pure-python.md

Pure Python

When to use

The task is pure deterministic logic with no need for LLM reasoning. For example:

• Word counting
• File reading / writing
• Data format conversion
• Math

Design points

• Do not use the @agentic_function decorator
• Do not call llm()
• No runtime parameter needed
• Use a standard Google-style docstring

Examples

def word_count(text: str) -> int:
    """Count the number of words in a text.

    Args:
        text: Input text.

    Returns:
        The word count.
    """
    return len(text.split())

def extract_emails(text: str) -> list[str]:
    """Extract every email address from a text.

    Args:
        text: Input text.

    Returns:
        List of email addresses.
    """
    import re
    return re.findall(r'[\w.+-]+@[\w-]+\.[\w.-]+', text)

Session DAG

Pure Python functions leave no node on the session DAG (unless decorated with @traced).

If you want the call recorded on the DAG, add @traced:

from openprogram.agentic_programming.function import traced

@traced
def word_count(text: str) -> int:
    """Count the number of words in a text."""
    return len(text.split())

The node records the function name, the bound arguments (with self/cls/runtime/callback stripped), and the return value, with expose fixed to 'io'. async def functions are also supported.

Pure Python vs. @agentic_function

Criterion	Pure Python	@agentic_function
Fixed input → fixed output	✓
Needs semantic understanding		✓
Needs natural-language generation		✓
Needs classification / judgement / reasoning		✓
Has a clear algorithm / rule	✓