Grafeo

Grafeo is a graph database built in Rust from the ground up for speed and low memory use. It runs embedded as a library or as a standalone server, with in-memory or persistent storage and full ACID transactions.

On the LDBC Social Network Benchmark, Grafeo is the fastest tested graph database in both embedded and server configurations, while using a fraction of the memory of some of the alternatives.

Grafeo supports both Labeled Property Graph (LPG) and Resource Description Framework (RDF) data models and all major query languages.

Features:

Core Capabilities

• Dual data model support: LPG and RDF with optimized storage for each
• Multi-language queries: GQL, Cypher, Gremlin, GraphQL, SPARQL and SQL/PGQ
• Embeddable with zero external dependencies - no JVM, no Docker, no external processes
• Multi-language bindings: Python (PyO3), Node.js/TypeScript (napi-rs), Go (CGO), C (FFI), C# (.NET 8 P/Invoke), Dart (dart:ffi), WebAssembly (wasm-bindgen)
• In-memory and persistent storage modes
• MVCC transactions with snapshot isolation

Query Languages

• GQL (ISO/IEC 39075)
• Cypher (openCypher 9.0)
• Gremlin (Apache TinkerPop)
• GraphQL
• SPARQL (W3C 1.1)
• SQL/PGQ (SQL:2023)

Vector Search & AI

• Vector as a first-class type: Value::Vector(Arc<[f32]>) stored alongside graph data
• HNSW index: O(log n) approximate nearest neighbor search with tunable recall
• Distance functions: Cosine, Euclidean, Dot Product, Manhattan (SIMD-accelerated: AVX2, SSE, NEON)
• Vector quantization: Scalar (f32 ? u8), Binary (1-bit) and Product Quantization (8-32x compression)
• BM25 text search: Full-text inverted index with Unicode tokenizer and stop word removal
• Hybrid search: Combined text + vector search with Reciprocal Rank Fusion (RRF) or weighted fusion
• Change data capture: Before/after property snapshots for audit trails and history tracking
• Hybrid graph+vector queries: Combine graph traversals with vector similarity in GQL and SPARQL
• Memory-mapped storage: Disk-backed vectors with LRU cache for large datasets
• Batch operations: Parallel multi-query search via rayon

Performance Features

• Push-based vectorized execution with adaptive chunk sizing
• Morsel-driven parallelism with auto-detected thread count
• Columnar storage with dictionary, delta and RLE compression
• Cost-based optimizer with DPccp join ordering and histograms
• Zone maps for intelligent data skipping (including vector zone maps)
• Adaptive query execution with runtime re-optimization
• Transparent spilling for out-of-core processing
• Bloom filters for efficient membership tests