Kwaai-AI-Lab

Democratizing access to artificial intelligence by designing, building and maintaining a free Personal AI

---

🎥 Introduction to Kwaai

▶️ Watch our introduction video

---

Table of Contents

• About Kwaai
• Work Tracks
  • Personal AI Development Track
  • Fundamental AI Research Track
  • Policy & Governance Track
• Supporting Infrastructure
• Quick Start Guides
• Network Roadmap
• Community
• Publications & Research
• Documentation

---

About Kwaai

Kwaai is building the future of decentralized, privacy-preserving artificial intelligence through a free Personal AI ecosystem. Our mission is to democratize AI access while maintaining user sovereignty, data privacy, and open collaboration.

Core Principles

• Decentralization: Peer-to-peer infrastructure without centralized control
• Privacy First: Homomorphic encryption and privacy-preserving computation
• Open Development: Open-source core components and transparent research
• User Sovereignty: Users own and control their data and AI agents

Technical Focus

We operate across three main work tracks that encompass development, research, and governance:

1. Personal AI Development - Building production-ready personal AI systems
2. Fundamental AI Research - Advancing the science of AI through novel approaches
3. Policy & Governance - Establishing ethical frameworks and standards

---

Work Tracks

🚀 Track 1: Personal AI Development

Building the infrastructure, applications, and experiences for a free Personal AI ecosystem.

Personal AI Agent Workgroup

Led by Balaji Lakshmanan

PAIAssistant - Email Assistant Demo Repository Demonstration application showcasing PAI capabilities in email management and automation. Features: Email summarization Smart reply suggestions Priority inbox management Calendar integration

Dormant Projects

pAI-OS - Personal AI Operating System Repository | Documentation Core operating system for personal AI agents, providing the foundation for autonomous AI assistance.	kwaai-pai - PAI Assistant Implementation Repository | Documentation Production implementation of the Personal AI Assistant with advanced reasoning capabilities.
pAI-EDUAssistant - Education Beta Educational AI assistant focused on personalized learning experiences.	kwaai-pai-assistant - PAI Assistant Service Backend service infrastructure for PAI Assistant deployment and orchestration.
assistant-temp - Legacy Version Repository Previous generation PAI implementation.

---

KwaaiNet DePIN Workgroup

Led by Reza Rassool

KwaaiNet - Decentralized AI Infrastructure Repository | Documentation Production-ready decentralized physical infrastructure network (DePIN) for AI computation and data sharing. Features: Peer-to-peer networking Distributed compute orchestration Incentive mechanisms Byzantine fault tolerance Cross-platform support (Linux, macOS, Windows)	OpenAI-Petal - API-Compatible Server Repository | Documentation OpenAI API-compatible server for decentralized model inference using the Petals protocol. Features: OpenAI API compatibility Distributed inference Load balancing Model versioning
kwaai-net - Supporting Infrastructure Repository | Documentation Supporting tools and utilities for KwaaiNet infrastructure deployment and management. Features: Node monitoring Network diagnostics Deployment automation Configuration management

---

Personal Agency Workgroup

Led by Steve Vitka

The Personal Agency workgroup focuses on user empowerment through the proprietary Lucid platform, which enables users to maintain control over their AI agents and data. This work is primarily conducted outside of public GitHub repositories.

Focus Areas:

• User consent management
• Agent behavior customization
• Personal data governance
• Agency verification mechanisms

---

User Experience & Applications

kwaai-ui - React Web Application Repository | Documentation Modern web interface for interacting with Kwaai Personal AI services. Features: Responsive design Real-time updates Multi-device support Accessibility compliance	open-webui-kwaai - Open WebUI Fork Repository | Documentation Customized fork of Open WebUI optimized for Kwaai ecosystem integration. Features: KwaaiNet integration Custom themes Enhanced privacy controls Plugin architecture
chrome-extension - Browser Integration Chrome browser extension for seamless PAI Assistant access and web integration. Features: Context-aware assistance Web page summarization Form auto-fill Privacy-preserving analytics

---

🔬 Track 2: Fundamental AI Research

Advancing the scientific foundations of AI through novel architectures, privacy-preserving methods, and interdisciplinary research.

