Documentation

Everything there is to know about running Sumimi: how to put it on a box, how to put it on a site, what it promises and how much of your traffic it will take.

Getting started

• Install

  Download the release archive, install the one executable, create the admin account, start the server and put the tracker on a site.

• Self-host

  A complete deployment on a cheap VPS: systemd, a reverse proxy, TLS certificates, the firewall, and backups you can actually restore from.

Using it

• Tracker

  Every way to put sumimi.js on a site: the script tag, TypeScript, self-hosting, a CDN, several websites at once, and the web vitals.

• HTTP API

  Every endpoint, with the request bodies, the response shapes, the authentication rules and what each metric means.

• CLI

  The offline analyser: read the database file directly or talk to a running server. One binary, two modes — and an MCP server for agents.

Reference

• Memory & capacity

  The promise, where the memory goes, the four limits that keep it true, and the table a deployment is sized from.

• Privacy

  Exactly what is collected, what is refused, how the visitor identifier works, and how to erase it.

• FAQ

  The questions that come up most: how it differs from Umami, whether you need a domain, and which binary to download.

And the rest

The blog has the story behind all of this — why the storage engine changed, how the numbers were measured, and the traps in measuring them. The repository has the design documents behind the pages: AGENTS.md is the architecture, MEMORY.md is the memory strategy.