Hi everyone!
I’d like to introduce a new plugin I’ve been developing for Grav CMS: Grav AI Chatbot (grav-plugin-ai-chatbot). 100% FREE AND OPEN SOURCE.
📦 Where to Download / Repo
-
GitHub Repository: https://github.com/milkboyinchina/grav-ai-chatbot
-
Live Demo: Test the floating chatbot live on https://www.milkboy.my.id (No CDN, may be slow if you are not in the same region as me)
Hardware Warning for Live Demo:
The live demo backend onmilkboy.my.idis powered by a custom remote Ollama endpoint running on a janky old Intel Celeron N4100 low-power mini PC servingQwen2.5-0.8B! It proves the RAG engine's extreme token efficiency by running on ultra-weak hardware, but your mileage may vary—please expect slight processing latency when testing live queries! 😉
❓ What is it & What is it for?
It’s a customizable AI Chatbot widget for Grav sites designed to deliver instant, factual answers to visitors while keeping your AI API token costs close to $0.00.
✨ Key Features:
- 🧠 Retrieval-Augmented Generation (RAG): Uses a local SQLite vector database to inject only relevant section chunks into AI prompts, achieving an 80%+ token & cost reduction.
- ⚡ Multi-Provider AI Support: Works with Google Gemini (Default Free Tier), Groq (Ultra-fast Llama 3), OpenAI, OpenRouter, or local self-hosted Ollama models.
- 🎨 Custom UI Themes & Quick Replies: Includes preset themes (Glassmorphic Blue, Emerald Dark, etc.), custom accent colors, position controls, and quick reply suggestion pills.
- 📅 Date Range Analytics & Data Export: Visual interaction volume bar charts, query distribution ratios, and CSV/JSON export endpoints.
- 🛡️ Safety Guardrails & Limits: Built-in rate limiting and blacklisted vocabulary filtering.
💡 Use it WITH or WITHOUT Paid AI:
- WITHOUT AI (100% Free / Zero API Calls): Enable Local Semantic FAQ Pre-Matching and the Local TF-IDF/BM25 engine. Visitor questions are matched locally against your
/faqpages with 0 AI API tokens used. - WITH AI: Enable cloud AI fallback (Google Gemini, Groq, OpenAI) or local/remote self-hosted Ollama (tested on
Qwen2.5-0.8B,nomic-embed-text) for natural language answers.
⚙️ Known Issues Being Worked On:
- Live error logs display & inline chart metric updates on the Grav Admin config page not updating (still have not figure this out yet, might give up and ask for help soon).
🙏 Request for Testing & Feedback:
Please test it out on your Grav setups and let me know your thoughts!
- Does the RAG indexing work smoothly on your pages and themes?
- Is it polished enough for official GPM (Grav Package Manager) submission?
Looking forward to your feedback and bug reports on GitHub! Thanks!
Star this project on GitHub too!