Architecture¶

System design and internal architecture of NornicDB.

📚 Documentation¶

• System Design - High-level architecture overview
• MVCC Lifecycle and Background Work - Debounced mutation work, lifecycle scheduling, and query-protection behavior
• Embedding Search - Embedding storage model and search paths
• Graph-RAG: NornicDB vs Typical - In-memory vs distributed Graph-RAG and latency comparison
• Replication - Clustering and replication internals
• Clustering Roadmap - Future sharding and scaling plans
• Plugin System - Extensibility architecture
• Norns Mythology - Project naming and philosophy

🏗️ Core Components¶

Storage Layer¶

• Badger KV store for persistence
• In-memory engine for testing
• Property graph model
• ACID transactions

Query Engine¶

• Cypher parser and planner
• Query optimizer
• Execution engine
• Result streaming

Index System¶

• HNSW vector index
• B-tree property index
• Full-text BM25 index
• Automatic index selection

Replication¶

• Hot Standby (2-node HA)
• Raft Consensus (3+ node strong consistency)
• Multi-Region (geographic distribution with async replication)
• WAL streaming and automatic failover
• Chaos-tested for extreme latency scenarios

GPU Acceleration¶

• Multi-backend support (Metal, CUDA, OpenCL)
• Automatic CPU fallback
• Memory-optimized operations
• Batch processing

📖 Learn More¶

• System Design - Complete architecture
• MVCC Lifecycle and Background Work - Maintenance behavior and background work control
• Replication - Clustering internals
• Clustering Guide - User documentation
• Performance - Benchmarks and optimization
• Development - Contributing guide

Dive deeper → System Design