Python
FastAPI
React
LangGraph
LangChain
Deep Thinking Trading System is a collaborative multi-agent AI framework for financial analysis, inspired by the research paper "TradingAgents: Multi-Agents LLM Financial Trading Framework". Traditional algorithmic trading systems rely solely on numeric indicators, often missing semantic context like news, stock fundamentals, and macroeconomic events. This system orchestrates specialized agents—such as Market Data, News, Social Sentiment, Fundamentals, Research, Trading, and Risk Management—to perform comprehensive semantic reasoning before suggesting simulated trades. Powered by NVIDIA-hosted LLMs and LangGraph, it features an advanced adversarial debate workflow (such as Bull vs. Bear) to stress-test investment ideas before a Portfolio Manager makes final decisions.
Features
- •Multi-Agent Architecture: Structured workflow with separate agents for Market Data, News, Social Sentiment, Fundamentals, Research, Trading, and Risk Management.
- •Adversarial Debates: Runs Bull vs. Bear and Risk Management debates to stress-test investment proposals.
- •Live Data Integration: Incorporates real-time feeds from Yahoo Finance, Finnhub, Tavily Search, Alpha Vantage, and Financial Datasets.
- •NVIDIA-Powered LLMs: Leverages NVIDIA-hosted LLMs via an OpenAI-compatible API for deep reasoning and processing.
- •Premium UI Dashboard: A dark-themed, glassmorphic React dashboard visualizing the agent's real-time thought processes.