---

State Space Models

Led by Daryle Serrant

Research on state space models for efficient sequence modeling and long-range dependencies. Primary research outputs are published through academic channels and conferences.

Focus Areas:

• Structured state space sequences (S4)
• Efficient attention mechanisms
• Long-context modeling
• Computational efficiency

---

AI 4 Med

Led by Diego Galeano

periodontal - Health Impact Simulator Repository | Documentation AI-powered simulator for periodontal health assessment and treatment outcome prediction. Features: Disease progression modeling Treatment recommendation Risk factor analysis Clinical decision support

---

Physics Inspired

Led by Reza Rassool

Research into novel neural architectures derived from first principles in quantum mechanics and fluid dynamics. Work spans continuous state representations, energy-based learning, non-local interactions, and equivariant geometric architectures. Outputs are published through academic channels; implementation work is available in kwaai-innovative-architectures.

---

Homomorphic Encryption

Led by Sulimon Sattari

VPK — Virtual Private Knowledge Private repository — Kwaai members can request access on Slack Research system enabling confidential vector search for privacy-preserving RAG applications using Partial Homomorphic Encryption (PHE). Encrypted embedding vectors are split across N untrusted shards — each shard sees only 1/N of the corpus and 1/N of queries. Exposes an MCP server for direct integration with Claude and LLM agents. Demonstrated on 1,000 encrypted medical documents with 95–99% ranking preservation. Research published at MPI Workshop, Claremont Graduate University (2025). Capabilities: Five encryption strategies (scrambling, noise, ROME, combined) Multi-shard fan-out with parallel query dispatch MCP server (search_knowledge_base, upload_document, get_audit_log) ~25 µs encryption overhead per 384-dim vector

homomorphic-encryption-mpi-workshop - Private RAG Workshop materials and implementations for privacy-preserving retrieval-augmented generation. Topics: Encrypted document retrieval Secure similarity search Private inference MPI parallelization

privacy-preserving-rag-book - Comprehensive Guide (2026) Comprehensive guide to privacy-preserving retrieval-augmented generation systems. Coverage: Theoretical foundations Implementation patterns Security analysis Production deployment

---

Distributed Knowledge Base & Beyond Graph RAG

CommunitySolidServer - Data Sovereignty Repository | Documentation Implementation of the Solid protocol for decentralized data storage and sovereign data management. Features: W3C Solid specification compliance Decentralized identity (DID) Fine-grained access control Linked data support (RDF)	vector_dbs_benchmarking - Performance Analysis Repository | Results Comprehensive benchmarking suite for vector database performance comparison. Metrics: Query latency Indexing throughput Memory efficiency Scalability characteristics
synthetic-data - Data Generation Repository | Notebooks Methods and tools for generating high-quality synthetic training data. Capabilities: Privacy-preserving synthesis Distribution matching Bias mitigation Quality evaluation

---

AI Alignment Research

kwaai-alignment - Alignment Research Repository | Documentation Research on AI alignment, interpretability, and value learning for trustworthy AI systems. Research Areas: Value alignment methods Interpretability techniques Robustness testing Alignment surveys and datasets

---

📜 Track 3: Policy & Governance

Establishing ethical frameworks, standards, and governance mechanisms for responsible AI development and deployment.

---

AI Policy and Alignment

Led by Steve Vitka

Development of policy frameworks and alignment methodologies to ensure AI systems serve human interests and values.

Focus Areas:

• Ethical AI principles
• Regulatory compliance frameworks
• Alignment verification methods
• Stakeholder engagement processes

Note: Policy research may be included in the kwaai-alignment repository or maintained in separate documentation.

---

AI Trust "Loyal Agents"

Research and development of verifiable mechanisms to ensure AI agents remain loyal to their users' interests.

Focus Areas:

• Loyalty verification protocols
• Behavioral constraints
• Audit mechanisms
• Trust metrics

---

"MyTerms" IEEE P7012

Participation in IEEE P7012 standard development for machine-readable personal privacy terms.

Objectives:

• User-defined privacy preferences
• Machine-readable consent formats
• Automated compliance verification
• Interoperability standards

Note: Standards development work may not have dedicated GitHub repositories as it follows IEEE processes.

