Skip to main content
Streaming allows you to receive agent responses incrementally as they’re generated, rather than waiting for the complete response. This provides real-time feedback to users and creates a more responsive experience. Since language models can take time to generate complete responses, streaming lets users see output as it’s produced, making applications feel faster and more interactive.

​
Basic Usage

Use the stream() method for synchronous calls or astream() for asynchronous calls: 
from hypertic.agents import Agent
from hypertic.tools import tool
from hypertic.models import OpenAIChat

@tool
def get_date() -> str:
    """Get today's date."""
    from datetime import date
    return str(date.today())

model = OpenAIChat(model="gpt-5.2")

agent = Agent(
    model=model,
    tools=[get_date]
)

# Stream responses synchronously
for event in agent.stream("What is today's date?"):
    if event.type == "content":
        print(event.content, end="", flush=True)
Output: 
Today's date is **January 3, 2026**.

​
Accessing Event Attributes

Access individual attributes of streaming events: 
from hypertic.agents import Agent
from hypertic.tools import tool
from hypertic.models import OpenAIChat

@tool
def get_date() -> str:
    """Get today's date."""
    from datetime import date
    return str(date.today())

model = OpenAIChat(model="gpt-5.2")
agent = Agent(
    model=model,
    tools=[get_date],
)

# Stream with event handling
for event in agent.stream("What is today's date?"):
    if event.type == "content":
        print(event.content, end="", flush=True)
    elif event.type == "tool_calls":
        print(f"\nTool Calls: {event.tool_calls}")
    elif event.type == "tool_outputs":
        print(f"\nTool Outputs: {event.tool_outputs}")
    elif event.type == "metadata":
        print(f"\nMetadata: {event.metadata}")
Output: 
Tool Calls: [{'id': 'call_Qu159KQcXuBOFx9S8VyXRFQf', 'type': 'function', 'function': {'name': 'get_date', 'arguments': '{"day":"today"}'}}]
Tool Outputs: {'get_date': '2026-01-03'}
Today's date is **January 3, 2026**.
Metadata: {'model': 'gpt-5.2', 'params': {}, 'finish_reason': 'stop', 'input_tokens': 349, 'output_tokens': 17}

​
Event Types

When streaming, events have different types that you can handle:
Event TypeDescriptionData TypeWhen It Appears
contentText content chunks as they’re generatedstrContinuously as the model generates text
tool_callsTool calls made during executionlist[dict]When the agent decides to use a tool
tool_outputsResults from tool executiondict[str, Any]After tools complete execution
metadataModel information and token usagedictAt the end of the response