iRaven

IRAVEN // ALL QUIET

iRaven

Building for learning

iRaven is a personal project, built for learning. It holds two things: a memory for a person's AI assistant, and access to models for the people who use an app built on iRaven. Both use the same sign-in. Two more things are in design.

Read the papersOpen the console

What iRaven is

iRaven holds two things: Muninn, a memory for a person's AI assistant, and Huginn, access to models for the people of a tenant's app. Three tools serve them: sign-in and devices, usage and ceilings, and sealed storage. Each tool is explained on the page of the thing it serves. Two more things, Aviary and Preen, are in design, and nothing of them is built. A tenant is an app and the team that runs it on iRaven. A person is someone who signs in to a tenant's app. Each tenant's team manages its settings in the console.

In design

Two more things are being designed. Neither is finished. Each text below says what exists.

How the parts fit together

Apps and their people

  • A tenant's app on a phone or a computer, holding no secret
  • A tenant's server, holding its secret key
  • The people who sign in, each on their own devices

iRaven

  • Muninn
  • Huginn

Tools that serve them

  • Sign-in and devices
  • Usage and ceilings
  • Sealed storage

Model providers

  • Providers that iRaven sets up
  • A tenant's own providers (planned)
A person signs in to a tenant's app. The app asks iRaven for a model by its name. iRaven checks the call against the person's ceilings, then sends it to a provider with a key that never leaves iRaven. iRaven stores a person's memory in files that the person's devices seal, so iRaven cannot read them.

Papers

The papers are written in the repository. Each one is public only after iRaven's owner publishes it.

No paper is published yet.

All papers