LaminarDB

A high-performance embedded streaming database in Rust. Think "SQLite for stream processing" with sub-microsecond latency.

Features

• Embedded-first: Single binary deployment, no cluster required
• Sub-microsecond latency: < 1μs event processing on the hot path
• SQL-native: Full SQL support with streaming extensions
• Thread-per-core: Linear scaling with CPU cores
• Exactly-once semantics: End-to-end delivery guarantees

Quick Start

# Build all crates
cargo build --release

# Run tests
cargo test

# Run benchmarks
cargo bench

# Start the server
cargo run --bin laminardb

Architecture

LaminarDB uses a ring architecture to separate concerns:

• Ring 0 (Hot Path): Event processing with zero allocations
• Ring 1 (Background): Checkpointing, WAL, compaction
• Ring 2 (Control Plane): Admin API, metrics, configuration

See docs/ARCHITECTURE.md for details.

Documentation

• Architecture
• Development Roadmap
• Contributing Guide
• API Documentation

Performance

License

Apache License 2.0 - see LICENSE for details.

lib.rs:

LaminarDB SQL

SQL interface for LaminarDB with streaming extensions.

This crate provides:

• SQL parsing with streaming extensions (windows, watermarks, EMIT)
• Query planning and optimization via DataFusion
• Streaming-aware physical operators
• SQL-to-operator translation

Streaming SQL Extensions

-- Tumbling window with EMIT
SELECT
  window_start,
  COUNT(*) as event_count
FROM events
GROUP BY TUMBLE(event_time, INTERVAL '5' MINUTE)
EMIT AFTER WATERMARK;

-- Stream-to-stream join
SELECT *
FROM orders o
JOIN order_items i
  ON o.order_id = i.order_id
  AND i.event_time BETWEEN o.event_time AND o.event_time + INTERVAL '1' HOUR;