MindGraph OS is an AI-powered knowledge operating system that transforms documents, codebases, and research into a living knowledge graph with autonomous reasoning agents.
Unlike traditional RAG systems that simply retrieve text, MindGraph OS builds a temporal knowledge graph of your ideas, tracks how your thinking evolves over time, detects contradictions, and proactively surfaces insights.
It acts as a second brain for developers, researchers, and teams.
Most AI tools work like this:
User Question
↓
Vector Search
↓
LLM Answer
MindGraph OS introduces knowledge reasoning:
Documents + Code + Notes
↓
Embedding Index (Qdrant)
↓
Knowledge Graph (Neo4j)
↓
Autonomous AI Agents
↓
Insight Generation
↓
Human Decision Making
This turns a static knowledge base into a self organizing intelligent system.
MindGraph can ingest and understand:
- PDFs
- Markdown notes
- Plain text documents
- Entire GitHub repositories
- Python codebases
It automatically extracts concepts, relationships, and structures.
Every concept is stored with timestamps.
You can ask:
- “How has my understanding of transformers evolved?”
- “When did this concept first appear in my research?”
- “What knowledge has become outdated?”
Graph nodes store:
Concept
Chunk
Document
Contradiction
Relationships:
MENTIONS
RELATED_TO
INVOLVES
BELONGS_TO
MindGraph includes background AI agents that continuously analyze your knowledge base.
Detects conflicting information between documents.
Example:
Doc A: Transformers outperform CNNs
Doc B: CNNs outperform Transformers
Agent flags:
⚠️ Knowledge Conflict Detected
Finds hidden relationships between ideas.
Example:
Graph Neural Networks
↕
Knowledge Graph Embeddings
Identifies knowledge that hasn't been updated recently.
Example:
Concept last updated: 92 days ago
Status: STALE
Traditional RAG:
Vector search only
MindGraph uses hybrid retrieval:
Vector similarity (Qdrant)
+
Graph traversal (Neo4j)
This allows context-aware reasoning across documents.
MindGraph can analyze entire GitHub repositories.
Using tree-sitter, it extracts:
- classes
- functions
- module structure
- imports
This enables queries like:
Explain this repository architecture
or
Where is authentication implemented?
Background agents push insights using WebSockets.
Examples:
⚠️ Contradiction detected in research notes
🔗 New connection discovered between concepts
📉 Knowledge staleness detected
These insights appear instantly in the UI.
MindGraph can generate a Knowledge Debt Report summarizing:
- contradictions
- outdated knowledge
- missing conceptual connections
This helps researchers and engineers maintain clean knowledge structures.
| Feature | Traditional RAG | MindGraph OS |
|---|---|---|
| Vector Search | ✅ | ✅ |
| Knowledge Graph | ❌ | ✅ |
| Temporal Knowledge Tracking | ❌ | ✅ |
| Contradiction Detection | ❌ | ✅ |
| Autonomous Agents | ❌ | ✅ |
| Codebase Understanding | ❌ | ✅ |
| Real-time Knowledge Insights | ❌ | ✅ |
Streamlit UI
│
WebSocket Updates
│
FastAPI
│
┌───────────────┼───────────────┐
│ │ │
Qdrant Neo4j Redis
(Vector DB) (Knowledge Graph) (Cache)
│
│
Ollama
(Local LLM + Embeddings)
Autonomous Agents
├── Contradiction Agent
├── Connection Agent
├── Staleness Agent
└── Knowledge Debt Generator
- FastAPI – high-performance async API
- Python 3.11
- Qdrant – vector database
- Neo4j – graph database
- Redis – caching and pub/sub
- Ollama – local LLM runtime
- Llama3 / Mistral / Phi3
- nomic-embed-text embeddings
- tree-sitter
- LangGraph
- Streamlit
mindgraph-os
│ |
│ |── api
│ │ ├── upload.py
│ │ ├── query.py
│ │ ├── graph.py
│ │ └── ws.py
│ │
│ ├── agents
│ │ ├── contradiction_agent.py
│ │ ├── connection_agent.py
│ │ ├── staleness_agent.py
│ │ └── workflow.py
│ │
│ ├── services
│ │ ├── parser.py
│ │ ├── chunker.py
│ │ ├── embedder.py
│ │ ├── graph_builder.py
│ │ ├── retriever.py
│ │ ├── generator.py
│ │ └── github_ingestor.py
│ │
│ └── db
│ | ├── qdrant_client.py
│ | └── neo4j_client.py
│ └── app.py
│
└── docker-compose.yml
git clone https://github.com/yourusername/mindgraph-os.git
cd mindgraph-os
pip install -r requirements.txt
docker compose up -d
This starts:
- Neo4j
- Qdrant
- Redis
ollama pull llama3
ollama pull nomic-embed-text
uvicorn app.main:app --reload
streamlit run app.py
Once running, you can ask:
Explain this research paper
What contradictions exist in my knowledge?
Explain this GitHub repository architecture
Which concepts in my research are outdated?
Track how your understanding of topics evolves.
Analyze large repositories with natural language.
Maintain a living knowledge graph of ideas.
Planned upgrades:
- Multi-tenant collaboration
- Knowledge personas (researcher / engineer modes)
- Graph visualization with D3
- Autonomous research agents
- arXiv paper ingestion
- citation graph analysis
- real-time knowledge dashboards
Contributions are welcome!
Steps:
- Fork the repository
- Create a feature branch
- Commit changes
- Open a pull request
Please follow clean code practices and include tests when possible.
MIT License.
You are free to use this project for research, commercial products, or learning.
MindGraph OS was inspired by the idea that:
Knowledge should not just be stored — it should evolve, reason, and challenge itself.
Please consider starring the repository and sharing it with others.
It helps the project grow and motivates further development.
MindGraph OS: Where your knowledge becomes intelligent.