translator

type LLMTokenUsage struct {
	// InputTokens is the number of tokens consumed from the input.
	InputTokens uint32
	// OutputTokens is the number of tokens consumed from the output.
	OutputTokens uint32
	// TotalTokens is the total number of tokens consumed.
	TotalTokens uint32
}

type OpenAIChatCompletionTranslator interface {
	// RequestBody translates the request body.
	// 	- `raw` is the raw request body.
	// 	- `body` is the request body parsed into the [openai.ChatCompletionRequest].
	//	- `onRetry` is true if this is a retry request.
	//	- This returns `headerMutation` and `bodyMutation` that can be nil to indicate no mutation.
	RequestBody(raw []byte, body *openai.ChatCompletionRequest, onRetry bool) (
		headerMutation *extprocv3.HeaderMutation,
		bodyMutation *extprocv3.BodyMutation,
		err error,
	)

	// ResponseHeaders translates the response headers.
	// 	- `headers` is the response headers.
	//	- This returns `headerMutation` that can be nil to indicate no mutation.
	ResponseHeaders(headers map[string]string) (
		headerMutation *extprocv3.HeaderMutation,
		err error,
	)

	// ResponseBody translates the response body. When stream=true, this is called for each chunk of the response body.
	// 	- `body` is the response body either chunk or the entire body, depending on the context.
	//	- This returns `headerMutation` and `bodyMutation` that can be nil to indicate no mutation.
	//  - This returns `tokenUsage` that is extracted from the body and will be used to do token rate limiting.
	ResponseBody(respHeaders map[string]string, body io.Reader, endOfStream bool) (
		headerMutation *extprocv3.HeaderMutation,
		bodyMutation *extprocv3.BodyMutation,
		tokenUsage LLMTokenUsage,
		err error,
	)
}