---

🛠 Supporting Infrastructure

Operations & DevOps

kwaai-devops - Infrastructure as Code Terraform infrastructure as code for automated deployment and management. Features: Multi-cloud support GitOps workflows Automated scaling Disaster recovery	kwaai-bots - Automation Systems Automation bots for CI/CD, monitoring, and operational tasks. Capabilities: Continuous integration Automated testing Deployment orchestration Incident response
kwaai-uploader - Meeting Tools Tools for recording, uploading, and managing meeting content. Features: Automated transcription Meeting summaries Action item extraction Integration with collaboration tools	kwaai-newsletter - Communication Pipeline Automated newsletter generation and distribution system. Features: Content aggregation Template management Subscriber management Analytics tracking

---

Emerging Projects

TwinTechHub - New Initiative Repository Emerging project focused on digital twin technology and IoT integration. Early Features: Digital twin modeling Real-time synchronization IoT device management Predictive analytics

TwinTechHub-Backend - Backend Infrastructure Backend services and APIs for TwinTechHub platform. Components: RESTful API WebSocket support Database management Authentication/authorization

---

🚀 Quick Start Guides

KwaaiNet - Join the Decentralized Network

Option 1 — Pre-built binary (no Rust required):

# macOS Apple Silicon
curl -L https://github.com/Kwaai-AI-Lab/KwaaiNet/releases/latest/download/kwaainet-aarch64-apple-darwin.tar.gz \
  | tar -xz && sudo mv kwaainet /usr/local/bin/

See the releases page for Linux and Windows downloads.

Option 2 — Build from source:

git clone https://github.com/Kwaai-AI-Lab/KwaaiNet.git
cd KwaaiNet
cargo build --release -p kwaai-cli
# Binary at: target/release/kwaainet

First-time setup:

kwaainet setup                                      # create dirs, write default config
kwaainet config --set public_ip <YOUR_PUBLIC_IP>    # required for map visibility
kwaainet config --set public_name "YourName@kwaai"  # shown on map.kwaai.ai
kwaainet benchmark                                  # measure throughput, saved to cache

Start your node:

kwaainet start --daemon   # reads live network map, auto-selects best local model
kwaainet status           # confirm it's running
kwaainet logs --follow    # watch live output

Model selection is automatic — kwaainet reads map.kwaai.ai, lists your locally installed Ollama models, and picks the best match. Your node appears on the map within 30–60 seconds.

Serve the OpenAI-compatible API:

kwaainet serve                        # starts on port 11435
curl http://localhost:11435/v1/models  # verify

Next Steps:

• Full Documentation
• Network Map
• Join Community

---

pAI-OS - Deploy Your Personal AI

Prerequisites:

• Python 3.9+
• Docker (optional)
• 8GB RAM minimum

Quick Start:

# Clone the repository
git clone https://github.com/Kwaai-AI-Lab/assistant.git
cd assistant

# Set up virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Configure your AI
cp config.example.yaml config.yaml
# Edit config.yaml with your preferences

# Launch pAI-OS
python main.py

Docker Deployment:

# Build the image
docker build -t pai-os .

# Run the container
docker run -d \
  --name my-pai \
  -p 8080:8080 \
  -v ~/.pai:/data \
  pai-os

Configuration Options:

# config.yaml
agent:
  name: "MyPAI"
  model: "gpt-4"
  temperature: 0.7

privacy:
  encryption: true
  local_only: false

integrations:
  - email
  - calendar
  - slack

---

PAI-Assistant - Integrate AI into Your Workflow

Prerequisites:

• Python 3.9+
• API keys (OpenAI or local model)
• Web browser (for UI)

Installation:

# Clone the repository
git clone https://github.com/Kwaai-AI-Lab/kwaai-pai.git
cd kwaai-pai

# Install dependencies
pip install -r requirements.txt

# Set up environment variables
cp .env.example .env
# Add your API keys to .env

# Run the assistant
python app.py

Environment Configuration:

# .env
OPENAI_API_KEY=your_key_here
KWAAI_NODE_URL=http://localhost:8000
PRIVACY_MODE=true
LOG_LEVEL=INFO

Access the UI:

