Agentic_ML

Overview

Agentic_ML leverages the smolagents library from HuggingFace to orchestrate multiple specialized AI agents, spawned by a single agent orchestrator. This could include tasks such as web browsing, PDF processing, file management, data inspection, and package installation.

The core of the system is the AgentOrchestrator, which manages the creation, configuration, and execution of these agents. It supports integration with various Large Language Models (LLMs) from providers like Ollama, or OpenAI.

Installation

Setup

1. Clone the repository:

   git clone https://github.com/your-username/Agentic_ML.git
   cd Agentic_ML

2. Create a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

   Or, if you prefer using pyproject.toml with a tool like pip install . or poetry install (if using Poetry):

   pip install .

Usage

To run the ML Pipeline Agent, you need to specify a configuration file and an initial directive.

First, ensure your config.yaml (or chosen config file) points to your initial directive file (e.g., initial_directive_fmri.txt). For example, in your config.yaml:

run:
  initial_prompt: initial_directive_fmri.txt
  agent_working_dir: "../mri_voxel_model/"

Then, execute the main script:

python src/main.py --config configs/config.yaml

Quick Run

To quickly run the agent and see it in action, use the following command:

python -m agentic_ml.main --config ./agentic_ml/configs/config.yaml --prompt 'Hi could you open the following webpage and extract details from it: https://arxiv.org/pdf/2407.00203, make a detailed and thorough structured report in mark down.'

License

[Specify your license here, e.g., MIT License]