prashant9501/meridian-wealth

Meridian Wealth Partners is a Python web application that acts as an AI-powered financial analyst assistant for a fictional wealth management company. The agent answers questions about client portfolios by pulling information from multiple sources automatically, rather than requiring a person to look up each piece of data manually. When a relationship manager types a question into the chat interface, the AI agent decides which tools to use: it can query a SQLite database containing client holdings, risk profiles, and market data, search a collection of firm policy PDFs to check compliance rules, or look up live market news through an external web search service called Tavily. The agent reasons through what it needs, runs the relevant tools, and composes a sourced answer. A live trace panel in the chat interface shows exactly which tools were called and in what order. The policy document search is powered by a technique called retrieval-augmented generation (RAG). On first startup, the app reads PDF files from a folder, converts them into searchable numerical representations using OpenAI's API, and stores those in a local index called FAISS. Subsequent lookups search the index rather than re-reading the PDFs each time. The project is built with FastAPI on the backend and uses Jinja2 HTML templates for the frontend. WebSockets handle streaming chat responses. An AWS EC2 deployment guide using Nginx and systemd is included for hosting the app on a public server behind a reverse proxy. Setup requires Python 3.12, an OpenAI API key, and a Tavily API key for the web search feature. The readme includes step-by-step Windows instructions for creating a virtual environment and running the server locally.