Eatmydata.ai has been introduced—a tool for local data analysis using artificial intelligence. The application runs entirely within the browser, allowing users to perform complex analytical operations without sending the actual data to remote LLMs, sending only sanitized queries instead.
What Happened
Developers have released Eatmydata.ai, a tool that implements a Question-to-SQL-to-Dashboard cycle on the client side. The technical stack includes the use of SQLite via OPFS (Origin Private File System), WASM-optimized engines for performing NER (Named Entity Recognition) tasks and creating embeddings, as well as a QuickJS sandbox for the secure execution of generated code and building visualizations using Apache ECharts.
Context
The project is an implementation of the Local-first AI concept. Unlike traditional cloud solutions, the Eatmydata.ai architecture moves heavy computations and storage directly into the user's browser, separating execution: the LLM receives only sanitized and obfuscated queries, which prevents the model from accessing the actual content of the files.
Why It Matters for the Industry
The project demonstrates the viability of complex agentic workflows performed entirely on the client side. This paves the way for creating a new generation of analytics tools that reduce dependence on centralized GPU clusters and lower server infrastructure costs by utilizing Edge computing resources.
Why It Matters for Users
For users, this means the ability to quickly and safely analyze sensitive CSV and Excel files using AI. This approach solves the privacy problem, as data is not uploaded to cloud chatbots, and ensures high performance thanks to local computations.
Sources
Author
Look at AI, Editorial Team