NiubiGEO v0.3.3 is an open-source, self-hosted Generative Engine Optimization (GEO) research tool for tracking brand visibility and competitor presence in AI model responses. Built with Node.js 22.13+ and deployed via Docker or npm, it enables users to run reproducible, multi-model tests using their own API keys (OpenRouter, direct provider APIs, or custom OpenAI-compatible endpoints) while retaining full data sovereignty. The platform structures research into two phases: an initial Domain Test analyzing how models describe a product (business description, category, competitors, associated keywords) and a Keyword Monitoring module for scheduled, repeatable measurements (hourly/daily/weekly) using classified prompts (Discovery, Alternative, Comparison, Brand, Use Case). Version 0.3 introduces a Diff View for comparing snapshots across runs—showing added/removed recommendations, position shifts, and source delta—alongside CSV/JSON exports, UTC date/model filters, per-observation mention charts, a CLI with dry-run support, worker status tracking, failed-model retries, and tag/alias fixes. Evidence is preserved at the granular level: original answers, provider citations, ordinary URLs, text locations, and failures are stored separately per endpoint/model/search configuration. The repository includes 20 real-world case studies (e.g., Notion, Figma, PostHog) demonstrating side-by-side model output analysis, citation tracing, and scheduled run histories. Licensed Apache-2.0, NiubiGEO covers infrastructure and model costs only, positioning itself as a transparent alternative to commercial GEO SaaS platforms like Profound or Semrush AI Visibility.