{"id":600,"date":"2025-01-11T15:33:15","date_gmt":"2025-01-11T15:33:15","guid":{"rendered":"https:\/\/smolagents.org\/?post_type=docs&p=600"},"modified":"2025-01-11T15:40:57","modified_gmt":"2025-01-11T15:40:57","password":"","slug":"orchestrate-a-multi-agent-system-%f0%9f%a4%96%f0%9f%a4%9d%f0%9f%a4%96","status":"publish","type":"docs","link":"https:\/\/smolagents.org\/hi\/docs\/orchestrate-a-multi-agent-system-\ud83e\udd16\ud83e\udd1d\ud83e\udd16\/","title":{"rendered":"Orchestrate a multi-agent system \ud83e\udd16\ud83e\udd1d\ud83e\udd16"},"content":{"rendered":"

In this notebook we will make a multi-agent web browser: an agentic system with several agents collaborating to solve problems using the web!<\/strong><\/p>\n\n\n\n

It will be a simple hierarchy, using a ManagedAgent<\/code> object to wrap the managed web search agent:<\/p>\n\n\n\n

Copied<\/p>\n\n\n\n

              +----------------+\n              | Manager agent  |\n              +----------------+\n                       |\n        _______________|______________\n       |                              |\n  Code interpreter   +--------------------------------+\n       tool          |         Managed agent          |\n                     |      +------------------+      |\n                     |      | Web Search agent |      |\n                     |      +------------------+      |\n                     |         |            |         |\n                     |  Web Search tool     |         |\n                     |             Visit webpage tool |\n                     +--------------------------------+<\/pre>\n\n\n\n
Let\u2019s set up this system.<\/p>\n\n\n\n
Run the line below to install the required dependencies:<\/p>\n\n\n\n
Copied<\/p>\n\n\n\n
!pip install markdownify duckduckgo-search smolagents --upgrade -q<\/pre>\n\n\n\n
Let\u2019s login in order to call the HF Inference API:<\/p>\n\n\n\n
Copied<\/p>\n\n\n\n
from huggingface_hub import login\n\nlogin()<\/pre>\n\n\n\n
\u26a1\ufe0f Our agent will be powered by\u00a0Qwen\/Qwen2.5-Coder-32B-Instruct<\/a>\u00a0using\u00a0HfApiModel<\/code>\u00a0class that uses HF\u2019s Inference API: the Inference API allows to quickly and easily run any OS model.<\/p>\n\n\n\n
Note:<\/em>\u00a0The Inference API hosts models based on various criteria, and deployed models may be updated or replaced without prior notice. Learn more about it\u00a0here<\/a>.<\/p>\n\n\n\n
Copied<\/p>\n\n\n\n
model_id = \"Qwen\/Qwen2.5-Coder-32B-Instruct\"<\/pre>\n\n\n\n
<\/a>\ud83d\udd0d Create a web search tool<\/h2>\n\n\n\n
For web browsing, we can already use our pre-existing\u00a0DuckDuckGoSearchTool<\/code><\/a>\u00a0tool to provide a Google search equivalent.<\/p>\n\n\n\n
But then we will also need to be able to peak into the page found by the DuckDuckGoSearchTool<\/code>. To do so, we could import the library\u2019s built-in VisitWebpageTool<\/code>, but we will build it again to see how it\u2019s done.<\/p>\n\n\n\n
So let\u2019s create our VisitWebpageTool<\/code> tool from scratch using markdownify<\/code>.<\/p>\n\n\n\n
Copied<\/p>\n\n\n\n
import re\nimport requests\nfrom markdownify import markdownify\nfrom requests.exceptions import RequestException\nfrom smolagents import tool\n\n\n@tool\ndef visit_webpage(url: str) -> str:\n    \"\"\"Visits a webpage at the given URL and returns its content as a markdown string.\n\n    Args:\n        url: The URL of the webpage to visit.\n\n    Returns:\n        The content of the webpage converted to Markdown, or an error message if the request fails.\n    \"\"\"\n    try:\n        # Send a GET request to the URL<\/em>\n        response = requests.get(url)\n        response.raise_for_status()  # Raise an exception for bad status codes<\/em>\n\n        # Convert the HTML content to Markdown<\/em>\n        markdown_content = markdownify(response.text).strip()\n\n        # Remove multiple line breaks<\/em>\n        markdown_content = re.sub(r\"\\n{3,}\", \"\\n\\n\", markdown_content)\n\n        return markdown_content\n\n    except RequestException as e:\n        return f\"Error fetching the webpage: {str(e)}\"\n    except Exception as e:\n        return f\"An unexpected error occurred: {str(e)}\"<\/pre>\n\n\n\n
Ok, now let\u2019s initialize and test our tool!<\/p>\n\n\n\n
Copied<\/p>\n\n\n\n
print(visit_webpage(\"https:\/\/en.wikipedia.org\/wiki\/Hugging_Face\")[:500])<\/pre>\n\n\n\n
<\/a>Build our multi-agent system \ud83e\udd16\ud83e\udd1d\ud83e\udd16<\/h2>\n\n\n\n
Now that we have all the tools search<\/code> and visit_webpage<\/code>, we can use them to create the web agent.<\/p>\n\n\n\n
Which configuration to choose for this agent?<\/p>\n\n\n\n