Skip to content

04-agent-and-workflow-creation-patterns.md

Latest commit

 

History

History
222 lines (168 loc) · 7.94 KB

04-agent-and-workflow-creation-patterns.md

File metadata and controls

222 lines (168 loc) · 7.94 KB
layout default
title Chapter 4: Agent and Workflow Creation Patterns
nav_order 4
parent AutoAgent Tutorial

Chapter 4: Agent and Workflow Creation Patterns

Welcome to Chapter 4: Agent and Workflow Creation Patterns. In this part of AutoAgent Tutorial: Zero-Code Agent Creation and Automated Workflow Orchestration, you will build an intuitive mental model first, then move into concrete implementation details and practical production tradeoffs.

This chapter focuses on effective natural-language prompts for agent and workflow generation.

Learning Goals

  • write clearer creation prompts for better outputs
  • separate capability requirements from implementation details
  • iterate profile/tool/workflow outputs with intent clarity
  • avoid over-specified or under-specified requests

Creation Strategy

  • define goal, constraints, and success criteria first
  • iterate in small prompt revisions
  • validate generated agents on representative tasks

Source References

Summary

You now have prompt patterns for more reliable AutoAgent creation flows.

Next: Chapter 5: Tooling, Python API, and Custom Extensions

Depth Expansion Playbook

Source Code Walkthrough

autoagent/fn_call_converter.py

The values interface in autoagent/fn_call_converter.py handles a key part of this chapter's functionality:

    """Exception raised when FunctionCallingConverter failed to validate a function call message.

    This typically happens when the LLM outputs unrecognized function call / parameter names / values.
    """

    def __init__(self, message):
        super().__init__(message)

# Inspired by: https://docs.together.ai/docs/llama-3-function-calling#function-calling-w-llama-31-70b
SYSTEM_PROMPT_SUFFIX_TEMPLATE = """
You have access to the following functions:

{description}

If you choose to call a function ONLY reply in the following format with NO suffix:

<function=example_function_name>
<parameter=example_parameter_1>value_1</parameter>
<parameter=example_parameter_2>
This is the value for the second parameter
that can span
multiple lines
</parameter>
</function>

<IMPORTANT>
Reminder:
- Function calls MUST follow the specified format, start with <function= and end with </function>
- Required parameters MUST be specified
- Only call one function at a time
- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after.
- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls

This interface is important because it defines how AutoAgent Tutorial: Zero-Code Agent Creation and Automated Workflow Orchestration implements the patterns covered in this chapter.

autoagent/util.py

The if class in autoagent/util.py handles a key part of this chapter's functionality:

from prompt_toolkit.styles import Style
def debug_print_swarm(debug: bool, *args: str) -> None:
    if not debug:
        return
    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    message = " ".join(map(str, args))
    print(f"\033[97m[\033[90m{timestamp}\033[97m]\033[90m {message}\033[0m")
def print_in_box(text: str, console: Optional[Console] = None, title: str = "", color: str = "white") -> None:
    """
    Print the text in a box.
    :param text: the text to print.
    :param console: the console to print the text.
    :param title: the title of the box.
    :param color: the border color.
    :return:
    """
    console = console or Console()

    # panel = Panel(text, title=title, border_style=color, expand=True, highlight=True)
    # console.print(panel)
    console.print('_'*20 + title + '_'*20, style=f"bold {color}")
    console.print(text, highlight=True, emoji=True)
    


def debug_print(debug: bool, *args: str, **kwargs: dict) -> None:
    if not debug:
        return
    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    message = "\n".join(map(str, args))
    color = kwargs.get("color", "white")
    title = kwargs.get("title", "")

This class is important because it defines how AutoAgent Tutorial: Zero-Code Agent Creation and Automated Workflow Orchestration implements the patterns covered in this chapter.

autoagent/util.py

The UserCompleter class in autoagent/util.py handles a key part of this chapter's functionality:

class UserCompleter(Completer):

    def __init__(self, users: List[str]):
        super().__init__()
        self.users = users
    def get_completions(self, document, complete_event):
        word = document.get_word_before_cursor()
        
        if word.startswith('@'):
            prefix = word[1:]  # 去掉@
            for user in self.users:
                if user.startswith(prefix):
                    yield Completion(
                        user,
                        start_position=-len(prefix),
                        style='fg:blue bold'  # 蓝色加粗
                    )
def pretty_print_messages(message, **kwargs) -> None:
    # for message in messages:
    if message["role"] != "assistant" and message["role"] != "tool":
        return
    console = Console()
    if message["role"] == "tool":
        console.print("[bold blue]tool execution:[/bold blue]", end=" ")
        console.print(f"[bold purple]{message['name']}[/bold purple], result: {message['content']}")
        log_path = kwargs.get("log_path", None)
        if log_path:
            with open(log_path, 'a') as file:
                file.write(f"tool execution: {message['name']}, result: {message['content']}\n")
        return

This class is important because it defines how AutoAgent Tutorial: Zero-Code Agent Creation and Automated Workflow Orchestration implements the patterns covered in this chapter.

autoagent/util.py

The debug_print_swarm function in autoagent/util.py handles a key part of this chapter's functionality:

from prompt_toolkit.formatted_text import HTML
from prompt_toolkit.styles import Style
def debug_print_swarm(debug: bool, *args: str) -> None:
    if not debug:
        return
    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    message = " ".join(map(str, args))
    print(f"\033[97m[\033[90m{timestamp}\033[97m]\033[90m {message}\033[0m")
def print_in_box(text: str, console: Optional[Console] = None, title: str = "", color: str = "white") -> None:
    """
    Print the text in a box.
    :param text: the text to print.
    :param console: the console to print the text.
    :param title: the title of the box.
    :param color: the border color.
    :return:
    """
    console = console or Console()

    # panel = Panel(text, title=title, border_style=color, expand=True, highlight=True)
    # console.print(panel)
    console.print('_'*20 + title + '_'*20, style=f"bold {color}")
    console.print(text, highlight=True, emoji=True)
    


def debug_print(debug: bool, *args: str, **kwargs: dict) -> None:
    if not debug:
        return
    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    message = "\n".join(map(str, args))
    color = kwargs.get("color", "white")

This function is important because it defines how AutoAgent Tutorial: Zero-Code Agent Creation and Automated Workflow Orchestration implements the patterns covered in this chapter.

How These Components Connect

flowchart TD
    A[values]
    B[if]
    C[UserCompleter]
    D[debug_print_swarm]
    E[print_in_box]
    A --> B
    B --> C
    C --> D
    D --> E
Loading