Overview
The Multi-Agent Financial Assistant is a web-based AI system built with the Phi multi-agent framework and Groq's ultra-fast LLaMA3. It handles real-time financial queries and intelligent web search by delegating tasks between specialized agents (Finance Agent and Web Search Agent). Integrated with yfinance for stock data and Google Search for live information, it delivers results in a Markdown-rendered chat UI with light/dark mode support.
Pipeline
1. User submits financial or search query via chat UI
2. Task delegation between Finance Agent (yfinance) and Web Search Agent (Google)
3. LLaMA3 model via Groq processes query for reasoning & synthesis
4. Responses formatted in Markdown (tables, code, citations)
5. Chat UI renders response with persistent light/dark mode toggle
Architecture & Design
The backend runs on Flask, orchestrating multi-agent communication via Phi. Finance queries use yfinance for real-time stock market data, while general knowledge queries trigger a Google Search agent. Responses are processed through Groq's LLaMA3-8B for low-latency inference and delivered to the frontend as Markdown-rendered messages. The system features a responsive UI with theme persistence and chat history support.
Features
• Multi-agent task delegation (Finance + Web Search)
• Groq-powered LLaMA3 integration
• Real-time stock data and fundamentals via yfinance
• Web search with citations
• Markdown chat UI with tables and code blocks
• Light/dark mode toggle with persistence
• Flask-based responsive web app
Tech. Stack
Python, Flask, Phi Multi-Agent Framework, Groq, yfinance, Google Search API, JavaScript, HTML/CSS, Marked.js