OpenAI Agents
OpenAI Agents is a Python framework to build agentic AI apps in a lightweight, easy-to-use package with very few abstractions. It’s a production-ready upgrade of the experimental framework, Swarm.
Getting Started
To start using OpenAI Agents, follow these steps:
- Install the package
pip install openai-agents
- Set up your OpenAI API key
export OPENAI_API_KEY="<YOUR_KEY>"
How It Works
The Agents SDK has a very small set of primitives:
Agents, which are LLMs equipped with instructions and toolsHandoffs, which allow agents to delegate to other agents for specific tasksGuardrails, which enable the inputs to agents to be validated
Used with Python, these building blocks make it easy to create real-world apps with tool-agent interactions and minimal learning curve. Plus, the SDK also includes tracing to help you debug, evaluate, and fine-tune your agent workflows.
Creating Your First Agents
Here’s a basic example of three agents:
- Triage Agent: Acts as the initial point of contact. It analyzes the user’s question and decides whether to route it to a specialized agent.
- Math Tutor: A specialist agent designed to help with math-related questions.
- History Tutor: A specialist agent focused on historical topics.
from agents import Agent, Runner
math_tutor_agent = Agent(
name="Math Tutor",
handoff_description="Specialist agent for math questions",
instructions="You provide help with math problems. Explain your reasoning at each step and include examples",
)
history_tutor_agent = Agent(
name="History Tutor",
handoff_description="Specialist agent for historical questions",
instructions="You provide assistance with historical queries. Explain important events and context clearly.",
)
triage_agent = Agent(
name="Triage Agent",
instructions="You determine which agent to use based on the user's homework question",
handoffs=[history_tutor_agent, math_tutor_agent],
)
# Run the interaction
result = Runner.run_sync(triage_agent, "I want some help with WW1.")
print(result.final_output)
Integrating with Qdrant
You can connect agents to retrieve or ingest data into a Qdrant collection. Thereby building your knowledge base. Here’s how to enable an agent to retrieve information from Qdrant.
Assume you have a Qdrant collection created using the "text-embedding-3-small" model. The payload structure includes a text field for knowledge storage.
import qdrant_client
from openai import OpenAI
from agents import Agent, function_tool
# Initialize clients
openai_client = OpenAI()
qdrant = qdrant_client.QdrantClient(host="localhost")
# Configuration
EMBEDDING_MODEL = "text-embedding-3-small"
COLLECTION_NAME = "help_center"
LIMIT = 5
SCORE_THRESHOLD = 0.7
@function_tool
def query_qdrant(query: str) -> str:
"""Retrieve semantically relevant content from Qdrant.
Args:
query: The query to search.
"""
embedded_query = openai_client.embeddings.create(
input=query,
model=EMBEDDING_MODEL,
).data[0].embedding
results = qdrant.query_points(
collection_name=COLLECTION_NAME,
query=embedded_query,
limit=LIMIT,
score_threshold=SCORE_THRESHOLD,
).points
if results:
return "\n".join([point.payload["text"] for point in results])
else:
return "No results found."
qdrant_agent = Agent(
name="Qdrant searcher",
handoff_description="Specialist agent for retrieving info from a Qdrant collection",
instructions="You help find answers for user queries using Qdrant. Do not make up any info on your own.",
tools=[query_qdrant],
)
Our qdrant_agent can now query a Qdrant collection whenever deemed necessary to answer a user query.