SentinelRAG — The Evidence Engine

SentinelRAG is a local-first AI investigation workbench that turns scattered, sensitive evidence such as PDFs, emails, screenshots, logs, chat exports, and web records into a structured, verifiable case file. Instead of acting like a black-box chatbot, SentinelRAG builds an evidence engine: it ingests and normalizes files, chunks them intelligently, stores them in a hybrid retrieval pipeline, and answers investigator questions only with source-grounded context, clickable citations, confidence signals, risk scoring, entity correlation, timelines, and exportable reports. The system is designed for privacy-sensitive domains like cybersecurity, compliance, legal review, academic investigation, and enterprise incident response, where every claim must be traceable back to the original evidence. It runs locally by default through Ollama and Milvus Lite, while optionally supporting OpenAI for higher-quality synthesis, voice-ready workflows, and demo-grade reasoning when the user chooses to enable it. Our goal is simple: help teams investigate faster without losing trust, privacy, or auditability. Every clue is preserved, every answer is cited, every risk is explainable, and every report can be defended.