Tutorials¶

A tutorial takes you from nothing to something that runs. Follow it in order, type the code, and you end up with a working program you can then change.

Start here¶

Getting started: your first AI chat, end to end. Activate a provider by blank import, resolve a credential, send a Chat message, then take a structured Ask response straight into a Go struct. About fifteen minutes.

Add a slash command: build a chat loop that answers /help, /usage and /clear, then write a verb of your own that acts on your program's state instead of asking the model. About twenty minutes.

When a tutorial is the wrong shape¶

Three other kinds of page exist here, and each answers a different question:

• How-to guides: you know what you are doing and want the recipe for one task: fail over across providers, cache a large prompt, run a batch statelessly.
• Reference: you want a fact, such as what a field defaults to, which environment variable is read, what an error means.
• Explanation: you want to understand why the module is shaped the way it is, or what it deliberately does not do.

What is not here yet¶

Two tutorials, and two is not many. The gap worth filling next is a small tool-using agent end to end, with real handlers and a bounded ReAct loop; until it exists, Call tools (the ReAct loop) covers the same ground in how-to form, assuming you already have a client.

No blog post has been backported into this section, and that is deliberate rather than pending. The published posts covering this module argue positions and explain designs (why a four-method interface beats a framework, why one provider is a CLI rather than an API) which is a different register. Moving one here would not make it a tutorial.