🏥 RAG Clinic Assistant

A production-style Retrieval-Augmented Generation (RAG) chatbot that answers questions about clinic documents. Built with LlamaIndex, ChromaDB, OpenAI, and served via a FastAPI backend with a clean, framework-free HTML/CSS/JS frontend.

---

✨ Features

• 💬 ChatGPT-style dark web UI — no React, no Vue, pure HTML/CSS/JS
• 🔍 Semantic search over clinic documents using vector embeddings
• 🤖 Answers grounded in your documents via gpt-4o-mini
• 💾 Automatic per-session chat backup to chatBackup/ as JSON
• ⚡ FastAPI backend with instant startup — index loaded once at boot
• 📱 Responsive layout — works on mobile and desktop
• 🗂️ Sidebar with document list and "New Chat" button

---

🧠 How It Works

This project implements the standard RAG pipeline in two phases:

Offline — Indexing (run once)

docs/ (your .txt / .pdf files)
       │
       ▼
[ingest.py]
  → split into 512-token chunks
  → embed with OpenAI text-embedding-ada-002
  → store vectors in ./chroma_db

Online — Query (every request)

User question
      │
      ▼
[Embed question] → vector
      │
      ▼
[ChromaDB] → top 3 semantically similar chunks
      │
      ▼
[gpt-4o-mini] ← chunks + question as context
      │
      ▼
  🤖 Grounded answer

---

📁 Project Structure

RAG-ChatBot/
├── docs/                              # Source documents (clinic knowledge base)
│   ├── blood_tests_reference.txt
│   ├── clinic_faq.txt
│   ├── colonoscopy_prep.txt
│   ├── emergency_escalation_rules.txt
│   ├── insurance_guide.txt
│   ├── patient_rights.txt
│   ├── prescription_refill_policy.txt
│   └── telehealth_guide.txt
├── static/
│   └── index.html                     # Full chat UI (single file, no frameworks)
├── chroma_db/                         # Auto-generated vector DB (git-ignored)
├── chatBackup/                        # Per-session chat logs as JSON (git-ignored)
├── ingest.py                          # Step 1: Chunk, embed, and index documents
├── query.py                           # Step 2: Terminal Q&A interface
├── api.py                             # Step 3: FastAPI backend
├── .env                               # API keys (git-ignored)
├── .gitignore
└── README.md

---

⚙️ Setup

1. Clone the Repository

git clone https://github.com/eziyoo/RAG-ChatBot.git
cd RAG-ChatBot

2. Create a Virtual Environment

python3 -m venv .venv
source .venv/bin/activate        # macOS / Linux
.venv\Scripts\activate           # Windows

3. Install Dependencies

pip install -r requirements.txt

4. Configure Your API Key

Create a .env file in the project root:

OPENAI_API_KEY=your_openai_api_key_here

Get your key from platform.openai.com/api-keys. Make sure billing is enabled — costs are minimal (fractions of a cent per query).

---

🚀 Usage

Step 1 — Ingest Documents

Run once to build the vector database from your documents:

python ingest.py

This reads all files from docs/, chunks them into 512-token pieces, embeds them with OpenAI, and stores everything in ./chroma_db.

Re-run this any time you add, remove, or update documents.

Step 2 — Start the Web App

uvicorn api:app --reload

Open your browser at http://localhost:8000.

Optional — Terminal Interface

For quick testing without the web UI:

python query.py

🙋 Your question: Do I need someone to drive me home after a colonoscopy?
🤖 Answer: Yes, you must arrange for a responsible adult to drive you home...

Type exit or quit to stop.

---

🛠️ Tech Stack

Layer	Technology
RAG Framework	LlamaIndex
Vector Database	ChromaDB — local, no cloud needed
Embedding Model	OpenAI text-embedding-ada-002
LLM	OpenAI gpt-4o-mini
Backend	FastAPI + Uvicorn
Frontend	Vanilla HTML / CSS / JavaScript (no frameworks)
Environment	Python 3.10+

---

📦 Dependencies

Install all at once:

pip install -r requirements.txt

Key packages:

llama-index
llama-index-vector-stores-chroma
llama-index-llms-openai
chromadb
fastapi
uvicorn
python-multipart
openai
python-dotenv

Generate a fresh lockfile after adding packages:

pip freeze > requirements.txt

---

💾 Chat Backup

Every conversation is automatically saved to chatBackup/ as a JSON file. One file is created per session — it is overwritten on each message (not duplicated).

chatBackup/
├── chat_2026-06-24_18-42-00.json   ← Session 1
├── chat_2026-06-24_20-15-00.json   ← Session 2 (New Chat clicked)
└── ...

Each file contains the full message history:

{
  "saved_at": "2026-06-24_18-43-15",
  "session_id": "2026-06-24_18-42-00",
  "messages": [
    { "role": "user",      "content": "What are my rights as a patient?" },
    { "role": "assistant", "content": "As a patient, you have the right to..." }
  ]
}

---

📄 Adding New Documents

1. Add .txt, .pdf, or .docx files to the docs/ folder
2. Delete the existing chroma_db/ folder to avoid stale data:

   rm -rf chroma_db/

3. Re-run the ingestion pipeline:

   python ingest.py

---

🔒 Environment Variables

Variable	Description
OPENAI_API_KEY	Your OpenAI API key

The .env file is listed in .gitignore and will never be committed to version control.

---

🤝 Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

---

📝 License

This project is licensed under the MIT License.