Crate synadb 

Syna DB

An embedded, log-structured, columnar-mapped database engine written in Rust.

Syna synthesizes the embedded simplicity of SQLite, the columnar analytical speed of DuckDB, and the schema flexibility of MongoDB. It exposes a C-ABI for polyglot integration and is optimized for high-throughput time-series data with AI/ML tensor mapping capabilities.

Features

• Append-only log structure - Fast sequential writes, immutable history
• Schema-free - Store heterogeneous data types without migrations
• AI/ML optimized - Extract time-series data as contiguous tensors
• C-ABI interface - Use from Python, Node.js, C++, or any FFI-capable language
• Delta & LZ4 compression - Minimize storage for time-series data
• Crash recovery - Automatic index rebuild on open
• Thread-safe - Concurrent read/write access

Quick Start

use synadb::{SynaDB, Atom, Result};

fn main() -> Result<()> {
    // Open or create a database
    let mut db = SynaDB::new("my_data.db")?;

    // Write different data types
    db.append("temperature", Atom::Float(23.5))?;
    db.append("count", Atom::Int(42))?;
    db.append("name", Atom::Text("sensor-1".to_string()))?;

    // Read values back
    if let Some(temp) = db.get("temperature")? {
        println!("Temperature: {:?}", temp);
    }

    // Extract history as tensor for ML
    let history = db.get_history_floats("temperature")?;
    println!("History: {:?}", history);

    Ok(())
}

Architecture

Syna uses an append-only log structure. Each entry consists of:

• A fixed-size LogHeader (15 bytes) with timestamp, lengths, and flags
• The key as UTF-8 bytes
• The value serialized with bincode

An in-memory index maps keys to file offsets for O(1) lookups.

Re-exports

pub use engine::close_db;

pub use engine::free_tensor;

pub use engine::open_db;

pub use engine::with_db;

pub use engine::DbConfig;

pub use engine::SynaDB;

pub use error::Result;

pub use error::SynaError;

pub use types::Atom;

pub use types::LogHeader;

pub use types::HEADER_SIZE;

pub use types::IS_COMPRESSED;

pub use types::IS_DELTA;

pub use types::IS_TOMBSTONE;

pub use tensor::optimal_chunk_size;

pub use tensor::DType;

pub use tensor::MmapTensorMeta;

pub use tensor::MmapTensorRef;

pub use tensor::TensorEngine;

pub use tensor::TensorMeta;

pub use tensor::CHUNK_SIZE_LARGE;

pub use tensor::CHUNK_SIZE_MEDIUM;

pub use tensor::CHUNK_SIZE_SMALL;

pub use tensor::DEFAULT_CHUNK_SIZE;

pub use mmap::MmapReader;

pub use gpu::GpuContext;

pub use gpu::GpuTensor;

Modules

compression

Compression utilities for Syna database.

distance

Distance metrics for vector similarity search.

engine

Core database engine for Syna.

error

Error types for Syna database operations.

experiment

Experiment tracking for ML workflows.

ffi

C-ABI Foreign Function Interface for Syna database.

gpu

GPU Direct memory access (optional feature)

hnsw

Hierarchical Navigable Small World (HNSW) index for approximate nearest neighbor search

mmap

Memory-mapped file access for zero-copy reads.

model_registry

Model registry for versioned artifact storage.

tensor

Tensor engine for batch operations on numerical data.

types

Core data types for Syna database.

vector

Vector store for embedding storage and similarity search.