💼 TrueWealth AI: Your AI-powered financial strategist

demo.mp4

---

Project Overview

TrueWealth AI is a end-to-end Multi-Agent Financial Advisor AI System that combines LangGraph-powered orchestration, advanced LLM reasoning (LLaMA-3 via Groq), and RAG with ChromaDB + HuggingFace embeddings to deliver professional, context-aware financial insights. The system features Planner, Retriever, Generator, News, Web Search, and Memory agents with intelligent tool routing, retry logic, and multi-source knowledge fusion for high accuracy and reliability. It supports PDF ingestion, financial news retrieval, and fallback web search to ensure comprehensive answers. Built with a modular architecture for scalability, it provides FastAPI APIs for testing, a Flask-based responsive UI (HTML, CSS, JS) for client interaction, and is fully Dockerized for portability. Deployed on Render with an integrated CI/CD pipeline, it reflects enterprise-level engineering practices, making it a industry-grade AI solution ready for real-world deployment.

---

🚀 Live Demo

🎯 Try the real-time TrueWealth AI:
👉 TrueWealth AI – Click Here

---

Real-World Use Cases

1. Personal Financial Advisory

• Helps individual investors make informed decisions on stocks, bonds, and long-term investments.

2. Financial Education & Self-Learning

• Acts as a tutor for users reading financial books (e.g., The Intelligent Investor).

3. Real-Time Market & News Analysis

• Provides up-to-date financial news from Yahoo Finance.

4. Small Business Financial Consulting

• Assists small business owners in budgeting, tax planning, and investment decisions.

5. Retirement & Wealth Management

• Offers retirement planning insights (e.g., 401(k), Roth IRA strategies).

---

Features & Functionalities

✅ Step	🧠 Feature	⚙️ Tech Stack / Tool Used
1️⃣	🧠 LLM-based Financial Query Understanding	Groq + LLaMA-3
2️⃣	✨ Professional Tone Personalization	Prompt Engineering + Advisor Persona Templates
3️⃣	📚 RAG-based Financial Answering	LangChain + ChromaDB + Sentence Transformers (all-MiniLM-L6-v2)
4️⃣	🔍 Financial Document Retriever Agent	RetrieverAgent + Vector Store Search
5️⃣	🧠 Answer Generator Agent	GeneratorAgent (LLM-based factual + professional financial style)
6️⃣	📰 Financial News Retrieval Agent	YahooFinanceNewsTool
7️⃣	🌐 Web Search Agent (Fallback)	DuckDuckGo Search Tool
8️⃣	🧠 Planner Agent	LangGraph Planner Node
9️⃣	🔄 Intelligent Tool Routing & Fallback	Retry Logic + Conditional Branching + Multi-step Tool Selection
🔟	🧠 Short-Term Conversational Memory	LangGraph Memory Integration (Buffer-based)
1️⃣1️⃣	📂 PDF Knowledge Ingestion	PyPDFLoader + RecursiveCharacterTextSplitter
1️⃣2️⃣	📊 Vector Embedding & Storage	HuggingFaceEmbeddings + ChromaDB
1️⃣3️⃣	⚙️ State-based Multi-Agent Orchestration	LangGraph StateGraph + Conditional Edges + Dynamic State Updates
1️⃣4️⃣	🗂 Multi-source Knowledge Fusion	LLM + RAG + Yahoo Finance + DuckDuckGo Combined Answer Synthesis
1️⃣5️⃣	🧪 API Testing & Integration	FastAPI (API endpoints for agent orchestration testing)
1️⃣6️⃣	🗃 Modular Code Architecture	Separation of Concerns + Service/Agent Modules
1️⃣7️⃣	💻 Responsive Web UI	Flask + HTML5, CSS3, JavaScript
1️⃣8️⃣	🌐 Cloud Deployment	Render (Production hosting)
1️⃣9️⃣	🔄 CI/CD Pipeline	GitHub Actions / CI/CD Workflows
2️⃣0️⃣	📦 Containerization for Portability	Docker (App + Dependencies + Environment)

