Customizing DeepSeek Responses with Model Parameters in Go

Exploring Model Parameters

Welcome back! In the previous lesson, you learned how to send a simple message to DeepSeek's language model and receive a response using Go . Now, let's take a step further by exploring model parameters that allow you to customize the AI's responses. These parameters are essential for tailoring the model's behavior to your needs. In this lesson, we will focus on four key parameters: MaxTokens, Temperature, PresencePenalty, and FrequencyPenalty . Understanding these parameters will help you control the creativity, length, and content of the AI's explanations.

Using openai.Int and openai.Float

When setting parameters in the openai-go client, you may notice that some fields require you to wrap values using helper functions like openai.Int() and openai.Float() . This is because the API expects pointers to these values, allowing you to omit parameters by passing nil . Using openai.Int() and openai.Float() makes it easy to provide optional values in a type-safe way. For example: openai.Int(150) returns a pointer to the integer value 150 . openai.Float(0.6) returns a pointer to the float value 0.6 . This pattern is used for all optional numeric parameters in the API.

Controlling Response Length with MaxTokens

The MaxTokens parameter sets a hard limit on the number of tokens the AI can generate in its response. A "token" can be a whole word or just part of a word. For example, "tutor" might be one token, while "explanation" could be split into multiple tokens. The exact number of tokens depends on the language and the model. When you set MaxTokens, you specify the maximum number of tokens the AI can produce. This is a strict limit, meaning the model will stop generating text once it reaches this count, even if the answer is incomplete. Here's how you can set MaxTokens to 150 in Go: Go MaxTokens: openai.Int(150), // Limits the response to 150 tokens By setting MaxTokens to 150, you limit the length of the AI's response. This is useful for managing usage and controlling the cost of API requests. Keep in mind that this parameter does not make the model summarize or shorten its response; it simply stops the output when the limit is reached.

Exploring Temperature

Join the 1M+ learners on CodeSignal

Be a part of our community of 1M+ users who develop and demonstrate their skills on CodeSignal

Example: Implementing Model Parameters in Code

Let's bring it all together with a complete Go code example that incorporates all four parameters:

package main

import (
    "context"
    "fmt"
    "log"
    "os"
    "github.com/openai/openai-go"
    "github.com/openai/openai-go/option"
)

func main() {
    // Read OpenAI credentials from environment variables
    apiKey := os.Getenv("OPENAI_API_KEY")
    baseURL := os.Getenv("OPENAI_BASE_URL")

    if apiKey == "" {
        log.Fatal("Please set the OPENAI_API_KEY environment variable.")
    }


    client := openai.NewClient(
        option.WithAPIKey(apiKey),
        option.WithBaseURL(baseURL),
    )

    prompt := "Tell me a fun fact about space."

    // Create a chat completion request with custom parameters
    chatCompletion, err := client.Chat.Completions.New(
        context.Background(),
        openai.ChatCompletionNewParams{
            Messages: []openai.ChatCompletionMessageParamUnion{
                openai.UserMessage(prompt),
            },
            Model:            "deepseek-ai/DeepSeek-V3",
            MaxTokens:        openai.Int(150),    // Limits explanation length
            Temperature:      openai.Float(0.6),  // Controls explanation creativity
            PresencePenalty:  openai.Float(0.5),  // Encourages new concepts
            FrequencyPenalty: openai.Float(0.2),  // Reduces repetitive details
        },
    )

    if err != nil {
        panic(err.Error())
    }

    if len(chatCompletion.Choices) == 0 {
        log.Println("No response choices were returned by the API.")
        return
    }
    reply := chatCompletion.Choices[0].Message.Content

    fmt.Println("Prompt:", prompt)
    fmt.Println("Response:", reply)
}

package main

import (
    "context"
    "fmt"
    "log"
    "os"
    "github.com/openai/openai-go"
    "github.com/openai/openai-go/option"
)

func main() {
    // Read OpenAI credentials from environment variables
    apiKey := os.Getenv("OPENAI_API_KEY")
    baseURL := os.Getenv("OPENAI_BASE_URL")

    if apiKey == "" {
        log.Fatal("Please set the OPENAI_API_KEY environment variable.")
    }


    client := openai.NewClient(
        option.WithAPIKey(apiKey),
        option.WithBaseURL(baseURL),
    )

    prompt := "Tell me a fun fact about space."

    // Create a chat completion request with custom parameters
    chatCompletion, err := client.Chat.Completions.New(
        context.Background(),
        openai.ChatCompletionNewParams{
            Messages: []openai.ChatCompletionMessageParamUnion{
                openai.UserMessage(prompt),
            },
            Model:            "deepseek-ai/DeepSeek-V3",
            MaxTokens:        openai.Int(150),    // Limits explanation length
            Temperature:      openai.Float(0.6),  // Controls explanation creativity
            PresencePenalty:  openai.Float(0.5),  // Encourages new concepts
            FrequencyPenalty: openai.Float(0.2),  // Reduces repetitive details
        },
    )

    if err != nil {
        panic(err.Error())
    }

    if len(chatCompletion.Choices) == 0 {
        log.Println("No response choices were returned by the API.")
        return
    }
    reply := chatCompletion.Choices[0].Message.Content

    fmt.Println("Prompt:", prompt)
    fmt.Println("Response:", reply)
}

In this example, all four parameters are used to customize the AI's response. Notice how each numeric parameter is wrapped with openai.Int() or openai.Float(). This is required by the SDK to properly handle optional values.

Exploring Model Parameters

Using openai.Int and openai.Float

Controlling Response Length with MaxTokens

Exploring Temperature

Join the 1M+ learners on CodeSignal

Encouraging New Topics with PresencePenalty

Reducing Repetition with FrequencyPenalty

Example: Implementing Model Parameters in Code

Summary and Preparation for Practice