🍵 CHAGEEPT

A RAG-powered AI chatbot for CHAGEE Philippines. Ask questions about the menu, store locations, and more!

What is this?

CHAGEEPT is a personal fun project - a chatbot that knows everything about CHAGEE Philippines. It uses:

• RAG (Retrieval-Augmented Generation) to answer questions based on scraped website data
• Qdrant Cloud for vector storage and semantic search
• Ollama Cloud LLM (via Ollama's OpenAI-compatible API) for natural language responses
• Gemini embeddings (via Google AI Studio's OpenAI-compatible API) for vector search
• Agentic tool-calling loop - the LLM plans across knowledge-base search, on-demand scraping, and a Tavily web-search fallback (CHAGEE-only, used only after the knowledge base comes up short)
• Chainlit for a ChatGPT-style web interface

Features

• 🍵 Browse the full CHAGEE menu (Fresh Milk Tea, Brewed Tea, Frappés, etc.)
• 📍 Find store locations in the Philippines
• ℹ️ Learn about CHAGEE's history and brand
• 🔄 Auto-updates database on startup and daily at 3 AM

Setup

1. Clone and install dependencies

   git clone <your-repo>
   cd chageept
   pip install -r requirements.txt

2. Set up environment variables

   cp .env.example .env
   # Edit .env and add your GEMINI_API_KEY (embeddings), OLLAMA_API_KEY (chat),
   # and QDRANT_URL/QDRANT_API_KEY (vector storage)
   # Optionally add TAVILY_API_KEY to enable web-search fallback

3. Build the knowledge base

   python -m scripts.seed_crawler

4. Run the chatbot

   chainlit run chat_ui.py

5. Open http://localhost:8000 in your browser 🎉

Project Structure

chageept/
├── chat_ui.py           # Main Chainlit app
├── chageept/
│   ├── agent.py         # Tool-calling planning loop (search/scrape/web_search/answer)
│   ├── scraper.py       # Web scraper for CHAGEE website
│   ├── retriever.py     # Qdrant vector search
│   ├── llm.py           # Ollama Cloud LLM integration
│   ├── websearch.py     # Tavily web search fallback
│   └── tools.py         # Data models
├── scripts/
│   └── seed_crawler.py  # Database builder
└── public/              # UI assets (logo, favicon)

Tech Stack

Component	Technology
UI	Chainlit
Vector DB	Qdrant Cloud
Embeddings	gemini-embedding-001 (via Google AI Studio)
LLM	gemma4 (via Ollama Cloud)
Scraping	BeautifulSoup4

Deployment

Configured for Railway deployment with Docker. Just push to your repo and connect to Railway.

---

This is a personal project and is not affiliated with CHAGEE.