NiubiGEO – Open‑source AI Visibility & GEO Research Tool

What it is – A self‑hosted web app that lets you query large‑language‑model (LLM) providers (via OpenRouter, OpenAI‑compatible endpoints, etc.) about a specific domain (your product or brand). It records the raw model answer, the citations the model returns, and extracts the brand mentions, competitor names, and associated keywords. You can repeat the test, schedule keyword‑monitoring runs, and compare snapshots to see how AI’s description of your product evolves.

---

Core capabilities

Feature	What you get
Domain‑level testing	Enter a domain (e.g., notion.so). Choose one or many LLMs and optionally enable web search. Each model answers independently; results are stored with the exact prompt, model, and search settings.
Evidence‑first view	For every answer you can open the original text, see the exact citations (URLs or provider‑specific references) and the locations in the answer where they were used.
Competitor & keyword extraction	The UI lists the brand name the model used, a short description, competitor products it mentions, and the keywords it associates with the brand.
Repeatable runs & history	Every test is saved as a run. You can rerun the same configuration later or schedule runs (hourly/daily/weekly) with the built‑in monitoring worker. Historical runs are shown side‑by‑side with a Diff view that highlights added/removed competitors, keyword rank changes, and citation count changes.
Keyword monitoring module (v0.3, beta)	Define custom keyword lists, import batches, classify prompts (Discovery, Comparison, etc.), and run them on a schedule. Results are stored independently and can be diffed across snapshots.
Multi‑provider model support	Connect to OpenRouter, direct provider APIs, or any OpenAI‑compatible endpoint. Mix providers in a single test; the UI shows which endpoint produced each answer.
Self‑hosted & data‑ownership	Runs on your own server or Docker container. All prompts, answers, and citations stay in your environment; only the external model APIs see the prompts you send.
Export & backup	Projects, runs, and configuration can be backed up and restored; Docker images are published as packages.

---

Who might use it

• Product & marketing teams that want to know how LLMs describe their brand and which competitors appear in AI‑generated answers.
• UX/researchers investigating the impact of content changes on AI perception (e.g., after a website rewrite).
• Start‑ups looking for a free alternative to commercial AI‑visibility dashboards.
• Developers building internal dashboards for brand monitoring or for compliance audits of AI‑generated content.

---

Tech stack & deployment

• Language: TypeScript / Node.js (requires Node 22.13+).
• Web framework: Express‑style server with a React‑based front‑end (served on http://localhost:8787).
• Containerisation: Official Docker image and docker compose setup (see docs/deployment/docker.md).
• Persistence: Uses a lightweight file‑based store (JSON) by default; can be swapped for a database in custom deployments.
• License: Apache‑2.0 (permissive, commercial‑friendly).

---

Quick start (3‑minute audit)

# clone the v0.3.0 tag
git clone --branch v0.3.0 --depth 1 https://github.com/Albert-Weasker/niubigeo.git
cd niubigeo
npm ci                     # install dependencies
cp .env.example .env       # copy env template
# add your OpenRouter API key
echo "OPENROUTER_API_KEY=your_key" >> .env
npm run server            # start the web UI

Open http://localhost:8787 in a browser, create a project, enter a domain, pick models, and hit Run.

---

Documentation & community

• Full docs – docs/ folder (model connections, deployment, monitoring, methodology, known issues).
• Examples – 20 real‑world cases (Notion, Figma, PostHog, etc.) with screenshots and raw answer data.
• Website – https://niubigeo.ai/ (product page, video advisor, promotion platform).
• Release notes – https://github.com/Albert-Weasker/niubigeo/releases (v0.3 adds keyword monitoring).
• Package registry – Docker images available via GitHub Packages.

---

Why it matters

Commercial AI‑visibility SaaS tools often present a single opaque “GEO score”. NiubiGEO flips that model: you get the evidence—the exact model answer and its sources—so you can verify, reproduce, and act on the findings. Because it’s open‑source and self‑hosted, you retain control over sensitive brand data while still being able to query any LLM you have access to.

---

Bottom line – NiubiGEO is a practical, evidence‑driven platform for anyone who needs to monitor how AI models talk about their product, track competitor mentions, and keep a reproducible history of those observations, all without paying for a proprietary dashboard.