Skip to content

Latest commit

 

History

History
96 lines (79 loc) · 4.38 KB

Overview.md

File metadata and controls

96 lines (79 loc) · 4.38 KB

Tools and Actions

Tools Overview

Tools are predefined functions or methods within our system that perform specific tasks. They are essential building blocks, each encapsulated within a YAML file that outlines its purpose, arguments, and usage. Tools can be utilized individually or combined to form Actions.

Detailed Guide: For a comprehensive guide on Tools, including their configurations and capabilities, please see Tools Detailed Guide.

Example Tool: Google Search

Name: Google Search
Args:
  - query (str)
  - number_result (int, optional)
Command: google_search
Description: |-
  The 'Google Search' tool performs a web search using the Google Custom Search API. It returns a specified number of search results, each containing a URL and a brief description.
Instruction: |-
  To use the 'Google Search' tool, follow these steps:
  1. Call the `google_search` function with the following arguments:
     - `query`: A string representing the search query.
     - `number_result`: (Optional) An integer specifying the number of results to return. Defaults to 5.
  2. The function returns a formatted string containing the search results.
  3. Use the output as needed in your application.
Example: |-
  # Example usage of the Google Search tool:
  from agentforge.tools.GoogleSearch import google_search

  # Search with default number of results
  results = google_search("Python programming")
  print(results)

  # Search with custom number of results
  results = google_search("Machine learning", number_result=10)
  print(results)
Script: agentforge.tools.GoogleSearch

Actions Overview

Actions are structured sequences of one or more Tools, designed to accomplish complex tasks. They allow the chaining of tool functionalities to achieve a desired outcome, orchestrated via YAML files which describe the process flow and inter-tool dynamics.

Detailed Guide: To understand Actions in depth, including how to create and manage them, refer to Actions Detailed Guide.

Example Action: Web Search

Name: Web Search
Description: >-
  The 'Web Search' action combines a Google search, web scraping, and text chunking operations. 
  It first performs a Google search for a specified query using the 'Google Search' tool. 
  Then, it scrapes the text from one of the returned URLs using the 'Web Scrape' tool. 
  Finally, it breaks the scraped text into manageable chunks using the 'Intelligent Chunk' tool.
Example: |-
  # Example usage of the Web Search action:
  query = "OpenAI GPT-4"
  number_result = 5
  
  # Perform Google search
  search_results = google_search(query, number_result)
  print(search_results)
  
  # Choose a URL from the search results
  url = search_results[2][0] # This is selecting the URL for the third result in the dictionary
  
  # Scrape the text from the chosen URL
  scraped_text = get_plain_text(url)
  print(scraped_text)
  
  # Break the scraped text into chunks
  chunk_size = 1
  text_chunks = intelligent_chunk(scraped_text, chunk_size)
  for i, chunk in enumerate(text_chunks):
      print(f"Chunk {i+1}: {chunk}")
Instruction: |-
  To perform the 'Web Search' action, follow these steps:
  1. Use the 'Google Search' tool to perform a Google search:
     - Call the `google_search` function with the query string and the number of results to retrieve.
     - The function returns a list of search results, each containing a URL and a snippet.
  2. Review the search results and choose a URL from the list.
  3. Use the 'Web Scrape' tool to scrape the text from the chosen URL:
     - Call the `get_plain_text` function with the chosen URL.
     - The function returns the textual content of the webpage as a string.
  4. Use the 'Intelligent Chunk' tool to break the scraped text into manageable chunks:
     - Call the `intelligent_chunk` function with the scraped text and the desired chunk size.
     - The function returns a list of text chunks.
  5. Utilize the outputs from each tool as needed for your application.
Tools:
  - Brave Search
  - Web Scrape
  - Semantic Chunk

Note: While Tools provide the fundamental functions of our system, Actions blend these functions to automate workflows and complex processes. Through the strategic use of both Tools and Actions, our system can cater to a variety of automation needs, offering users a versatile platform for their operational requirements.