Welcome to KnowledgeFlowDB
KnowledgeFlowDB is a next-generation distributed knowledge graph database written in Rust, optimized for AI agentic systems.
Try it now!
Want to get started immediately? Check out our Hosted Service for a fully managed database instance, or follow the quickstart below to self-host.
Status: Production-Ready ๐โ
- 264 tests passing (100%)
- Zero warnings across workspace
- Performance: Up to 13.4x faster than targets
- 5 production crates: core, storage, graph, vector, query
- 3 storage backends: Memory, RocksDB, ScyllaDB
- Horizontally scalable: Linear scaling with cluster nodes
Quick Startโ
Get started with KnowledgeFlowDB in just a few commands:
# Clone the repository
git clone https://github.com/cambriannetwork/knowledgeflow_db.git
cd knowledgeflow_db
# Start the interactive CLI
cargo run -p kfdb
# Run a query
cargo run -p kfdb -- query "MATCH (f:File) RETURN f.path, f.size"
Key Featuresโ
๐ High Performanceโ
- Storage: 45ns (memory), 8ยตs (RocksDB), 2ms (ScyllaDB)
- Graph: BFS/DFS in <10ยตs for 1000 nodes
- Vector Search: 13.4x faster than targets (<0.4ms for 10K vectors)
๐ Horizontal Scalingโ
Throughput = 100,000 QPS ร Number of Nodes
Storage = 500 GB ร Number of Nodes
๐ Powerful Query Language (KQL)โ
MATCH (f:File)-[r:DEFINES]->(fn:Function)
WHERE fn.visibility = 'public'
RETURN f.path, fn.name
ORDER BY fn.name
LIMIT 10
๐ง AI-Native Featuresโ
- Dual embeddings (content + context)
- Semantic code search
- Session tracking
- Self-improvement mechanisms
Architectureโ
KnowledgeFlowDB is built with a layered architecture:
- kfdb-core: Core types and traits (NodeId, EdgeId, Value, Embedding)
- kfdb-storage: Storage engines (Memory, RocksDB, ScyllaDB)
- kfdb-graph: Graph storage and traversal algorithms
- kfdb-vector: HNSW vector search index
- kfdb-query: KQL parser and query executor
- kfdb: CLI tool for interactive queries
Next Stepsโ
- Installation Guide
- Quick Start Tutorial
- CLI Reference
- Architecture Overview
- KQL Syntax
- Deployment Guide
Communityโ
Licenseโ
KnowledgeFlowDB is dual-licensed under MIT OR Apache-2.0.