Getting Started with Yoker

You installed an app, package, or module that uses Yoker as its AI backend, and now you are wondering what Yoker is and what you need to do. This page explains it in plain terms and gets you through the one-time setup.

What Yoker is

Yoker is a provider-neutral AI backend for running agentic workflows. It is the layer that sits between the app you are using and the language model that powers it. Instead of each app building its own agent plumbing, they all share Yoker.

In practice this means: any app, package, or module that supports Yoker can run on top of the same Yoker installation and configuration. You set Yoker up once, and every Yoker-based tool benefits.

Why that is useful for you

  • One-time setup for all agentic workflows. Configure Yoker once and every Yoker-based app, package, or module you run afterwards reuses that same setup. No per-app configuration, no repeated wizard runs.

  • Provider- and model-neutral. Yoker does not lock you into a single vendor. It supports Ollama (via its native SDK, free tier available), OpenAI, Anthropic, Google Gemini, and 100+ providers via LiteLLM. You choose the provider and model that fit your cost, privacy, and performance needs. Not once! You can change model and/or provider whenever you want, your apps just follow you. And in the near future, when models go local, you are already set up.

  • Transparent. Yoker does not hide what it does. Prompts, tool calls, and agent decisions are visible and inspectable, so you stay in control of what the AI does on your behalf.

  • Open Yoker is Open Source and free to use. You can change it to your liking.

One config, many apps

Yoker stores its configuration in a single file in your home directory:

~/.yoker.toml

This file describes which backend to use, which model, and any other Yoker settings. Every Yoker-based app reads the same ~/.yoker.toml, so a change you make there applies to all of them at once.

You normally do not edit this file by hand. The bootstrap wizard (below) creates and fills it in for you.

Get started: the bootstrap wizard

The first time you run any Yoker-based app (or yoker itself), Yoker detects that no configuration exists yet and launches the bootstrap wizard. The wizard walks you through the one-time setup in a few short steps:

  1. Provider selection — choose your LLM provider (Ollama, OpenAI, Anthropic, or Google Gemini). For Ollama, choose how Yoker reaches it: through the local app or with an API key. See Getting Started with Ollama for the exact steps to create an account, install the app, and (optionally) generate a key. For Gemini, see Getting Started with Gemini.

  2. Model selection — pick a model from a curated list, or accept the default. You can change this later in ~/.yoker.toml.

  3. Confirmation — Yoker writes the configuration to ~/.yoker.toml and you are ready to go.

After the wizard completes, every other Yoker-based app you start will pick up the same configuration automatically. You only run the wizard once.

Run Yoker directly

If you want to try Yoker on its own before using it through another app, run:

pip install yoker
python -m yoker

The wizard appears on first launch. Once configured, Yoker drops you into an interactive chat session so you can confirm the model responds.

Next steps

  • Getting Started with Ollama — create an Ollama account, install the local app/proxy, and optionally generate an API key. This is the detailed companion to step 1 of the bootstrap wizard.

  • Quick Start — an overview of running Yoker, including interactive and batch modes.

  • Why Yoker? — the design rationale behind Yoker and how it compares to other solutions.