Skip to main content
kitedb — bash — 120×40

KITEDB

v0.2.18● onlinerust core

KiteDB - The Graph Database Built for Speed

❯

$bun add @kitedb/core
$npm install @kitedb/core
./installgit clone

Performance Statistics

SYSTEM_METRICS
NODE_LOOKUP
~125ns
1_HOP_TRAVERSAL
~208ns
DEPENDENCIES
0
CORE
Rust
PERFORMANCE_LEVELBenchmark snapshot: 2026-02-03
CORE_FEATURES
01

UNIFIED_DATA_MODEL

Combine graph relationships and vector similarity in one coherent API—no glue code, no extra services.

Graph + Vector

Traverse relationships and run similarity search in the same query chain.

IVF Vector Index

Approximate nearest-neighbor search tuned with nProbe and threshold.

02

BLAZING_PERFORMANCE

Memory-mapped storage + zero-copy reads keep latency ultra-low without external processes.

~125ns Lookups

~125ns node lookups, ~208ns traversals (p50, Apple M4).

Zero Dependencies

Single-file storage that's easy to back up, sync, and deploy.

03

DEVELOPER_EXPERIENCE

Rust core with idiomatic bindings, MVCC transactions, and type-safe schemas across languages.

Multi-Language

First-class bindings for TypeScript, Python, and more.

MVCC Transactions

Consistent reads with serialized commits for durable writes.

CODE_EXAMPLES
01

SCHEMA_FIRST

Define your schema once, get idiomatic APIs in every language. Type safety where your language supports it.

  • ✓Typed nodes with vector embeddings
  • ✓Typed edges with properties
  • ✓Single-file storage
schema.ts
import { kite } from '@kitedb/core';

// Open database with schema
const db = await kite('./knowledge.kitedb', {
  nodes: [
    {
      name: 'document',
      props: {
        title: { type: 'string' },
        content: { type: 'string' },
        embedding: { type: 'vector' },
      },
    },
    {
      name: 'topic',
      props: { name: { type: 'string' } },
    },
  ],
  edges: [
    {
      name: 'discusses',
      props: { relevance: { type: 'float' } },
    },
  ],
});
02

QUERY_API

Fluent, chainable queries that read like the graph—traversal, vectors, and CRUD in one place.

traversalvector_searchcrud
typescript
// Find all topics discussed by Alice's documents
const topics = db
  .from(alice.id)
  .out('wrote')           // Alice -> Document
  .out('discusses')       // Document -> Topic
  .nodes();

// Multi-hop traversal
const results = db
  .from(startNode.id)
  .out('knows')
  .out('worksAt')
  .take(10)
  .nodes();
ARCHITECTURE

CSR_STORAGE_FORMAT

Compressed Sparse Row format stores adjacency data contiguously for cache-efficient traversal. Memory-mapped files enable zero-copy reads.

RUST_CORE

Memory safety and predictable performance with zero-cost FFI.

IVF_INDEX

Approximate nearest-neighbor search with tunable probe count.

APPEND_ONLY_WAL

Write-ahead logging for durability. Periodic compaction reclaims space.

USE_CASES

RAG_PIPELINES

Store document chunks with embeddings and traverse relationships for context-aware retrieval. Combine vector similarity with graph context for superior RAG results.

Vector embeddings + Graph traversal

KNOWLEDGE_GRAPHS

Model complex relationships with semantic similarity.

RECOMMENDATIONS

Hybrid user-item graphs with embedding similarity.

LOCAL_FIRST_APPS

Embedded architecture with single-file storage. Perfect for desktop apps, CLI tools, and edge computing.

Ready to Start?

Build your first graph database in 5 minutes with our Quick Start guide.

./install
2 min setup
./quickstart
First graph