🏗️ CiviAssist – IS Code RAG System (FAISS + Streamlit)

🚀 Overview

CiviAssist is an AI-powered Retrieval-Augmented Generation (RAG) system that recommends relevant Indian Standard (IS) codes based on user queries.

It uses FAISS-based semantic search to retrieve the most relevant standards from a structured dataset and displays them via a Streamlit interface.

---

🧠 Key Features

• 🔍 Semantic search using FAISS
• 🧠 RAG-based retrieval pipeline
• ⚡ Fast vector similarity search
• 📚 Domain-specific IS standards dataset
• 🖥️ Interactive Streamlit UI

---

📂 Project Structure

CiviAssist/
│
├── data/
│   └── bis_standards.json
│
├── venv/
│
├── app.py
├── build_index.py
├── rag_engine.py
├── faiss_index.bin
├── metadata.pkl
├── requirements.txt

---

⚙️ Complete Setup Guide

🔹 1. Clone Repository

git clone https://github.com/Suvanwita/civiassist.git
cd BIS-RAG

---

🔹 2. Create Virtual Environment

▶️ Linux / Mac:

python3 -m venv venv
source venv/bin/activate

▶️ Windows:

python -m venv venv
venv\Scripts\activate

---

🔹 3. Upgrade pip

pip install --upgrade pip

---

🔹 4. Install Dependencies

pip install -r requirements.txt

---

🔹 5. Build FAISS Index

python build_index.py

This generates:

• faiss_index.bin
• metadata.pkl

---

🔹 6. Test Retrieval Engine (CLI Mode)

Run the RAG engine directly:

python rag_engine.py

👉 This allows you to:

• Enter queries in terminal
• See retrieved IS codes
• Debug retrieval logic

---

🔹 7. Run Streamlit App

streamlit run app.py

---

🔹 8. Open in Browser

http://localhost:8501

---

🧪 Example Queries

• IS code for cement
• Standard for reinforced concrete
• Concrete mix design guidelines
• IS code for steel reinforcement
• Aggregates standard in construction

---

⚡ How It Works

🔹 Indexing (build_index.py)

• Loads dataset
• Converts text → embeddings
• Stores vectors in FAISS

🔹 Retrieval (rag_engine.py)

• Converts query → embedding
• Searches FAISS index
• Returns top matches

🔹 UI (app.py)

• Takes user input
• Displays ranked IS codes

---

📊 Sample Output

[
  {
    "id": "IS 456",
    "title": "Plain and Reinforced Concrete",
    "score": 0.92
  }
]

---

🔥 Future Enhancements

• 🤖 Add LLM for full RAG (answer generation)
• 📊 Visualization of similarity scores
• 🌐 FastAPI backend
• 🧠 Hybrid search (keyword + vector)
• 📄 Full document retrieval

---

⚠️ Limitations

• Dataset is limited
• Retrieval-only (no generative explanation yet)
• Performance depends on embeddings quality

---

🤝 Contributing

1. Fork the repo
2. Create branch
3. Commit changes
4. Open PR

---

📄 License

MIT License

---

💼 Author

CiviAssist is built to simplify IS code discovery using AI-driven semantic search for civil engineering applications.

---

⭐ Support

If you found this useful, consider giving a ⭐ on GitHub!