🏦 Fintria

Welcome to the Fintria AI project — an intelligent assistant built with Streamlit, FastAPI (REST), gRPC, WebSocket, and Langchain, designed to provide banking services like balance inquiries, loan checks, transaction history, and more through a conversational interface.

This chatbot is engineered with workflow principles for reliable, modular, and scalable interactions, mimicking real-world banking operations.

---

✨ Features

• Streamlit Web UI for live interaction.
• REST API integration with FastAPI.
• gRPC API for high-performance communication.
• WebSocket API for real-time messaging.
• Document Loading with custom SimpleDirectoryLoader.
• LLM (Large Language Model) powered by Ollama LLM (using models like Mistral).
• Workflow Engine integration for:
  • Account balance inquiry
  • Loan eligibility check
  • Transaction history retrieval
  • Payments/Transfers
  • Human agent escalation
• Session State Management.
• Backend Service Abstraction (easy to plug in real databases).
• Easy extensibility for future banking use cases.

---

📚 Architecture

flowchart TD
    A[User Input] --> B["Streamlit Frontend / REST / gRPC / WebSocket"]
    B --> C[Chatbot Service]
    C --> D[Document Loader]
    C --> E[LLM Client]
    C --> F[Workflow Manager]
    F --> G[Banking Backend Services]
    G --> H[Bank Database / External APIs]

Loading

• Presentation Layer: Streamlit frontend and APIs.
• Service Layer: Business logic and workflows.
• Model Layer: LLM integration for smart responses.
• Persistence Layer: (Future) Database for users, transactions, etc.

---

🚀 Setup Instructions

1. Clone the Repository

git clone https://github.com/yourusername/banking-assistant-ai.git
cd banking-assistant-ai

2. Install Requirements

pip install -r requirements.txt

Make sure you have Python 3.11+ installed.

---

3. Running Servers

Server Type	Command	Description
Streamlit UI	streamlit run src/app/streamlit_app.py	Interactive chatbot frontend
REST API	python src/app/rest_api.py	FastAPI-based REST server
gRPC Server	python src/app/grpc_server.py	gRPC server for efficient communication
WebSocket Server	python src/app/websocket_server.py	Real-time WebSocket server

---

4. Folder Structure

src/
  app/
    streamlit_app.py         # Streamlit Frontend
    rest_api.py               # REST API
    grpc_server.py            # gRPC Server
    websocket_server.py       # WebSocket Server
    proto/                    # Protobuf definitions
  core/
    services/
      simple_directory_loader.py
      ollama_llm_client.py
      banking_chatbot_service.py
      backend_service.py      # Abstract banking operations
tests/
  functional/
    test_rest_api.py
    test_grpc_api.py
    test_websocket_api.py
docs/                         # Preloaded documents for chatbot

---

🛠️ Workflows Overview

Workflow	Description	Entry Point
Account Balance Inquiry	Get your account balance securely	"check balance"
Loan Eligibility Check	See if you're eligible for loans	"check loan eligibility"
Transaction History Retrieval	View recent transactions	"view transactions"
Payments/Transfers	Send money securely	"make payment"
Escalate to Agent	Talk to a human support agent	"escalate issue"

---

📦 Example API Requests

REST Example (using curl)

curl -X POST http://localhost:8000/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "What is my account balance?"}'

---

gRPC Example (Python Client)

import grpc
import src.app.proto.chat_pb2 as chat_pb2
import src.app.proto.chat_pb2_grpc as chat_pb2_grpc

channel = grpc.insecure_channel('localhost:50051')
stub = chat_pb2_grpc.ChatServiceStub(channel)
request = chat_pb2.ChatRequest(user_message="Tell me about loans")
response = stub.Chat(request)
print(response.bot_message)

---

WebSocket Example (Python)

import websocket
import json

def on_message(ws, message):
    print("Received:", message)

ws = websocket.WebSocketApp("ws://localhost:8001/chat", on_message=on_message)
ws.run_forever()

---

🧠 Future Improvements

• Integrate with real banking backend (databases/APIs)
• Add user authentication (OAuth, JWT)
• Fine-tune LLM for banking-specific knowledge
• Rich chat UI with Streamlit Components
• Alert systems for suspicious activities

---

📽️ Live Demo

🎥 Watch Demo Video

💡 Smart Conversations | 🏦 Banking Knowledge | 🔥 Built with LLMs, Workflows, and FinTech APIs

📜 License

This project is licensed under the MIT License.
Feel free to modify and extend it!

---

🤝 Contributions

Contributions are welcome!
Please open issues and pull requests to improve workflows, models, or APIs.

---