📈 PortfolioPilot AI

PortfolioPilot AI is an intelligent, multi-agent financial chatbot that securely connects to your live Zerodha account. It leverages a sophisticated backend using LangGraph to orchestrate specialized AI agents, OpenAI for language understanding, and provides a real-time conversational UI with Streamlit.

---

🚀 Features

🧠 Interactive Financial Analysis

Run the web application, securely log in to your Zerodha account, and ask complex financial questions in plain English.

🧩 Intelligent Multi-Agent Framework

Built with LangGraph to create a robust workflow that intelligently routes your request to the best-specialized agent for the job:

• 📊 Portfolio Agent – Connects to your live Zerodha account to fetch holdings and positions.
• 🌐 Market Research Agent – Explores real-time news and market data using the Tavily search API.
• ⚠️ Risk Analyst – Intelligently assesses portfolio diversification and sector concentration.
• 📈 Technical Analyst – Calculates key indicators like RSI from historical stock data and uses real-time data for analysis.

🔁 Stateful & Context-Aware

The bot remembers the full context of your conversation, allowing for natural follow-up questions (e.g., “what are the risks in that sector?”)

📡 Data-Driven & Reliable

Agents are grounded with strict prompts to base their answers on factual data from APIs, reducing LLM hallucinations.

💻 User-Friendly Web Interface

A clean UI built with Streamlit that handles secure authentication flow and provides a seamless chat experience.

---

🛠️ Technology Stack

Layer	Tech Used
Orchestration	LangGraph
Language Model	OpenAI (GPT-4o-mini)
Web Interface	Streamlit
Broker API	Zerodha Kite Connect
Web Search	Tavily AI
Stock Data	yfinance, Tavily Search
Package Manager	uv

---

⚙️ Prerequisites

🖥️ System

• Python 3.11+
• uv (recommended Python package manager)
• graphviz (for agent graph visualization)

🔑 API Keys

Create a .env file in the root of the project with the following keys:

# Required for agent functionality
OPENAI_API_KEY="sk-..."
TAVILY_API_KEY="tvly-..."

# Required for live broker connection
ZERODHA_API_KEY="your_zerodha_api_key"
ZERODHA_API_SECRET="your_zerodha_api_secret"

---

🧰 Installation & Setup

1️⃣ Clone the Repository

git clone https://github.com/your-username/PortfolioPilot-AI.git
cd PortfolioPilot-AI

Replace your-username with your actual GitHub username.

---

2️⃣ Create the Environment File

touch .env

Populate .env with your API keys (see above).

---

3️⃣ Setup Backend with uv

# Create the virtual environment
uv venv

# Activate the environment
# On Windows (PowerShell):
.venv\Scripts\activate

# On macOS/Linux:
source .venv/bin/activate

# Install dependencies
uv pip install -r requirements.txt

---

▶️ Usage

1️⃣ Run the Application

streamlit run app.py

---

2️⃣ Log In and Interact

• Your browser will open with the PortfolioPilot AI login screen.
• Follow the instructions to log in via Zerodha.
• You will be redirected to a URL with a request_token.
• Paste that token back into the Streamlit app.

---

✅ Welcome to PortfolioPilot AI!

🔗 Please log in using this URL:
https://kite.zerodha.com/connect/login?api_key=...

🔑 Please paste the request_token here:
[_________________]