Sift – Analyse every file with local LLMs (github.com)

Sift, a new tool for analyzing files using local large language models (LLMs), has been launched, enabling users to summarize and classify various file types entirely offline. By pointing Sift at a file or a folder, it detects the file's actual type regardless of its extension, extracts its content with specialized parsers, and utilizes locally hosted models like Ollama or LM Studio to generate meaningful summaries. This privacy-focused solution ensures that users' data remains solely on their machines, addressing growing concerns over cloud-based AI processing and data security. The significance of Sift lies in its comprehensive, zero-dependency approach that runs seamlessly on any system with Python 3.9 or newer. It supports a wide array of file formats including text, spreadsheets, images, and code, providing structure-aware extraction that can identify key characteristics such as potential security risks in scripts. Importantly, Sift doesn’t solely rely on model outputs; it incorporates validation measures to flag discrepancies in the data it processes, which enhances reliability. This capability makes Sift a valuable asset for developers, data analysts, and IT professionals who require efficient, secure, and private file analysis in an increasingly data-driven environment.