5 New Tools for Building AI Agents by OpenAI

OpenAI's March 11, 2025 Agent SDK release empowers developers and businesses to build more robust and sophisticated AI agents. This significant update introduces a suite of powerful tools designed to streamline AI application development, improve reliability, and offer deeper insights into agent performance. Key components include the Responses API, built-in tools, the OpenAI Agents SDK itself, and comprehensive observability tools. These enhancements ultimately facilitate the creation of more intelligent, responsive, and efficient AI solutions.

Table of Contents

• OpenAI's Latest Enhancements
• Responses API: A Deep Dive
• Built-in Tools: Expanding AI Capabilities
• OpenAI Agents SDK: Orchestrating AI Workflows
• Utilizing the OpenAI Agents SDK
• Integrated Observability: Monitoring and Optimization
• Conclusion

OpenAI's Latest Enhancements

This release provides developers with a new set of tools for constructing dependable and efficient AI agents – autonomous AI systems performing tasks on users' behalf. Building on advancements in reasoning, multimodal interactions, and safety features, the update tackles the challenges of transitioning these capabilities into production-ready agents. The previous process often involved extensive prompt engineering, custom logic, and lacked built-in monitoring.

This new suite of APIs and tools simplifies AI agent creation and deployment:

• Responses API: Combines the simplicity of the Chat Completions API with the tool-use capabilities of the Assistants API, simplifying agent development.
• Built-in Tools: Includes web search, file search, and computer interaction, expanding agent functionality.
• Agents SDK: A framework for efficient management of single and multi-agent workflows.
• Integrated Observability Tools: Provides visibility into agent workflows for improved debugging and optimization.

Responses API: A Deep Dive

Terminal Execution Example:

Step 1: Necessary Installations

<code>pip install openai --upgrade</code>

Step 2: Setting the OpenAI API Key

<code>export OPENAI_API_KEY="your-openai-api-key-here"</code>

Step 3: Creating the app.py file

<code>touch app.py</code>

Step 4: app.py Code

from openai import OpenAI
client = OpenAI()

response = client.responses.create(
model="gpt-4o",
input="Give me warm up exercises to do before start of Half Marathon?"
)

print(response.output_text)

Step 5: Executing the Code

<code>python app.py</code>

Sample Output

<code>Warming up before a half marathon is crucial. Here's a sample routine:

1. Dynamic Stretching (5-10 minutes): Leg swings, arm circles, hip circles, torso twists.

2. Light Jogging (5-10 minutes):  Start slow, gradually increasing heart rate.

3. Dynamic Drills (5 minutes): High knees, butt kicks, skipping, bounding.

4. Strides (3-5 bouts): 20-30 second accelerations, then decelerate.

Remember hydration and listen to your body. Good luck!</code>

Key Differences: Responses API vs. Chat Completions

The Responses API represents OpenAI's API evolution, merging Chat Completions' ease of use with Assistants API's power. Key improvements include:

• Stateful vs. Stateless: Responses API is stateful, automatically storing responses and using previous_response_id for conversation continuity, unlike the stateless Chat Completions.
• Enhanced Functionality: It uses Items, a flexible structure for inputs/outputs, supporting file search, web search, structured outputs, and hosted tools.
• Improved Streaming & Event Handling: Uses semantic events, replacing the less user-friendly delta streaming of previous APIs.
• Hosted Tools & Vector Search: Offers one-line integration for various tools and a new Vector Stores Search API.
• Improved API Design: Simplified structure, flattened JSON responses, and form-encoded input support.

Built-in Tools: Expanding AI Capabilities

Step 1: Necessary Installations

<code>pip install openai --upgrade</code>

Step 2: Setting the OpenAI API Key

<code>export OPENAI_API_KEY="your-openai-api-key-here"</code>

Step 3: Creating the app.py file

<code>touch app.py</code>

Step 4: app.py Code

from openai import OpenAI
client = OpenAI()

response = client.responses.create(
model="gpt-4o",
tools=[{"type": "web_search_preview"}],
input="Give me the news of ICC Champions Trophy 2025."
)

print(response.output_text)

Step 5: Executing the Code

<code>python app.py</code>

(Note: The output will depend on current real-time web search results.) File search requires a provided vector store ID.

OpenAI Agents SDK: Orchestrating AI Workflows

The open-source OpenAI Agents SDK simplifies multi-agent workflow orchestration, improving upon the previous Swarm SDK. Key enhancements include:

• Smarter Agents: Easily configure AI models with instructions and built-in tools.
• Seamless Handoffs: Agents efficiently transfer control between each other.
• Robust Guardrails: Built-in safety checks ensure reliable input/output validation.
• Improved Debugging: Visualize agent execution traces for performance optimization.

Utilizing the OpenAI Agents SDK

Step 1: Necessary Installations

<code>pip install openai --upgrade
pip install openai-agents</code>

Step 2: Setting the OpenAI API Key

<code>export OPENAI_API_KEY="your-openai-api-key-here"</code>

Step 3: Creating the app.py file

<code>touch app.py</code>

Step 4: app.py Code (Example using Triage Agent)

# ... (Code as provided in the original input,  handling multiple agents and guardrails) ...

Step 5: Executing the Code

<code>python app.py</code>

(Output will show the responses from the different agents based on the input queries.)

Integrated Observability: Monitoring and Optimization

The integrated observability tools provide detailed insights into agent workflows, including:

• End-to-End Tracing: Visualizes multi-step processes, showing execution times for each agent.
• Step-by-Step Breakdown: Displays API calls and function executions.
• Performance Analysis: Identifies performance bottlenecks based on execution times.
• Resource Monitoring: Tracks AI model usage and token consumption.
• Contextual Information: Shows system instructions and workflow context.

Conclusion

OpenAI's Agent SDK and API updates represent a significant leap forward in AI agent development. The streamlined development process, enhanced reliability, and improved observability tools empower developers to create more sophisticated and efficient AI-driven applications. The combined power of the Responses API, built-in tools, Agents SDK, and observability features simplifies the creation of intelligent, autonomous, and high-performing AI agents. Future updates promise further advancements in AI capabilities.

The above is the detailed content of 5 New Tools for Building AI Agents by OpenAI. For more information, please follow other related articles on the PHP Chinese website!