AMA-CMFAI/LAMBDA

This is the offical repository of paper "LAMBDA: A large Model Based Data Agent". https://www.polyu.edu.hk/ama/cmfai/lambda.html

What it solves

LAMBDA is designed to bridge the gap between natural language questions and complex data analysis. It eliminates the need for users to manually write code or perform manual data exploration, allowing them to perform sophisticated data analysis, visualization, and reporting through a conversational interface.

How it works

LAMBDA operates as an agentic system that uses an OpenAI-compatible LLM to translate natural language queries into executable Python and shell tools. It maintains a persistent workspace per conversation, ensuring that variables and generated files (like charts and tables) remain available throughout the session. It can autonomously explore datasets (CSV, Excel, text), write and run code, and then synthesize findings into reports, Jupyter Notebooks, or PDF documents.

Who it’s for

It is intended for data analysts, researchers, and anyone who need to perform data analysis on uploaded datasets without needing to deep technical coding skills for every step of the process.

Highlights

  • Autonomous Exploration: The system can independently inspect datasets, plan analysis, and produce reports.
  • Multilingual Support: Supports English and Chinese UI and report generation.
  • Comprehensive Exports: Ability to export results to Jupyter Notebooks, PDF reports, Markdown bundles, and slides.
  • ** qualitative Artifact Tracking**: Includes a Files panel to track all generated charts, tables, and reports.
  • Flexible Model Configuration: Compatible with any OpenAI-style API endpoint and configurable model list.

Related

  • Project
  • Project
  • Project
  • Project
  • Project