Introduction

Redis-compatible key-value store on SQLite in Rust.

Quick Example

Python

from redlite import Redlite

db = Redlite("app.db")

db.set("user:1", "Alice")
print(db.get("user:1"))  # "Alice"

db.hset("profile:1", {"name": "Alice", "score": "100"})
print(db.hgetall("profile:1"))

TypeScript

import { Redlite } from 'redlite';

const db = new Redlite('app.db');

await db.set('user:1', 'Alice');
console.log(await db.get('user:1'));  // "Alice"

await db.hset('profile:1', { name: 'Alice', score: '100' });
console.log(await db.hgetall('profile:1'));

Rust

use redlite::Db;

let db = Db::open("app.db")?;

db.set("user:1", b"Alice", None)?;
println!("{:?}", db.get("user:1")?);  // Some(b"Alice")

db.hset("profile:1", &[("name", b"Alice"), ("score", b"100")])?;
println!("{:?}", db.hgetall("profile:1")?);

Go

import "github.com/patte/redlite-go"

db, _ := redlite.Open("app.db")

db.Set("user:1", []byte("Alice"), 0)
fmt.Println(string(db.Get("user:1")))  // "Alice"

db.HSet("profile:1", map[string][]byte{"name": []byte("Alice"), "score": []byte("100")})
fmt.Println(db.HGetAll("profile:1"))

Why Redlite?

Redis has operational overhead—separate process, memory constraints, replication complexity. For embedded systems, mobile apps, or simple services, you just need a fast key-value store that persists. Redlite gives you that: Redis semantics on SQLite, durable by default, single-file backup, compatible with existing clients.

Modes

Mode	Storage	Use Case
Embedded	Memory	Testing, ephemeral caches
Embedded	File	Single-process apps, mobile
Server	Memory	Ephemeral shared cache
Server	File	Multi-client persistent cache

• Embedded — Native bindings (Python, TypeScript, Elixir) or FFI. No network overhead.
• Server — TCP server speaking Redis protocol. Use redis-cli or any Redis client.

How It Works

• Redis types → SQLite tables (hashes → rows, sorted sets → scores, lists → gap-based positioning)
• Writes persist immediately (WAL mode)
• Extras: HISTORY (time-travel), VACUUM, full-text search, vectors, geo

Performance

Mode	GET	SET	Mixed
Embedded (file)	200k+ ops/sec	45k+ ops/sec	90k ops/sec

Benchmarks: macOS M1, 64MB cache. See redlite-bench for details.

Durability: All writes persisted immediately (SQLite WAL mode). Memory: Configurable page cache (default 64MB).

Performance notes:

• History tracking adds write overhead (extra row per mutation)
• FTS/vector indexing adds write latency
• Geo features require computation per query

What’s Implemented

Feature
Strings, Hashes, Lists, Sets, Sorted Sets, Streams	✅
Transactions (MULTI/EXEC/WATCH)	✅
Pub/Sub (server), Blocking Reads (polling/async)	✅
Full-Text Search, JSON, Vectors, Geospatial	✅
History Tracking, Memory/Disk Eviction	✅

Not Supported

• Clustering/replication (use Litestream for backups)
• Lua scripting

Next Steps

Installation · Quick Start