FinTrack AI is a modern financial dashboard that combines a sleek UI with a high-performance AI-Powered Chatbox. Instead of relying on static data, this platform uses an agentic workflow to research the live web, providing users with real-time financial insights and market analysis.
The heart of this project is the Agentic Chat Assistant. Unlike standard chatbots, this assistant doesn't just "chat"—it researches. It is powered by a custom RAG (Retrieval-Augmented Generation) pipeline.
When you ask a question like "What is the current trend for Nvidia stock?" or "Should I invest in Gold today?", the AI follows a 3-step process:
- Web Retrieval: The system identifies the need for real-time data and triggers the DuckDuckGo Search API to scan the latest financial news and market reports.
- Context Augmentation: The top search results are parsed and injected into the AI's "short-term memory" as factual context.
- Llama 3.1 Reasoning: Using the Meta Llama-3.1-8B-Instruct model (via Hugging Face), the agent synthesizes the live data and the user query to provide a professional, data-backed response.
- Zero Knowledge Cutoff: By using live web search, the AI is always aware of market moves that happened minutes ago.
- Hallucination-Free: Every financial insight is grounded in real-time search results, drastically reducing the "guessing" common in standard AI.
- Agentic Decision Making: The system intelligently decides when to search the web and when to use its internal reasoning.
- Backend: Python & Flask
- LLM: Llama 3.1 (8B Instruct)
- Inference: Hugging Face API
- Search Engine: DuckDuckGo Search API
- Frontend: Custom CSS/JS Fintech Dashboard
- Deployment: Render Cloud
├── app.py # Main Flask server & AI/RAG logic
├── requirements.txt # Production-ready dependencies
├── static/ # CSS and JavaScript assets
├── templates/ # HTML Structure (Dashboard & Chat Widget)
└── .gitignore # Security shield for private tokens