Plans to support DeepSeek?

[bamzi]

Reviewing the documentation of DeepSeek, the example codes use python OpenAi library which indicates API parity with OpenAI.

Is there a plan to support DeepSeek models and endpoint?

Example: DeepSeek Structured output

from openai import OpenAI

client = OpenAI(
    api_key="<your api key>",
    base_url="https://api.deepseek.com", # <--- deepseek api endpoint
)

[desprit]

config := openai.DefaultConfig(os.Getenv("OPENAI_API_KEY"))
config.BaseURL = "https://api.deepseek.com"
client := openai.NewClientWithConfig(config)

resp, err := client.CreateChatCompletion(
		context.Background(),
		openai.ChatCompletionRequest{
			Model: "deepseek-chat",
			Messages: []openai.ChatCompletionMessage{
				{
					Role:    openai.ChatMessageRoleSystem,
					Content: MakeSystemPrompt(),
				},
				{
					Role:    openai.ChatMessageRoleUser,
					Content: prompt,
				},
			},
			ResponseFormat: &openai.ChatCompletionResponseFormat{
				Type: openai.ChatCompletionResponseFormatTypeJSONObject,
			},
		},
	)

[feng626]

deepseek-reasoner What about this?

[feng626]

status code: 400, status: 400 Bad Request, message: %!s(), body: {"code":20024,"message":"Json mode is not supported for this model.","data":null}

[desprit]

status code: 400, status: 400 Bad Request, message: %!s(), body: {"code":20024,"message":"Json mode is not supported for this model.","data":null}

Why are you asking it here? You are getting an error from deepseek, not from go-openai lib.

[feng626]

type ChatCompletionStreamChoiceDelta struct {
openai.ChatCompletionStreamChoiceDelta
ReasoningContent string json:"reasoning_content,omitempty"
}

type ChatCompletionStreamChoice struct {
openai.ChatCompletionStreamChoice
Delta ChatCompletionStreamChoiceDelta json:"delta"
}

type ChatCompletionStreamResponse struct {
openai.ChatCompletionStreamResponse
Choices []ChatCompletionStreamChoice json:"choices"
}

response := ChatCompletionStreamResponse{}
rawLine, streamErr := stream.RecvRaw()

Rewrote it. OK

[panzhongxian]

👍 @feng626 adding the reasoning_content single line can store the deepseek-r1 model streaming response.