This project provides an open-source multiagent framework designed to automate the complete scientific research lifecycle, from hypothesis generation to publication-ready manuscripts. It targets researchers and power users seeking to establish a 24/7 automated research lab tailored to their specific scientific problems. The primary benefit is the ability to autonomously conduct research, generate experiments, and draft papers with minimal human configuration, offering significant time and resource savings.

How It Works

The system employs a dynamic, multiagent architecture where specialized agents (e.g., IdeationAgent, ExperimentationAgent, WriteupAgent) collaborate under the orchestration of a ManagerAgent. Its core advantage lies in its real-time adaptability to research findings, enabling dynamic workflows rather than predetermined paths. This approach supports continual research through robust context management and allows for natural, real-time human feedback integration. Customization for specific scientific domains is facilitated by the ability to easily add, modify, or remove agents and their associated tools.

Quick Start & Requirements

• Primary Install/Run:
  1. Clone the repository: git clone https://github.com/ltjed/freephdlabor.git && cd freephdlabor
  2. Create and activate Conda environment: conda env create -f environment.yml && conda activate freephdlabor
  3. Set up API keys in .env file for LLM providers (OpenAI, Anthropic, Google, etc.).
  4. Run a task: python launch_multiagent.py --task "Your research idea or direction here"
• Prerequisites: Python 3.11+, Conda environment manager, CUDA-compatible GPU (recommended for experiments), API keys for chosen LLM providers.
• Links: Full demo on YouTube, blog post for introduction, technical report for specifics.

Highlighted Details

• Dynamic Workflows: Agents adapt in real-time to research findings, offering flexibility beyond fixed pipelines.
• Fully Customizable: Agents and tools can be added, modified, or removed with built-in support, allowing adaptation to diverse scientific domains.
• Human-in-the-Loop: Seamless integration of human feedback at any stage of the research process.
• Continual Research: Advanced context management enables sustained, iterative exploration of research problems.
• Automated Prompt Optimization: Utilizes built-in slash commands with Claude Code to analyze LLM call logs and refine agent prompts iteratively.
• HPC Support: Includes a launch_multiagent_slurm.sh template for running on SLURM-based High-Performance Computing clusters.
• Fine-grained LLM Configuration: .llm_config.yaml allows specifying different LLM models for various agents and tools.

Maintenance & Community

The project provides a contribution guide. Support and questions are handled via GitHub issues. Specific community channels (e.g., Discord, Slack), active maintainer lists, or sponsorship details are not explicitly mentioned in the README.

Licensing & Compatibility

The project is licensed under the MIT License. This permissive license generally allows for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

Operation requires obtaining and configuring API keys for various LLM providers. While a GPU is recommended for computational experiments, it is not strictly mandatory for all functionalities. The SLURM script for HPC clusters may require adjustments based on specific cluster configurations. The system's complexity suggests a learning curve for advanced customization.