🚀 Intelligent Banking Support Powered by LangGraph · LLaMA 3.1 · Groq
👉 Try the App Here
🔗 Frontend UI (Streamlit): https://banking-agent-phmivszmcasyzw5fdv3ea3.streamlit.app/
🔗 GitHub Repository: https://github.com/hari9618/Banking-Agent
SecureBank AI Agent is an intelligent, LangGraph-powered autonomous banking assistant designed to automate customer support entirely.
Traditional banking chatbots follow rigid rule trees that break the moment a user asks something unexpected. SecureBank AI Agent solves this with a hybrid architecture — combining fast rule-based routing for known queries (balance, transactions, FAQs) with full LLM reasoning via Groq's LLaMA 3.1 for complex, open-ended questions.
The result is a banking assistant that is fast, accurate, and capable of handling real-world conversations — all within a sleek, interactive Streamlit chat interface.
| Feature | Description |
|---|---|
| 🧠 Intent Classification | Automatically detects balance, transaction, FAQ, or general queries |
| 🔀 Intelligent Routing | LangGraph routes each query to the right handler node |
| 🤖 LLM Reasoning | Groq + LLaMA 3.1 handles complex, open-ended questions |
| 📋 Rule-Based FAQ Engine | Instant keyword-matched answers for common banking questions |
| 💰 Account Balance Retrieval | Displays live-formatted balance cards per customer |
| 📊 Transaction History | Generates and displays recent transaction records |
| 🛡️ Error Handling | Handles rate limits, API errors, and model issues gracefully |
| 🔑 Secure Key Management | API key loaded via .env — never hardcoded |
| 🎨 Interactive Chat UI | Streamlit chat interface with quick-action sidebar buttons |
| ⚡ Rate Limit Protection | Built-in throttling and automatic retry logic |
| Technology | Purpose |
|---|---|
| 🐍 Python 3.10+ | Core programming language |
| 🎨 Streamlit | Interactive frontend chat UI |
| 🔗 LangGraph | Stateful AI workflow & node routing engine |
| 🦜 LangChain | LLM integration and message formatting |
| ⚡ Groq API | Ultra-fast LLM inference |
| 🧠 LLaMA 3.1 (8B Instant) | Language model for intelligent responses |
| 🔐 python-dotenv | Secure environment variable management |
SecureBank-AI-Agent/
│
├── app.py # Streamlit Frontend + LangGraph workflow
├── .env # API keys (not committed)
├── requirements.txt # Python dependencies
└── README.md
│
│ ── Internal Architecture ──
│
├── 🎨 UI Layer # Streamlit chat interface + sidebar quick actions
├── 🔀 LangGraph Workflow # StateGraph with classify → route → handle nodes
├── 🤖 LLM Layer # ChatGroq (LLaMA 3.1) via LangChain
└── 🗄️ Mock Database Layer # Customer profiles + transaction generation
User Input
│
▼
[classify node] ──► intent detected
│
▼
[route function]
│
├──► "balance" ──► [handle_balance] ──► Balance Card UI
├──► "txn" ──► [handle_txn] ──► Transaction Table UI
├──► "faq" ──► [handle_faq] ──► FAQ Answer UI
└──► "other" ──► [handle_other] ──► LLM (Groq + LLaMA 3.1)
│
▼
AI Response UI
git clone https://github.com/your-username/securebank-ai-agent.git
cd securebank-ai-agentpip install -r requirements.txtOr manually:
pip install streamlit langgraph langchain langchain-core langchain-groq python-dotenvCreate a .env file in the project root:
GROQ_API_KEY=your_groq_api_key_here🔑 Get your free Groq API key at console.groq.com
streamlit run app.py
⚠️ Always run from the folder containing your.envfile.
1️⃣ User types a message or clicks a Quick Action in the sidebar
│
2️⃣ LangGraph StateGraph receives the input
│
3️⃣ classify node detects intent:
├── "balance" → contains keywords like balance, funds, how much
├── "txn" → contains keywords like transaction, history, recent
├── "faq" → matches FAQ keyword map (rates, loans, KYC, etc.)
└── "other" → everything else
│
4️⃣ route function directs to the correct handler node
│
5️⃣ Handler executes:
├── Rule-based → instant response (balance / txn / faq)
└── LLM-based → Groq API call → LLaMA 3.1 generates response
│
6️⃣ Formatted response rendered in Streamlit chat UI
<img width="957" height="398" alt="LG" src="https://github.com/user-attachments/assets/4dc52af6-8609-45af-bfd9-e42f8d58709f" />
✔ Building stateful AI agents with LangGraph StateGraph
✔ Designing hybrid rule-based + LLM systems for reliability
✔ Integrating Groq API with LangChain for fast LLM inference
✔ Implementing intent classification and conditional routing
✔ Building production-grade error handling (rate limits, model deprecation)
✔ Deploying interactive AI apps with Streamlit
✔ Managing LLM client caching with @st.cache_resource
🔹 Real banking API integration (Plaid / RazorpayX)
🔹 User authentication & multi-account support
🔹 Voice input support
🔹 RAG-based document Q&A (loan agreements, terms)
🔹 SHAP-based explainability for model decisions
🔹 Multilingual support (Hindi, Telugu, etc.)
🔹 Persistent chat history with vector database
🔹 Fraud detection alert system
Special thanks to everyone who supported this project:
🏫 Innomatics Research Labs — For providing a strong AI learning environment
👨🏫 Manohar Chary V. Sir — For guidance and continuous mentorship
And a heartfelt thank you to:
Raghu Ram Aduri Sir · Kanav Bansal Sir · Vishwanath Nyathani Sir · Kalpana Katiki Reddy Ma'am
Hari Krishna
AI Enthusiast | Gen AI Engineer | AI Builder
If you found this project useful:
⭐ Star the repository
📢 Share with your network
🐛 Report issues or suggest features via GitHub Issues
AI LangGraph LangChain Groq LLaMA Banking Assistant Streamlit Python Generative AI AI Agents Chatbot NLP Machine Learning Autonomous Agent FastAPI