NornicDB is a high-performance graph and vector database engineered for AI agents and knowledge systems. It provides seamless compatibility with Neo4j (Bolt/Cypher) and Qdrant (gRPC), allowing zero-code migration for existing applications. NornicDB enhances knowledge graphs with AI-native features like memory decay, auto-relationships, and GPU-accelerated vector search, enabling emergent meaning and intelligent data interaction.

How It Works

NornicDB utilizes a hybrid execution model, combining shape-specialized streaming executors for common graph traversals with a general Cypher engine. It offers runtime parser mode switching for optimized production performance or detailed debugging. Vector search is accelerated via HNSW construction optimization and a shared seed strategy. Core to its design is hardware acceleration through Metal/CUDA/Vulkan pathways, ensuring high-throughput graph and semantic workloads.

Quick Start & Requirements

• Docker (Recommended): Use timothyswt/nornicdb-arm64-metal-bge:latest for Apple Silicon or timothyswt/nornicdb-amd64-cuda-bge:latest for NVIDIA GPUs. Full images include embedding models.
• From Source: Clone the repository, build with go build, and run ./nornicdb serve.
• Prerequisites: Docker, Go (for source build), GPU with CUDA (NVIDIA) or Metal support (Apple Silicon) for accelerated images.
• Resource Footprint: Docker images vary from ~148MB (headless) to ~4.5GB (GPU + embeddings).
• Links: Docker Hub, Docs

Highlighted Details

• Performance: Demonstrates significant speedups over Neo4j on LDBC benchmarks (e.g., 12x-52x faster), leveraging GPU acceleration.
• Compatibility: Offers drop-in Neo4j compatibility (Bolt, Cypher) and Qdrant gRPC API compatibility.
• AI-Native Features: Integrates memory decay, automatic relationship discovery, and native vector search with hybrid retrieval capabilities.
• Canonical Ledger: Implements versioned facts, temporal validity, and audit-ready transaction logs.
• Heimdall AI Assistant: Provides natural language querying, database status checks, and a plugin system for custom AI actions.

Maintenance & Community

• Contributors: Acknowledges community contributions.
• Roadmap: Key planned features include hybrid retrieval enhancements, distributed fabric, distributed transactions, and GDPR compliance.
• Community: Primarily managed via the GitHub repository.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive MIT license supports commercial use and integration with closed-source projects. Third-party licenses for bundled components are detailed separately.

Limitations & Caveats

• Ongoing Development: Core features like distributed fabric, transactions, and GDPR compliance are still planned, indicating the project is not yet fully mature in these areas.
• Hardware Dependency: Optimal performance relies on specific GPU hardware (NVIDIA CUDA or Apple Metal).
• Benchmark Specificity: Performance claims are based on specific benchmark suites and hardware configurations.