AI Investment Advisor Agent with Portfolio Simulation & Risk Guardrails

An agentic AI investment advisor that combines LangGraph tool orchestration, RAG over your portfolio and documents, scenario simulation, and compliance guardrails — built for the 2025–2026 fintech/wealth-management stack.

What It Does

• LLM agent (LangGraph): ReAct-style agent that chooses when to call tools and how to answer.
• Multi-tool integration:
  • yfinance: Real-time and historical prices, company info, holding values.
  • Math: Sharpe ratio, portfolio metrics, CAGR.
  • Scenario simulation: Monte Carlo–style portfolio projection (e.g. AAPL:0.5,MSFT:0.5 over 5 years).
  • News/sentiment: Optional Alpha Vantage or News API for headlines/sentiment.
• RAG: Vector store (Chroma + HuggingFace embeddings) over uploaded portfolio CSV and optional user documents.
• Conversation memory: Optional RAG over recent conversation for context.
• Risk guardrails: Compliance checks (no specific tax/legal advice, no guarantees), concentration flags, and automatic disclaimer when needed.
• Trajectory view: See which tools the agent called (debugging and explainability).
• Streamlit UI: Upload portfolio, chat, and inspect agent steps.

Setup

1. Clone and enter the project

   cd "AI Investment Advisor Agent with Portfolio Simulation & Risk Guardrails"

2. Create a virtual environment

   python -m venv .venv
   source .venv/bin/activate   # macOS/Linux
   # Windows: .venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

4. Configure environment

   • Copy .env.example to .env.
   • Set GROQ_API_KEY (get one at console.groq.com).
   • Optionally set NEWS_API_KEY or ALPHA_VANTAGE_API_KEY for news/sentiment.

Run the App

streamlit run app.py

Open http://localhost:8501. Upload a portfolio CSV (or use data/sample_portfolio.csv), then ask questions such as:

• "What's in my portfolio?"
• "What's the current price of AAPL?"
• "Simulate a portfolio 60% AAPL, 40% MSFT over 5 years."
• "What's the Sharpe ratio if my return is 12% and volatility 18%?"

Portfolio CSV Format

Column	Description
ticker	Stock symbol
shares	Number of shares
purchase_price	Price per share

Example: data/sample_portfolio.csv.

Project Layout

• app.py – Streamlit UI and session state.
• config.py – Config and constants.
• agent/graph.py – LangGraph ReAct agent and guardrail wrapper.
• tools/ – Market, math, scenario, sentiment tools.
• rag/store.py – Vector store for portfolio/docs.
• rag/memory.py – Conversation memory (optional).
• guardrails/compliance.py – Response checks and disclaimers.

Why Recruiters Care

• Agentic AI: Tool-calling agents are central in 2025–2026 fintech.
• LangChain/LangGraph: Demonstrates orchestration and state.
• Safety and explainability: Guardrails and trajectory view matter in regulated domains.
• RAG + memory: Personalization over user data and conversation.

Stretch Goals (Senior-Level)

• Reinforcement learning: Optimize strategy in simulation (e.g. reward Sharpe, penalize drawdowns).
• A/B testing: Log and compare response variants for quality.
• Multimodal: Ingest PDF statements and chart images (vision + OCR).

Disclaimer

For educational use only. Not financial, tax, or legal advice.