Installation
Overview
The@ai-billing/moonshotai package provides middleware for tracking token usage and calculating costs when using Moonshot AI (Kimi) models with the Vercel AI SDK.
It captures Moonshot-specific metrics, such as reasoningTokens, ensuring that reasoning costs are accurately reflected. Reasoning models such as kimi-k3 always reason before answering, and Moonshot’s usage payload reports reasoning tokens as a subset of the completion tokens — the middleware splits reasoning tokens out of the completion total before billing, so reasoning is never double-counted.
Usage
To use the middleware, wrap your Moonshot AI model usingwrapLanguageModel from the ai package and pass the createMoonshotaiMiddleware.
1
Initialize the Moonshot AI provider
First, set up the provider using your API key.
2
Define model pricing
Set up a price resolver to define the costs for the models you’ll be using. For Moonshot AI’s
kimi-k3, you can specify costs for prompt/completion tokens, cache-read tokens (inputCacheReadTokens), and reasoning tokens (internalReasoningTokens).3
Create the billing middleware
Initialize the Moonshot AI billing middleware. You need to provide a destination (such as
consoleDestination) where billing events will be sent, along with your priceResolver.4
Wrap the model
Use
wrapLanguageModel from the ai package to apply the billing middleware to your Moonshot AI model.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 reasoning metrics, and calculate costs.