Large Language Models

​

Model Capabilities

• Text Generation Generates logically coherent text content based on context and adjusts the output style as needed.
• Language Understanding Accurately understands the meaning of input text and supports conversations that incorporate context.
• Text Translation Provides cross-language generation and understanding capabilities, enabling text translation between different languages.
• Knowledge Q&A Has a rich knowledge base and can answer questions in various fields such as culture, science, and history.
• Code Understanding and Generation Understands and generates code (such as Python, Java, C++, etc.), supports identifying code errors, and provides code suggestions.
• Text Classification and Summarization Understands complex sentences, performs information classification and extraction, and extracts key points from text for automatic summarization.

​

Model Selection

​

API Calls

• ChatCompletion, supporting both streaming mode and standard mode.
• Completion, supporting both streaming mode and standard mode.

​

Code Examples

​

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://api.highwayapi.ai/openai",
    api_key="<Your API Key>",
)

model = "deepseek/deepseek-r1"
stream = True  # or False
max_tokens = 512

chat_completion_res = client.chat.completions.create(
    model=model,
    messages=[
        {
            "role": "system",
            "content": "You are a professional AI documentation assistant.",
        },
        {
            "role": "user",
            "content": "What scenarios can the models provided by JieKou AI be used for?",
        }
    ],
    stream=stream,
    max_tokens=max_tokens,
)

if stream:
    for chunk in chat_completion_res:
        print(chunk.choices[0].delta.content or "", end="")
else:
    print(chat_completion_res.choices[0].message.content)

​

Curl

export API_KEY="<Your API Key>"

curl "https://api.highwayapi.ai/openai/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${API_KEY}" \
  -d '{
    "model": "deepseek/deepseek-r1",
    "messages": [
        {
            "role": "system",
            "content": "You are a professional AI documentation assistant."
        },
       {
            "role": "user",
            "content": "What scenarios can the models provided by JieKou AI be used for?"
        }
    ],
    "max_tokens": 512
}'

​

Key Parameters

​

Basic Parameters

​

Message Roles

Applies only to ChatCompletion.

• content: The message content.
• role: The role of the message author.
  • system: Sets the AI role, telling the model the role or behavior it should adopt.
  • user: The text entered by the user for the model.
  • assistant: The response generated by the model. Users can also prefill examples to tell the model how it should respond to the current request.
• name: Optional. Used to distinguish message authors with the same role.

​

Prompt

Applies only to Completion.

​

Controlling Generation

Both temperature and top_p can control the diversity of generated text. We recommend setting only one of them. The larger the value, the more diverse the generated text. The smaller the value, the more deterministic the generated text.

• temperature: Sampling temperature, which adjusts the randomness of generated text.
• top_p: Nucleus sampling, which controls the cumulative probability of candidate words.
• top_k: Limits the number of candidate words.

• presence_penalty: Presence penalty, which controls the degree of content repetition when the model generates text. If a Token has already appeared in the text, it will be penalized, causing the model to introduce more new Tokens.
• frequency_penalty: Frequency penalty, which controls how often certain words appear in the generated text. It penalizes Tokens each time they appear in the text, thereby reducing the probability of these Tokens appearing in future generation and preventing the model from repeatedly using the same Tokens.
• repetition_penalty: Repetition penalty value, used to suppress or encourage repetition.

​

Output Limits

• max_tokens: The maximum number of Tokens returned in a single request. If the number of Tokens generated by the model exceeds the value of max_tokens, the truncated content will be returned.
• stream: Controls whether the output is streamed. For some models that produce a large amount of output, we recommend setting this to streaming output to prevent overly long output from causing a timeout.
  • true: Streaming output, meaning output is returned as it is generated. The model returns a chunk each time it generates part of the content.
  • false: Returns the result all at once after the model has generated all content.
• stop: Stop sequence. When the text generated by the model contains the string set in stop, the model stops outputting.