🤖 Labguru Chatbot

A command-line assistant to query structured data from Labguru, using a combination of LLM-based natural language parsing and structured data retrieval.

🔍 What It Does

This chatbot allows users to:

• Count or list cases, specimens, organoids, xenografts, assays, and various omics data (WES, WGS, RNA-seq, ATAC, etc.)
• Use natural language queries like:
  • how many specimens
  • list organoids starting with CSC
  • how many WES
  • list patients that name starts with BPTO
• Parse queries using LLaMA 3 for intent and filter extraction, then execute structured lookups via Python backend.

🛠️ Project Structure

.
├── chatbot/
│   ├── __init__.py
│   ├── labguru_api.py       # Core functions to retrieve structured Labguru data
│   ├── nlu_llm.py           # LLM-based intent + filter parsing (via LLaMA3)
├── interface/
│   ├── cli.py               # CLI interface for the chatbot
├── pixi.toml                # Pixi environment and dependencies
├── README.md

🚀 How to Run

1. Install Pixi (if not already)

   See: Pixi installation doc

2. Run the chatbot

pixi run chat

3. Try some example prompts:

how many specimens
list organoids name starts with CSC
list patients that name contains PHLC

🧠 Powered By

• LLaMA3 — for lightweight intent parsing
• Structured retrieval — deterministic querying of structured Labguru metadata
• Pixi — reproducible Python environment

📦 Dependencies

All dependencies are managed via pixi.toml. Notable one include:

• python-dotenv (database API configuration)

🧪 Example Intent JSON

A typical parsed LLM response looks like:

{
  "intent": "list_organoids",
  "filters": {
    "organoids_name_startswith": "CSC"
  }
}

✅ Features

• ✅ Count and list commands for structured entities
• ✅ Case-insensitive and typo-tolerant alias resolution (organlids → organoids)
• ✅ Natural language understanding using LLMs
• ✅ CLI interface for offline use

🔮 Roadmap

• integrate Q & A base LLM agent (langchain)
• integrate Knowledge graph based LLM agent (langchain)
• Add RAG-based querying for unstructured text (path reports, notes)
• Web UI
• Authentication and multi-user support

👤 Author

Guanqiao Feng

📄 License

MIT License