How to use this repo

This can serve as a starting point for any AI-driven application that you wanna use the Vercel AI SDK in. I have included few examples and structured the code in a way that should allow you to switch between providers easily and to write modular, easy-to-maintain code.

api

src
├── app
│   ├── api
│   │   ├── completion
│   │   │   └── route.ts
│   │   ├── recipe
│   │   │   └── route.ts
│   │   └── weather
        └── route.ts

Each route contains a handler that will recieve a request from one of our pages, notice that these usually come in pairs. the page.tsx will have something like import { experimental_useObject as useObject } from "ai/react"; from the ai client library and it will be expecting the api route to utilize import { streamObject } from "ai"; something equivalent from the ai library.

Providers

src
└── providers
     ├── ollama.ts
     └── openai.ts

Here you will find a quick way to switch between different providers eg; ollama, openai, etc.. Just import one of the many providers offered by Vercel (full list here) add your config, and export it for usage in your application.

Schemas

src
└── schemas
    ├── recipe.ts
    └── weather.ts

In this directory, you will export a Zod object describing the shape of the shape of data your LLM is going to return the data in. This helps it to provide more accurate and predictable results.

Tools

src
└── tools
    ├── command.ts
    └── weather.ts

Here you will write tools which are a set of functions that can perform some pieces of logic eg; interact with external API to supercharge your model with access to external data or give it the ability to perform operations specific to your application needs.

The template my require some minor clean up, removing some unused files, but for the most part I believe it can serve as a very good basis to get started with!