This project is a financial agent that given a ticker, validates it, fetches news, performs sentiment analysis, and generates a financial report. It provides human-in-the-loop feedback to validate the results, and has a 24h cache so that it can reuse previously fetched data. The agent is powered by LangGraph and its responses can be traced with Langfuse. Is also provides a simple user interface. The project uses yfinance to fetch financial data and Tavily to fetch news. It uses Ollama as the local LLM to generate the reports.
The project uses the following dependencies:
- ty: very fast type checking
- fastapi: web framework for the server mode
- uv: dependency management
- yfinance: fetch financial data
- tavily: fetch news
- ruff: very fast linting
- pytest: testing framework
- pydantic: data validation
- langgraph: graph-based agent framework
- streamlit: web UI