# Connect Vercel AI SDK

https://modelportal.ai/docs/integrations/vercel-ai-sdk

> TypeScript toolkit for LLM apps, connected through its OpenAI-compatible provider.

## Use cases

Good for TypeScript server code such as Next.js or Node.js. @ai-sdk/openai-compatible calls `/chat/completions` over the OpenAI-compatible protocol, so generateText, streamText and tool calling all work.

## Tips

Create the provider only in server routes or backend code, keep the key in an environment variable, and never ship it to the browser.

## If it cannot connect

1. Use `https://modelportal.ai/v1` as the API host, without `/chat/completions`.
2. The key starts with `mp-` (older keys: `key_`) and is not paused or revoked; `invalid_api_key` means the key cannot be used.
3. The model ID matches the [catalog](/catalog) and is in the key's allowed models.
4. `insufficient_balance` means the balance is too low; see [common errors](/docs/errors?lang=en) for more.

## Setup

1. Install the dependencies: npm install ai @ai-sdk/openai-compatible.
2. Create a provider with createOpenAICompatible, using the API host below as baseURL and your own key as apiKey.
3. Pass the provider's model to generateText or streamText; keep the key in server-side code.

## Settings

- API host：`https://modelportal.ai/v1`
- API key：`YOUR_API_KEY`
- Model：`MODEL_ID`

```typescript
import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { generateText } from "ai";

const modelportal = createOpenAICompatible({
  name: "modelportal",
  baseURL: "https://modelportal.ai/v1",
  apiKey: "YOUR_API_KEY",
});

const { text } = await generateText({
  model: modelportal("MODEL_ID"),
  prompt: "你好",
});
console.log(text);
```

## FAQ

### Why not @ai-sdk/openai?

Recent @ai-sdk/openai versions default to the Responses API, so the openai-compatible package is the more direct fit. The platform supports Responses too, but not stateful conversations.