naialorente/data-analyst

This is a free, open-source web app that helps non-technical and data-adjacent people understand why their numbers changed. You upload one or two CSV files, and the app computes a plain-English breakdown of what moved, which group or segment drove it, and whether the change is statistically real or likely noise. The type of analysis it automates, called root-cause drift analysis, is normally found only in expensive enterprise tools. The most distinctive feature is called "What Changed." You give it two snapshots of similar data, such as this week's sales versus last week's, and it breaks down each metric's change by segment, flags when an overall trend is driven by a shift in your customer or product mix, and runs a statistical test to tell you whether a drop is significant or just random variation. Every number shown is computed by the app itself using standard data science code, the AI is only used to write the plain-English explanation of numbers that were already calculated. Beyond drift analysis, the app includes data cleaning tools that let you fix missing values, remove duplicates, cap outliers, and convert wrongly typed columns with single clicks. You can ask plain-English questions about your data in a chat interface, explore individual columns, run simple predictive models to see which columns drive a target metric, do basic cohort retention analysis, and export a self-contained notebook that reproduces everything you did. The app is bring-your-own-key: you supply an API key from whichever AI provider you already use, including Anthropic, OpenAI, Google Gemini, Groq, or xAI. You can also run it entirely locally with no internet connection for your data by pointing it at a locally running model via Ollama. There is a live demo you can try without installing anything. The license is MIT. The full README is longer than what was shown.