How to Use the New Capabilities of OpenAI’s Responses API

What is the OpenAI Responses API?

Key Features

• Multi-turn Interactions: The Responses API allows for stateful conversations, meaning that the model can maintain context over multiple interactions, resulting in more coherent & engaging discussions.
• Integrated Tools: Built-in tools for functions like web search, file search, and computer use make it easier for developers to create applications that are not just chatbots, but comprehensive digital assistants.
• Granular Control: With the Responses API, developers can specify the context and the tools the model should use, facilitating complex tasks without the complexity of juggling multiple API calls.

Setting Up the API

Step 1: Get an API Key

Step 2: Install the OpenAI Library

1
pip install openai

Step 3: Making Your First API Call

1
2
3
4
5
6
7
8
9
10
import openai

openai.api_key = 'your_api_key'

response = openai.responses.create(
    model="gpt-4o",
    input="What is the capital of France?"
)

print(response.output[0].content[0].text)

Exploring the Features of the Responses API

Multi-Turn Conversations

1
2
3
4
5
6
7
8
9
10
11
12
13
14
# Initial message
initial_response = openai.responses.create(
    model="gpt-4o-mini",
    input="What's your favorite sport?"
)

# Continuing the conversation based on the initial response
follow_up_response = openai.responses.create(
    model="gpt-4o-mini",
    input="Why do you like that sport?",
    previous_response_id=initial_response.id
)

print(follow_up_response.output[0].content[0].text)

Utilizing Built-In Tools

Web Search Tool

1
2
3
4
5
6
7
8
9
web_search_response = openai.responses.create(
    model="gpt-4o",
    input="What are the latest advancements in AI?",
    tools=[
        {"type": "web_search"}
    ]
)

print(web_search_response.output[0].content[0].text)

File Search Tool

1
2
3
4
5
6
7
8
9
file_response = openai.responses.create(
    model="gpt-4o-mini",
    input="Give me the research findings from the provided papers.",
    tools=[
        {"type": "file_search"}
    ]
)

print(file_response.output[0].content[0].text)

Computer Use Tool

1
2
3
4
5
6
7
8
9
computer_use_response = openai.responses.create(
    model="computer-use-preview",
    input="Find the best pizza places nearby.",
    tools=[
        {"type": "computer_use_preview"}
    ]
)

print(computer_use_response.output[0].content[0].text)

Best Practices for Leveraging the Responses API

1. Use Clear Instructions: It’s vital to provide clear & concise prompts for the best results. The more specific you are, the better the responses will be.
2. Utilize the Function Calling Feature: By using function calling, your application can instruct the AI to perform specific actions based on user inputs, streamlining the process.
3. Test & Iterate: Always A/B test different prompts & tool interactions to find out what works best for your application.
4. Monitor Token Usage: Keep an eye on your token usage to avoid unexpected costs, especially when relying heavily on the API for information retrieval.

Making the Most Out of OpenAI's New Capabilities

Why Choose Arsturn?

• Effortless, No-Code Creation: Instantly build your AI chatbot tailored to your brand.
• Adaptable Data Usage: Use your own data seamlessly to train your chatbot, ensuring responses are relevant.
• Insightful Analytics: Gain valuable insights into your audience's interests and questions.

Conclusion