Skip to content

Latest commit

 

History

History
148 lines (104 loc) · 4.67 KB

code_tool_workflow.md

File metadata and controls

148 lines (104 loc) · 4.67 KB
Raw

ReXia.AI CodeToolWorkflow Class

Overview

The CodeToolWorkflow class is a specialized workflow implementation in the ReXia.AI framework. It extends the BaseWorkflow class and implements a process for generating and executing custom code tools to solve complex tasks. This workflow is designed to be used within an Agent instance and leverages a large language model (LLM) to dynamically create Python functions that can assist in completing the given task.

Table of Contents

Class Attributes

  • llm: The language model used by the workflow.
  • task: The task that the workflow is designed to perform.
  • verbose: A flag used for enabling verbose mode.
  • channel: The collaboration channel for the workflow.
  • max_attempts: The maximum number of attempts for code generation.

Methods

__init__(self, llm: Any, task: str, verbose: bool = False, max_attempts: int = 3) -> None

Initializes a CodeToolWorkflow instance.

Parameters:

  • llm: The language model used by the workflow.
  • task: The task assigned to the workflow.
  • verbose: A flag for enabling verbose mode. Defaults to False.
  • max_attempts: The maximum number of attempts for code generation. Defaults to 3.

_run_task(self) -> None

Internal method that executes the main CodeToolWorkflow process.

run(self) -> str

Public method to initiate the workflow execution.

Usage

Here's an example of how to use the CodeToolWorkflow class with an Agent in ReXia.AI:

import os
from rexia_ai.llms import RexiaAIOpenAI
from rexia_ai.agents import Agent
from rexia_ai.workflows import CodeToolWorkflow

# Set up the language model
YI_LARGE_API_KEY = os.getenv("YI_LARGE_API_KEY")
llm = RexiaAIOpenAI(
    base_url="https://api.01.ai/v1",
    model="yi-large",
    temperature=0,
    api_key=YI_LARGE_API_KEY,
)

# Define the task
task = "Calculate the sum of squares for the first 100 prime numbers."

# Create an Agent instance with CodeToolWorkflow
agent = Agent(
    llm=llm,
    task=task,
    workflow=CodeToolWorkflow,
    verbose=True
)

# Run the agent (which will use the CodeToolWorkflow)
response = agent.invoke()

# Print the response
print("Task:", task)
print("Response:", response)

Components

The CodeToolWorkflow primarily utilizes the following components:

  1. Large Language Model (LLM): Generates Python code based on the given task.
  2. ContainerisedCodeTester: Executes the generated code in an isolated Docker container.

Large Language Model (LLM)

The LLM is responsible for generating Python code that can solve the given task. Key features include:

  • Receiving a task description and generating appropriate Python functions
  • Handling complex tasks by breaking them down into manageable steps
  • Iteratively improving the generated code based on execution results

ContainerisedCodeTester

The ContainerisedCodeTester class is responsible for executing the generated Python code in isolated Docker containers. Key features include:

  • Initialization with a specified Docker image (default: "python:3.12-slim") and execution timeout
  • Execution of code in a controlled Docker environment
  • Writing code to a temporary directory
  • Running the Docker container with the provided code
  • Parsing and returning the execution results

Key methods:

  • execute_code(code): Executes the provided code in a Docker container
  • _write_files(tmpdir, code): Writes the code files to a temporary directory
  • _run_container(tmpdir): Runs the Docker container with the provided code
  • _parse_output(output): Parses the output from the Docker container

The ContainerisedCodeTester ensures that the generated code runs in a secure, isolated environment with:

  • Limited memory (128MB)
  • Limited CPU quota
  • No network access
  • Read-only access to the code files

Dependencies

  • typing
  • docker
  • tempfile
  • os
  • json
  • ReXia.AI components (BaseWorkflow, BaseMemory, CollaborationChannel, TaskStatus, Component)
  • ReXia.AI Agent class

Ensure all dependencies are installed and properly imported.

Contributing

We welcome contributions to improve the ReXia.AI framework. Please follow these steps to contribute:

  1. Fork the repository.
  2. Create a new branch (git checkout -b feature-branch).
  3. Make your changes.
  4. Commit your changes (git commit -m 'Add new feature').
  5. Push to the branch (git push origin feature-branch).
  6. Create a new Pull Request.

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.