Open Source Friday with LIDA - Generate Infographics with LLMS

()

LAA (Language-Agnostic Assistant)

LAA is an open-source project from Microsoft that generates visualizations and infographics for data using a declarative visualization language.
It understands the data, summarizes it, and automatically generates potential visualization goals and hypotheses.
LAA does not focus on cleaning the data and assumes it exists in a potentially formatted manner.

Lia is a tool that creates visualizations from data using natural language.
It provides a Python API and a web API for easy integration.
Lia summarizes the data using an LM to add semantic types and descriptions, then uses an EDA module to generate questions about the data.
For each question, Lia generates code to create a visualization, executes the code, and streams the visualization back to the user interface.
Users can modify the generated code or ask the LM to explain the visualization.
Lia can evaluate visualization code across multiple dimensions, such as bugs, transformation appropriateness, and visualization type, and automatically repair charts based on these assessments.

LiA is a tool that allows users to evaluate and chart their own data.
Users can drag and drop a CSV file into the LiA interface to start the process.
LiA will automatically generate a summary of the data and suggest questions and charts based on the data.
Users can modify the visualizations and ask their own questions using text chat.
LiA can also generate infographics from existing visualizations using a diffusion model.
LiA is open-source and can be used with multiple LLM providers, including OpenAI, Cohere, PaLM, and Hugging Face models.

Lighter is a tool that helps users visualize data and build models, making it more accessible for non-experts.
It supports various input data sources such as CSV and JSON, but may require customization for proprietary visualization tools or large-scale data.
Lighter is not designed to compare data or visualizations directly, but users can generate multiple visualizations and evaluate their quality using the evaluation module.
The project is open-source and welcomes contributions.