Summary

Econometrics AI Agent is a specialized LLM-driven system designed to automate complex econometric analysis. It targets researchers, students, and practitioners, offering zero-shot learning capabilities to democratize advanced econometric methods, boost research productivity, and enhance reproducibility without requiring costly LLM fine-tuning.

How It Works

Built on the MetaGPT framework, the agent employs a zero-shot learning approach by integrating deep econometric knowledge through a dedicated 'tool library' and precisely crafted prompts. This design bypasses the need for extensive LLM retraining. Core functionalities include strategic task planning, automated Python code generation and execution for econometric analyses, an error-based reflection mechanism for self-correction, and multi-round conversational capabilities for iterative refinement.

Quick Start & Requirements

Installation involves cloning the repository, configuring API keys and model settings in .env and config/config2.yaml, setting up a Python 3.11 virtual environment using uv, installing Python dependencies (uv pip install -r requirements.txt), installing and building frontend dependencies (cd web && npm install && npm run build), and finally running the application via bash start.sh. Key prerequisites include Python 3.11, uv, Node.js (>= 18 recommended), and valid LLM provider API credentials. The application is accessible at http://localhost:1280. A demo video is available at GitHub Assets, and the related research paper can be found on arXiv: 2506.00856.

Highlighted Details

• Demonstrates superior performance over general LLMs and AI agents on expert-level econometric tasks.
• Leverages a specialized econometric 'tool library' supporting methods like IV-2SLS, DID, RDD, and Propensity Score Methods.
• Features an error-based reflection mechanism for robust solution development.
• Enables iterative task refinement through multi-round user conversations.

Maintenance & Community

The project acknowledges foundational contributions from ChatPilot and MetaGPT. No explicit community channels (e.g., Discord, Slack) or roadmap details are provided in the README.

Licensing & Compatibility

The project is licensed under the Apache License 2.0, which generally permits commercial use and integration into closed-source projects.

Limitations & Caveats

The agent's functionality is dependent on valid API keys and configurations for external LLM providers. The setup process requires managing both Python and Node.js environments and their respective dependencies.