run_agent

Runs an AI agent loop from Python code.

Use this function inside regular workflow stages (Tasklet, Job, Hook, or Form logic) when you need multi-step reasoning and tool use.

Syntax

from abstra.ai import run_agent

result = run_agent(
    prompt,
    tools=None,
    max_steps=30,
)

Parameters

Name	Description	Type	Default
prompt (required)	Prompt input for the agent. Can be a string or a list of prompt items (text/files/images).	Prompt | list[Prompt]	None
tools	List of tools the agent can call. Each item can be a callable function or an AgentTools instance.	list[Callable | AgentTools] | None	None
max_steps	Maximum number of reasoning/tool-use iterations.	int	30

Example: Use Task Data

from abstra.ai import run_agent
from abstra.tasks import get_trigger_task

task = get_trigger_task()

result = run_agent(
    prompt=f"""
You are handling a support request.
Customer: {task.payload.get("customer_name")}
Message: {task.payload.get("message")}

Return a concise suggested response.
""",
    max_steps=20,
)

print(result)

Example: Custom Tool

from abstra.ai import run_agent

def lookup_order(order_id: str) -> str:
    """Returns the status of a given order."""
    return "Order is in transit."

result = run_agent(
    prompt="Check order 123 and explain status to the user.",
    tools=[lookup_order],
)

Notes

• Tool names are derived from Python function/method names.
• Add type hints and docstrings to improve tool calling accuracy.