Vercel AI SDK Integration

Use Cases

The Vercel AI SDK is a TypeScript toolkit for AI applications, commonly used in Next.js, React, Vue, Svelte, and Node.js projects. It provides an OpenAI Compatible provider that lets you integrate Tokener CA into generateText, streamText, tool calls, and chat interfaces.

The integration values are as follows:

Setting	Value
package	`@ai-sdk/openai-compatible`
baseURL	`https://a.skypool.xyz/v1`
apiKey	Consumer API Key, for example `stc-...`
model	Platform model name, for example `gemma4:26b`

Install Dependencies

pnpm add ai @ai-sdk/openai-compatible

npm:

npm install ai @ai-sdk/openai-compatible

Configure Environment Variables

Add the following to .env.local:

SKYPOOL_API_KEY=stc-...

Do not commit real API keys to the repository.

Non-Streaming Call

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";import { generateText } from "ai"; const skypool = createOpenAICompatible({  name: "skypool",  baseURL: "https://a.skypool.xyz/v1",  apiKey: process.env.SKYPOOL_API_KEY,}); const { text, usage, finishReason } = await generateText({  model: skypool.chatModel("gemma4:26b"),  prompt: "In one sentence, explain that the Vercel AI SDK is already integrated with Tokener CA.",  maxOutputTokens: 128,}); console.log(text);console.log(usage);console.log(finishReason);

If your AI SDK version supports calling the provider directly, you can also write the model as skypool("gemma4:26b"). To reduce version differences, the examples in this article use the explicit chatModel.

Streaming Call

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";import { streamText } from "ai"; const skypool = createOpenAICompatible({  name: "skypool",  baseURL: "https://a.skypool.xyz/v1",  apiKey: process.env.SKYPOOL_API_KEY,}); const result = streamText({  model: skypool.chatModel("gemma4:26b"),  prompt: "Introduce Tokener CA in three points.",  maxOutputTokens: 512,}); for await (const textPart of result.textStream) {  process.stdout.write(textPart);}

Next.js Route Handler Example

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";import { streamText } from "ai"; const skypool = createOpenAICompatible({  name: "skypool",  baseURL: "https://a.skypool.xyz/v1",  apiKey: process.env.SKYPOOL_API_KEY,}); export async function POST(request: Request) {  const { messages } = await request.json();   const result = streamText({    model: skypool.chatModel("gemma4:26b"),    messages,    maxOutputTokens: 1024,  });   return result.toUIMessageStreamResponse();}

Tool Call Recommendations

The Vercel AI SDK converts tool definitions into model call parameters. When using Tokener CA, we recommend first validating a minimal tool with a model that supports tool calls, and then integrating it into complex workflows.

If you run into structured output or tool call parameter incompatibilities, first confirm:

Whether the target model supports tool calls
Whether the request goes through https://a.skypool.xyz/v1/chat/completions
Whether you are sending provider-specific parameters that the model does not support

Verify with curl First

curl -X POST "https://a.skypool.xyz/v1/chat/completions" \  -H "Authorization: Bearer <consumer_api_key>" \  -H "Content-Type: application/json" \  --data-raw '{    "model": "gemma4:26b",    "messages": [      {        "role": "user",        "content": "Return a short message confirming the integration succeeded."      }    ],    "max_tokens": 64,    "stream": false  }'

Troubleshooting Checklist

Symptom	How to handle
Environment variable is empty	Make sure `SKYPOOL_API_KEY` is set in `.env.local`, and restart the dev server
`401`	Check whether the Consumer API Key is valid
`404`	Check whether `baseURL` is `https://a.skypool.xyz/v1` and whether the model ID is correct
No output from the streaming endpoint	Verify with the non-streaming `generateText` first, then switch to `streamText`
Tool call fails	Switch to a model that supports tool calls, and reduce provider-specific parameters