Oatflake: Federated Knowledge Systems¶
The breakthrough moment that shaped federated knowledge systems
Project Overview¶
The breakthrough moment that shaped federated knowledge systems
Oatflake was exhibited at the 2025 Design Dialogues 2 of the MDEF faculty, marking a pivotal moment in community-driven AI development. Trained on student websites, it demonstrated comprehensive knowledge of projects, methods, resources, and definitions, enabling users to evaluate data through Q&A and voting while creating community servers for free.
This exhibition became the breaking moment that shaped our research toward federated knowledge-sharing systems. The success demonstrated that local AI communities could effectively manage and share knowledge while maintaining complete privacy and control over their data.
Key Innovation¶
🔧 Technical Stack
Combines Python, JavaScript, and HTML with Ollama and ngrok to create local community AI systems that operate entirely on-device, ensuring data sovereignty while enabling collaborative intelligence.
📊 Data Processing
Integrates advanced file processing tools from LangChain to split and prepare text data, along with web scraping capabilities for resource analysis.
🤝 Community Interface
Community members can add data through a remote interface (see BLOB Browser) and retrieve information for search and generation functionality.
Dual-Track Architecture¶
Oatflake operates on two complementary tracks: live chat response for immediate user interaction and background learning for continuous model improvement. This design ensures real-time responsiveness while maintaining the system's learning capabilities through local processing with quantized models.
Local Processing System¶
Complete data sovereignty with responsive AI interactions
1. Local Model Hosting
Utilizes Ollama to run quantized language models locally on consumer hardware, eliminating the need for external API calls and ensuring complete privacy of conversations and data.
2. Secure Tunneling
Integrates ngrok for secure community access, allowing remote users to interact with local AI systems through encrypted tunnels while maintaining host control and privacy.
3. Data Processing Pipeline
Leverages LangChain for advanced text processing, document splitting, and web scraping capabilities, enabling sophisticated data ingestion and preparation for local AI models.
4. Community Integration
Supports Q&A functionality and voting mechanisms for content evaluation, enabling communities to collaboratively assess and improve their knowledge systems.
Federated Knowledge Sharing
The system enables communities to create autonomous knowledge-sharing networks where each node maintains full control over its data while contributing to collective intelligence through voluntary participation and transparent evaluation processes.
Exhibition Results¶
Real-world testing and community validation
The Design Dialogues 2 exhibition demonstrated Oatflake's effectiveness in real community knowledge scenarios, validating our approach to local AI systems:
📚 Knowledge Base
Successfully trained on comprehensive MDEF student website data, demonstrating accurate knowledge retrieval across projects and resources
🔒 Local Processing
100% local operation with quantized models - no external API dependencies or data transmission required
💻 Device Compatibility
Runs efficiently on consumer hardware through Ollama, making advanced AI accessible to community organizers
🌐 Community Features
Integrated Q&A and voting systems enable collaborative content evaluation and continuous improvement
Breakthrough Moment
The exhibition marked a pivotal moment in our research, demonstrating that local AI communities can effectively manage knowledge while maintaining complete autonomy. This success directly influenced our transition toward federated knowledge-sharing systems that prioritize community control and data sovereignty.
Future Development¶
Building the federated knowledge ecosystem
Research Direction
The success of Oatflake at Design Dialogues 2 has shaped our research toward comprehensive federated knowledge-sharing systems that preserve community autonomy while enabling collaborative intelligence.
Enhanced Local Processing: Improved quantized model efficiency and expanded language model support through Ollama integration
Community Tools: Advanced voting mechanisms, content curation systems, and collaborative knowledge validation features
Cross-Platform Integration: Seamless integration with BLOB Browser for enhanced data management and sharing
Privacy-First Architecture: Advanced secure tunneling and encrypted communication protocols for distributed communities
This experiment continues to evolve based on community feedback and technical discoveries. Regular updates reflect ongoing learnings and adaptations.