Project Introduction: INSIGHT
INSIGHT is an innovative, autonomous artificial intelligence system specifically designed to conduct medical research. It operates through a structured and interactive workflow involving a "boss" agent, "worker" agents, and a dynamic task management system. The system is robust, capable of integrating human-readable data, and can answer complex queries through its interactive components.
Key Components and Workflow
-
Boss Agent: The boss agent is the central coordinator in the INSIGHT platform. It sets objectives, summarizes completed tasks and their results, and formulates a task list that guides the subsequent research actions. It adapts the task list based on the outcomes of each task, ensuring a responsive and efficient research process.
-
Worker Agents: These agents take on specific tasks from the list provided by the boss agent and work to complete them. Once a task is finished, the results are stored in the llama index for reference and further analysis. Worker agents have access to specialized databases through APIs, such as PubMed and MyGene, allowing them to draw on extensive scientific resources to fulfill their tasks.
-
Llama Index: This serves as a repository for storing the results from each task. It also provides context to worker agents, aiding them in completing their tasks by supplying relevant background information and previously gathered data.
-
Task Management and Execution: The generated task list guides the research activities. As tasks are completed, their results can include both text and code, which are processed appropriately before being stored for future reference or to aid in decision-making.
-
Interact with Your Data: Users can engage directly with the llama index, querying it to retrieve insights from the data. By running the
talk_to_index.pyscript, users can ask questions about their data and receive informative responses, enhancing the interactive experience of the platform.
Getting Started with INSIGHT
To begin using INSIGHT, users must complete a few setup steps:
- Register with OpenAI to acquire necessary API keys.
- Configure environment variables or add the API key to a configuration file, ensuring sensitive data like API keys are not exposed to version control systems.
- Install the required dependencies by running
pip install -r requirements.txt. - Start the application with
python main.py.
Output and Data Management
INSIGHT meticulously documents each step of the research process. Results from tasks are saved in an organized directory structure within the out directory. Key findings are distilled into a comprehensive markdown file, which provides:
- A high-level summary and analysis of the data.
- A detailed listing of the key insights gained.
- Innovative hypotheses generated from the findings.
- Suggestions for further research directions.
These outputs are designed for a thorough understanding of the research conducted and facilitate the exploration of future opportunities in medical research.
Financial Considerations
When using INSIGHT, it's important to monitor execution costs. Running tasks typically incurs minimal charges, but costs may increase with the use of more advanced models like GPT-4. Users should be mindful of potential expenses to manage their budget effectively while harnessing the powerful capabilities of INSIGHT.