Open your browser to http://localhost:5000

API Usage:

from kwaai_pai import PAIClient

# Initialize client
client = PAIClient(api_key="your_key")

# Send a query
response = client.query("Summarize my emails from today")
print(response.answer)

# Stream responses
for chunk in client.stream("Write a report"):
    print(chunk, end="")

---

🗺 Network Roadmap

✅ Phase 1: Architecture & Foundation

• Native Rust CLI (kwaainet) with full node lifecycle management
• Persistent node identity (Ed25519 keypair, did:peer: DIDs)
• Decentralized Trust Graph (W3C Verifiable Credentials, trust scoring)
• Full Petals/Hivemind DHT compatibility
• OpenAI-compatible API server (kwaainet serve)
• Metal GPU inference — 33+ tok/s on Apple Silicon
• Cross-platform binaries (macOS, Linux, Windows)
• SummitAttendeeVC — live credential issuance at Kwaai Personal AI Summits

Phase 2: Platform Deployment

• 1K+ nodes
• Browser SDK — one-line website integration
• Mobile foundation — iOS/Android native apps
• Enterprise compliance tools (GDPR/HIPAA)
• Optional integration framework (modular storage/identity)
• FiduciaryPledgeVC + VerifiedNodeVC credential issuance

Phase 3: Market Expansion

• 10K+ nodes
• UptimeVC + ThroughputVC peer-witnessed credentials
• Environmental tracking — carbon accountability
• Enterprise routing (minimum trust score thresholds for HIPAA/GDPR workloads)

Phase 4: Scale

• 100K+ nodes
• EigenTrust propagation (Sybil-resistant endorsement-weight decay)
• OS-level integration

---

🤝 Community

Join the Conversation

• Slack Workspace: kwaaiailab.slack.com

  • #general - General discussions
  • #kwaai-net - KwaaiNet development
  • #pai-os - Personal AI OS
  • #research - Research collaborations
  • #support - Technical support

• Website: www.kwaai.ai

• Workgroups: www.kwaai.ai/workgroups

Contributing

We welcome contributions from the community! Here's how to get started:

1. Pick a Project: Browse our repositories
2. Read Guidelines: Check CONTRIBUTING.md in the repository
3. Find an Issue: Look for good first issue or help wanted labels
4. Submit a PR: Follow our pull request template

Contribution Areas:

• 🐛 Bug fixes and testing
• 📝 Documentation improvements
• ✨ Feature development
• 🔬 Research collaborations
• 🎨 UI/UX enhancements
• 🌍 Translations and localization

Code of Conduct

Kwaai-AI-Lab is committed to providing a welcoming and inclusive environment. All participants are expected to:

• Be respectful and considerate
• Provide constructive feedback
• Focus on what's best for the community
• Show empathy towards others

Report any concerns to [email protected]

---

📚 Publications & Research

Active Research Areas

Personal AI Systems

• Decentralized agent coordination
• Privacy-preserving inference
• User preference learning
• Multi-agent collaboration

Novel Architectures

• Physics-inspired neural networks
• Quantum-fluid dynamics models
• State space models for sequences
• Geometric deep learning

Privacy & Security

• Homomorphic encryption for AI
• Secure multi-party computation
• Privacy-preserving RAG
• Federated learning

Medical AI

• Periodontal disease prediction
• Clinical decision support
• Privacy-compliant healthcare AI
• Diagnostic assistance

AI Alignment & Ethics

• Value alignment methods
• Interpretability techniques
• Robustness verification
• Ethical framework development

Publications

Research publications and preprints from Kwaai contributors are listed on the website: www.kwaai.ai/research

---

📖 Documentation

Project Documentation

• KwaaiNet: Wiki | API Docs
• pAI-OS: User Guide | Developer Docs | API Reference
• PAI-Assistant: Integration Guide | API Docs
• CommunitySolidServer: Solid Protocol | Wiki

Technical Resources

• Architecture Overview: System Architecture
• Security Model: Security Documentation
• API Documentation: Unified API Docs
• Deployment Guide: Production Deployment

Learning Resources

• Tutorials: Getting Started Series
• Video Guides: YouTube Channel
• Blog: Kwaai Blog
• Workshops: Workshop Materials

---

📬 Contact & Support

• General Inquiries: [email protected]
• Technical Support: [email protected]
• Research Collaboration: [email protected]
• Security Issues: [email protected]

---

Building the future of decentralized AI, together.