Welcome | Liteset

Liteset is an async port of Apache Superset built on Litestar/ASGI — same dashboards, datasets, users and roles, a completely new web layer.

Liteset 6.0.0 · based on Apache Superset 6.0.0

StarUpstreamFork

Get Started

Overview

Liteset keeps every familiar feature of Apache Superset and replaces the synchronous Flask backend with a single-loop ASGI server. Higher concurrency and lower tail latency at a modest memory cost, identical UX.

Powerful, but async
Liteset preserves the no-code chart builder and SQL Lab from Apache Superset, but the entire web layer runs on a single ASGI event loop instead of pre-forked Flask workers.
Modern databases, native async drivers
Postgres, MySQL, ClickHouse, and Trino use native async drivers (asyncpg, asyncmy, aiochclient, aiotrino). Other databases keep working through a sync-fallback wrapper.
Clean async architecture
Four layers — Controllers, Commands, DAOs, AsyncSession — built on Litestar, SQLAlchemy 2.0 and msgspec. No Flask, no synchronous I/O on the hot path.
Drop-in compatibility
The metadata DB schema, REST API, WebSocket contract and SPA frontend are inherited 1:1 from Apache Superset 6.0.0. Stop Superset, start Liteset on the same database.

Self-serve analytics for anyone

Dashboards
Chart Builder
SQL Lab
Datasets

Explore data and find insights from interactive dashboards.

Key features

40+ pre-installed visualizations inherited from Apache Superset

Full async stack: Litestar + Uvicorn + uvloop + SQLAlchemy 2.0

Native async drivers for Postgres, MySQL, ClickHouse, Trino

msgspec-powered serialization (replaces Marshmallow and Pydantic v1)

Drop-in compatibility: same metadata DB schema, same REST API, same SPA frontend

Native Litestar WebSocket — no separate Node.js superset-websocket service

Flask session cookie / CSRF token compatibility — sessions survive the migration

Auto-generated OpenAPI docs at /swagger/v1

structlog JSON logging out of the box

Performance & Benchmarks

Liteset was benchmarked against Apache Superset 6.0.0 on identical hardware and workloads (SSB SF=10 dataset, Locust load generator). Below — the headline metrics; full methodology and results live in the testing report.