lida

Why we included this project

Turning a raw CSV into a useful chart usually means a loop of cleaning data, guessing at questions, and tweaking code. LIDA hands that loop to a language model: it summarizes the dataset and proposes analysis goals before writing the visualization code for libraries like matplotlib, seaborn, or Altair, then runs it and can fix its own mistakes. Because it is grammar agnostic and works with OpenAI, Azure, PaLM, Cohere, and Hugging Face models, it slots into an existing stack rather than locking you into a proprietary environment. The same engine will also explain what a chart's code does and flag problems against visualization best practices, which helps when you are doing quick exploratory work or teaching someone how chart code should look. One honest caveat: LIDA generates and executes code, so run it in a sandbox, not against untrusted production data.