---

Project Structure

TrueWealth AI/
   │
   ├── .github/
   │   └── workflows/
   │       └── main.yml
   │
   ├── agents/
   │   ├── __init__.py
   │   ├── duckduckgo.py
   │   ├── executor.py
   │   ├── generator.py
   │   ├── llm.py
   │   ├── memory_store.py
   │   ├── memory.py
   │   ├── planner.py
   │   ├── rag.py
   │   └── yfinance.py
   │
   ├── core/
   │   ├── __init__.py
   │   ├── config.py
   │   ├── state.py
   │   └── workflow.py
   │
   ├── data/
   │   └── The Intelligent Investor - BENJAMIN GRAHAM.pdf
   │
   ├──finance_db/
   │   └── chroma.sqlite3
   │
   ├── notebook/
   │   └── experiment.ipynb
   │
   ├── static/
   │   ├── css/
   │   │   └── style.css
   │   ├── images/
   │   │   └── logo.png
   │   └── js/
   │       └── script.js
   │
   ├── templates/
   │   └── index.html
   │
   ├── tests/
   │   └── test_app.py
   │
   ├── tools/
   │   ├── __init__.py
   │   ├── document_loader.py
   │   ├── llm_client.py
   │   ├── search_tools.py
   │   └── vector_store.py
   │
   ├── .gitignore
   ├── api.py
   ├── app.py
   ├── app.png
   ├── demo.mp4
   ├── Dockerfile
   ├── LICENSE
   ├── main.py
   ├── README.md
   ├── render.yaml
   ├── requirements.txt
   └── setup.py

---

System Architecture

flowchart TD
    A[User Query] --> B[Planner]
    B --> C[Recall Memory]
    C --> D[LLM Direct Answer Attempt]
    D -->|Success| E[Generate Response]
    D -->|Failure| F[Executor: Retry Logic]
    F --> G{Retry Count < 3?}
    G -->|Yes| H[RAG: Search PDF Knowledge]
    G -->|No| I[Yahoo Finance Search]
    H -->|Found Docs| E
    H -->|No Docs| I
    I -->|Found News| E
    I -->|No News| J[DuckDuckGo Web Search]
    J --> E
    E --> K[Store in Memory]
    K --> L[Return Final Answer]

    %% External Tools
    H --> M[(ChromaDB)]
    I --> N[[Yahoo Finance]]
    J --> O[[DuckDuckGo]]
    
    %% LLM Core
    D --> P[[Groq-LLaMA3]]
    E --> P

Loading

---

FastAPI Endpoints

POST /chat

Process a financial question and return an AI-generated response with source information.

Request:

POST /chat HTTP/1.1  
Content-Type: application/json  
Host: localhost:8000  

{
  "message": "What are the top performing stocks this week?",
  "session_id": "optional_existing_id"
}

Parameters:

• message (required) → The financial question to process
• session_id (optional) → Existing session ID for context continuity (default: "default")

Response:

{
  "response": "Based on Yahoo Finance data, the top performing stocks this week are...",
  "session_id": "20250813123045",
  "source": "YahooFinance"
}

Status Codes:

• 200 → Successful response
• 400 → Invalid request (missing message)
• 500 → Internal server error

---

Deployment Process with Docker

Dockerization:

1. Service Containerization: The entire TrueWealth AI system is broken down into microservices, each housed within its own Docker container.
2. Docker Compose: Used to coordinate and manage multi-container services for local development and testing.

Deployment Steps:

1. Build Docker Image:

   docker build -t truewealth-ai .

2. Run Application in Docker:

   docker-compose up --build

---

CI/CD Pipeline

Continuous Integration and Continuous Deployment (CI/CD) is implemented using GitHub Actions to automate testing, building, and deploying the project.

CI/CD Features:

• Automated Testing: Every code push triggers automated unit and integration tests.
• Automated Deployment: Successful builds are automatically deployed to production or staging environments.

---

Future Enhancements

• Multilingual Support: Future iterations will include multilingual capabilities to cater to a global client base.
• Advanced Financial Analytics: Integration with more sophisticated financial forecasting models and real-time market analysis.
• Long-Term Memory: Extended memory for better long-term personalized financial advice.

---

Personal Information

• Developer: Md Emon Hasan
• GitHub: Md-Emon-Hasan
• LinkedIn: Md Emon Hasan
• Email: iconicemon01@gmail.com
• WhatsApp: +8801834363533

---