The Open-Source
Distributed Cluster Database
Built from scratch in Rust. Unify relational SQL, NoSQL Documents, sub-millisecond HNSW Vector Search, Raft Consensus Clustering,Universal Migrations, and native Model Context Protocol (MCP) in a single, lightweight binary.
Core Capabilities & Architecture
Every Database Capability You Need, Built Into One Binary
Eliminate sprawling infrastructure stacks. PingDart DB replaces separate databases for SQL, vector embeddings, caching, graphs, and full-text search.
Sub-Millisecond HNSW Vector Engine
High-dimensional proximity graph supporting Cosine, Euclidean, and Dot Product ANN search across 384 to 3072 dimensions in sub-millisecond P50.
First-Class Model Registry & Auto-Embeddings
Define columns using 'VECTOR GENERATED FROM text USING model'. Never manage float vectors or external embedding services manually.
Cognitive AI Memory (Ebbinghaus Decay)
Hierarchical memory (Working, Episodic, Semantic, LongTerm) with Ebbinghaus exponential decay ranking and automatic memory consolidation.
BM25 & Vector Hybrid Search
Combines dense HNSW vector similarity with BM25 inverted keyword search using normalized α-weighted rank fusion.
Neo4j-Style Labeled Property Graph
Multi-labeled nodes, typed directional relationships (WORKS_AT, KNOWS), and breadth-first search (BFS) shortest-path traversals.
Native Model Context Protocol (MCP)
Built-in JSON-RPC 2.0 stdio MCP server for direct zero-code integration with Claude Desktop, Cursor, Antigravity, and LangChain agents.
Unified SQL + Document NoSQL
One engine for relational schemas, strict DDL types, and JSON document collections. No need to manage separate Postgres and MongoDB clusters.
Raft Consensus Clustering & Sharding
Multi-node consensus with automatic leader election, 16-virtual-node consistent hash ring, and distributed scatter-gather vector queries.
Vectorized Columnar Analytics (OLAP)
SIMD-optimized contiguous memory chunk arrays supporting grouped aggregations (COUNT, SUM, AVG, GROUP BY) at 118,700+ rows/sec.
Cost-Based Query Optimizer (CBO)
Postgres-grade optimizer using HyperLogLog (HLL) NDV cardinality, typed equi-depth histograms, and calibrated cost weights.
OpenTelemetry Tracing & EXPLAIN ANALYZE
Live query execution profiling with recursive PlanNode trees, stage microsecond breakdown, and native OTLP export to Jaeger/Datadog.
Zero-Leak Row-Level Security (RLS)
Injected dynamically into SELECT pipelines using CURRENT_TENANT() and CURRENT_USER() policies for leak-proof multi-tenancy.
AES-256-GCM Envelope Encryption
Authenticated encryption at rest using unique 96-bit nonces, PBKDF2/SHA-256 key derivation, and Poly1305 tamper detection.
Granular RBAC & Revocable API Keys
20 atomic permissions, preloaded roles (SuperAdmin, DataScientist, AppDeveloper), and high-entropy API keys with instant revocation.
Universal Migration Framework
Stream live data from PostgreSQL, MySQL, MongoDB, CSV, and JSON with automated schema drift handling and crash checkpoints.
Protobuf (.pb) ZSTD Backup & PITR
Canonical Protocol Buffers schema, dictionary Zstandard compression (3x-5x smaller), and CDC commit log Point-In-Time Recovery.
Cloud Object Storage Tiering
Async ObjectStore tier supporting AWS S3, Cloudflare R2, MinIO, and Local Filesystem for snapshot offloading and cold archiving.
Cloud-Native Kubernetes Operator
Custom Resource Definition (CRD), HA StatefulSet manifests, pod anti-affinity, and automated Prometheus /metrics endpoint.
Hands-On Interactive Console
Run, Cluster, and Query in Seconds
Choose a workflow below to inspect ready-to-run terminal commands and native SQL syntax.
Why PingDart DB?
PingDart DB vs Traditional Stack
Replace a complex, expensive patchwork of 4+ disconnected databases with one consolidated, high-throughput Rust engine.
| Capability | PingDart DB | PostgreSQL | MongoDB | Pinecone / Vectors |
|---|---|---|---|---|
| Unified SQL & JSON Documents | Native | Partial | ||
| Sub-ms HNSW Vector Search | Native | Slow Ext | Limited | |
| Auto-Generated Text Embeddings | Native | |||
| Cross-Table Semantic JOINs | Native | |||
| Cognitive AI Memory (Ebbinghaus) | Native | |||
| Native Model Context Protocol (MCP) | Native | |||
| Raft Consensus Clustering & Sharding | Native | Complex | Replica Sets | Proprietary |
| Universal Migration (Postgres/MySQL/Mongo) | Native | |||
| Single Self-Contained Binary | Native | SaaS Only |
Built For Every Workload
Who Uses PingDart Cluster DB?
Engineered for engineers, AI researchers, enterprise architects, and self-hosters.
AI & Agent Developers
Store conversational episodes, knowledge vectors, and entities. Plug PingDart directly into Claude Desktop, Cursor, or Antigravity via the native MCP server.
Full-Stack & DevOps Teams
Migrate existing Postgres, MySQL, or Mongo setups in one command. Deploy resilient 3-node Raft clusters with automated failover and K8s Operator.
Enterprise & SaaS Platforms
Cryptographic Row-Level Security, strict tenant QPS/storage quotas, AES-256-GCM envelope encryption at rest, and Point-In-Time Recovery.
Ready to deploy your first cluster?
PingDart DB is open-source and free to use. Spin up a single node in 10 seconds or cluster multiple servers across regions with Raft.