Skip to content

README.md

Latest commit

 

History

History
147 lines (103 loc) · 2.3 KB

README.md

File metadata and controls

147 lines (103 loc) · 2.3 KB

Enterprise LangChain Financial Research Assistant

Production-style LangChain project demonstrating a financial research assistant that analyzes US stock market data using:

  • LangChain
  • yfinance market data
  • technical indicators (SMA / RSI)
  • LLM reasoning
  • FastAPI API layer
  • C4 architecture diagrams

Author: Satish Gopinathan – The Pragmatic Architect

Features

  • Real US stock market data retrieval
  • Technical indicator calculation
  • LLM financial analysis
  • Agent-based tool orchestration
  • Health endpoint for Yahoo Finance + OpenAI connectivity checks
  • Graceful API error handling with clear HTTP status codes
  • Enterprise repo structure
  • C4 architecture documentation

Quick Start

1) Create & activate a virtual environment (recommended)

PowerShell (Windows)

python -m venv .venv
.\.venv\Scripts\Activate.ps1

Bash / WSL / macOS

python -m venv .venv
source .venv/bin/activate

2) Install dependencies

pip install -r requirements.txt

3) Set your OpenAI API key (required for /analyze/{ticker})

Recommended: create a .env file in the repo root:

OPENAI_API_KEY=sk-your-real-key-here

PowerShell

$env:OPENAI_API_KEY = "<your-key-here>"

Bash / WSL / macOS

export OPENAI_API_KEY="<your-key-here>"

4) Run API

uvicorn api.main:app --reload --app-dir src

Alternative (without --app-dir):

uvicorn src.api.main:app --reload

5) Open API docs

http://127.0.0.1:8000/docs

API Endpoints

Analyze a ticker

GET /analyze/{ticker}

Example:

http://127.0.0.1:8000/analyze/AAPL

Health check

GET /health

Example:

http://127.0.0.1:8000/health

Response shape:

{
  "status": "ok | degraded",
  "timestamp_utc": "ISO-8601 timestamp",
  "checks": {
    "yahoo": {
      "ok": true,
      "error": "optional error message"
    },
    "openai": {
      "ok": true,
      "error": "optional error message"
    }
  }
}

Expected for a healthy setup:

  • checks.yahoo.ok = true
  • checks.openai.ok = true

Error Handling

  • /analyze/{ticker} returns 400 for invalid ticker/data errors.
  • /analyze/{ticker} returns 503 for upstream dependency/config issues.
  • If OPENAI_API_KEY is missing, analysis calls return a clear 503 message.