements:

• "My favorite color is blue"
• "I prefer Python over Java"
• "I work as a software engineer"

The filter will extract and remember these preferences for future conversations.

Features

• User-Specific Memory: Each user has isolated memory storage
• Multi-LLM Support: Works with Ollama, OpenAI-compatible APIs, and Gemini
• Automatic Memory Extraction: Identifies and stores important information
• Context Injection: Seamlessly adds relevant memories to conversations
• Flexible Memory Banks: Organize memories by categories
• Deduplication: Prevents duplicate memory storage
• Sub-3B Model Support: JSON repair system for smaller models

Configuration

Quick Setup (Recommended)

Most users only need to configure these 4 settings:

1. 🎯 Configuration Mode: Choose "simple" (auto-configure everything) or "advanced" (full control)
2. 🤖 LLM Provider: Select your AI service (Ollama, OpenAI-compatible, or Google Gemini)
3. 📝 Model Name: Specify your model (e.g., "llama3:latest", "gpt-4", "gemini-pro")
4. 🧠 Memory Mode: Choose how much to remember ("minimal", "balanced", or "comprehensive")

Advanced Configuration

For power users, the filter includes configurable settings for:

• API endpoints and authentication
• Memory sensitivity and thresholds
• Memory organization and categories
• Advanced processing options
• Performance and reliability settings
• Custom system prompts

Requirements

• OpenWebUI instance (latest version recommended)
• Python 3.8+ (for OpenWebUI)
• No additional installation steps required

File Structure

adaptive_memory_v4.0.py    # Main filter file (copy this to OpenWebUI) - THE ONLY FILE YOU NEED
requirements.txt           # Dependencies reference (for development only)
Other Versions/            # Previous versions for reference
OWUI tech-docs/           # OpenWebUI architecture documentation

Troubleshooting

Filter Not Working

• Ensure the filter is enabled for your model
• Check OpenWebUI logs for error messages
• Verify the filter saved successfully without syntax errors

Memory Not Persisting

• Check that the filter is processing messages (look for log entries)
• Ensure the user is properly authenticated
• Verify the model has the filter assigned

Import Errors

• The filter includes all necessary imports
• If you see import errors, ensure you copied the complete file
• Check that your OpenWebUI Python environment has basic packages

Development

For development and testing:

# Install development dependencies
pip install -r requirements.txt

# Run tests
python -m pytest tests/

Security Notes

• All user memories are isolated by user ID
• No cross-user data access is possible
• API keys should be configured securely in OpenWebUI
• Dependencies have been updated to patch known vulnerabilities

License

MIT License - See LICENSE file for details

Contributing

Contributions are welcome! Please ensure:

• Code maintains the monolithic structure (single file)
• All features remain functional
• Tests pass successfully
• Security best practices are followed