Build AI Agents for US and Korean stock investing from scratch. On an Ollama-based local setup that runs on CPU alone, without CUDA, implement ReAct, RAG, Tool Use, and Multi-Agent step by step.
Development environment for this series
Combines the quantitative screener scores in the AutoTrading SQLite database with yfinance indicators to rank the top stocks. Click AI Analysis on any stock to watch the agent reason and hand the analysis straight to Antigravity.
Install Ollama + Llama 3.1 8B locally and set up LangChain, FastAPI, yfinance, and pykrx. Runs right away on a CPU-only laptop.
A ReAct agent that looks up real-time prices, indexes, and FX rates for the US (S&P 500, NASDAQ) and Korea (KOSPI, KOSDAQ). A full implementation of the Tool Use pattern.
A RAG agent that collects and embeds Korean DART filings and US SEC documents to analyze financial statements and earnings releases in natural language.
A Tool Use agent that computes RSI, MACD, Bollinger Bands, and moving averages and generates buy/sell signals. Includes backtesting code.
Three agents โ Researcher โ Analyst โ Risk Manager โ collaborate through LangGraph to produce a portfolio rebalancing plan.