Skip to main content

​
Installation

npm install @ai-billing/deepseek @ai-billing/core @ai-sdk/deepseek ai

​
Overview

The @ai-billing/deepseek package provides middleware for tracking token usage and calculating costs when using DeepSeek models with the Vercel AI SDK. It captures DeepSeek-specific metrics, such as inputCacheReadTokens, ensuring that Prompt Caching costs are accurately reflected.

​
Usage

To use the middleware, wrap your DeepSeek model using wrapLanguageModel from the ai package and pass the createDeepSeekMiddleware.
1

Initialize the DeepSeek provider

First, set up the DeepSeek provider using your API key.
import { createDeepSeek } from '@ai-sdk/deepseek';

const deepSeek = createDeepSeek({
  apiKey: process.env.DEEPSEEK_API_KEY,
});
2

Define model pricing

Set up a price resolver to define the costs for the models you’ll be using. For DeepSeek, you can specify costs for both standard prompt/completion tokens and cached tokens (inputCacheReadTokens).
import { createObjectPriceResolver } from '@ai-billing/core';

const priceResolver = createObjectPriceResolver({
  'deepseek-v4-pro': {
    promptTokens: 0.14 / 1_000_000,
    completionTokens: 0.28 / 1_000_000,
    inputCacheReadTokens: 0.028 / 1_000_000,
  },
});
3

Create the billing middleware

Initialize the DeepSeek billing middleware. You need to provide a destination (such as consoleDestination) where billing events will be sent, along with your priceResolver.
import { createDeepSeekMiddleware } from '@ai-billing/deepseek';
import { consoleDestination } from '@ai-billing/core';

const billingMiddleware = createDeepSeekMiddleware({
  destinations: [consoleDestination()],
  priceResolver: priceResolver,
});
4

Wrap the model

Use wrapLanguageModel from the ai package to apply the billing middleware to your DeepSeek model.
import { wrapLanguageModel } from 'ai';

const wrappedModel = wrapLanguageModel({
  model: deepSeek('deepseek-v4-pro'),
  middleware: billingMiddleware,
});
5

Use the wrapped model

Finally, use the wrapped model with AI SDK functions like generateText or streamText. The billing middleware will automatically track tokens, handle caching metrics, and calculate costs.
import { generateText } from 'ai';

const result = await generateText({
  model: wrappedModel,
  prompt: 'What is the capital of Sweden?',
});