AMA-CMFAI/LAMBDA

What it solves

LAMBDA is a code-free data analysis system that allows users to perform complex data science tasks using natural language instructions instead of writing code manually. It aims to reduce the technical barrier for data analysis and automate the tedious parts of report writing and formatting.

How it works

The system employs a multi-agent architecture consisting of a programmer agent and an inspector agent. These agents work iteratively and generatively to create and debug code based on the user's natural language requests. The system includes a local Code Interpreter (via Jupyter kernel) to execute the code and a user interface that allows humans to intervene in the operational loop.

Who it’s for

It is designed for individuals who need to perform complex data analysis but may lack the advanced coding skills required for traditional data science workflows, as well as researchers and educators who want to accelerate their data-driven applications.

Highlights

• Code-Free Workflow: Perform data analysis via human language instructions.
• Multi-Agent Debugging: A dedicated programmer and inspector agent pair for seamless code generation and debugging.
• Human-in-the-Loop: A robust UI that allows users to directly intervene in the process.
• Reproducibility: Ability to export results and code to Jupyter Notebooks for further analysis.
• Automated Reporting: Automatic generation of reports to save time on formatting.
• Flexible Model Integration: Supports OpenAI and other OpenAI-style interfaces (e.g., Ollama, LiteLLM, LLaMA-Factory) for local or cloud LLMs.