🏆 Claude Certified Architect — Portfolio Project

Deep Agentic AI for
Indian Stock Analysis

8 specialized AI agents orchestrated via LangGraph. Works with Groq (free), Ollama (local), or Claude Sonnet. Institutional-grade equity research for NSE/BSE stocks — run locally with your own API key.

⭐ View on GitHub 🚀 Run Locally
Python 3.11+
Claude Sonnet 4.6
LangGraph
LangChain
Streamlit
yfinance NSE/BSE Data
Plotly Charts
scipy Portfolio Optimization
ta Technical Indicators

🎯 What you can ask

Natural language queries — the orchestrator routes to the right agents automatically

Full analysis of RELIANCE Screen IT stocks with P/E < 25 & ROE > 15% DCF valuation for TCS at 12% growth Risk analysis for HDFCBANK Build a ₹5L diversified portfolio High dividend yield stocks — NSE INFY technical analysis — buy or sell? HINDUNILVR competitive moat vs peers

8 Specialized AI Agents

Each agent is a LangChain tool-calling agent with a focused system prompt and dedicated tools

🔍
Stock Screener
Filters 30+ NSE large-caps by P/E, ROE, market cap, debt, dividend yield, and sector.
💰
DCF Valuation
Discounted Cash Flow with base/bull/bear scenarios, margin of safety, and intrinsic value per share.
⚠️
Risk Analysis
Beta, alpha, Sharpe ratio, VaR (95%/99%), CVaR, max drawdown vs Nifty 50 benchmark.
📊
Earnings Breakdown
Quarterly EPS actuals vs estimates, revenue trend, margin expansion/compression, beat/miss history.
🗂️
Portfolio Builder
Markowitz mean-variance optimization, optimal weights, expected return/volatility, sector diversification.
📉
Technical Analysis
RSI, MACD, Bollinger Bands, EMA/SMA crossovers, ADX, Stochastic, ATR, OBV — buy/sell signals.
💸
Dividend Strategy
Dividend yield, payout ratio, FCF coverage, multi-year growth history, consistency scoring.
🏆
Competitive Advantage
Economic moat analysis, Porter's Five Forces, peer comparison, moat durability rating (Wide/Narrow/None).

Multi-Agent Architecture

LangGraph StateGraph with supervisor routing and parallel agent execution

User Query (natural language)
        │
        ▼
┌──────────────────────┐
│   Supervisor Node    │  ─── keyword routing + ticker extraction
│   (LangGraph)        │
└──────────┬───────────┘
           │  parallel dispatch to relevant agents
    ┌──────┴───────────────────────────────────────────┐
    ▼        ▼        ▼        ▼       ▼       ▼        ▼       ▼
[Screener] [DCF]  [Risk]  [Earnings] [Portf.] [Tech.] [Div.] [Comp.]
    │        │        │        │       │       │        │       │
    └────────┴────────┴────────┴───────┴───────┴────────┴───────┘
                              │
                              ▼
                    ┌──────────────────┐
                    │  Synthesis Node  │  ─── progressive summarization
                    │  (Claude Sonnet) │       context positioning
                    └────────┬─────────┘
                             │
                             ▼
                    Final Investment Report
                    (BUY/HOLD/SELL + rationale)

Claude Certified Architect Concepts

Applied concepts from the 5 exam domains

Domain 1
Agentic Architecture & Orchestration
LangGraph StateGraph Supervisor routing Agentic loop Task decomposition Multi-agent coordination
Domain 2
Tool Design & MCP Integration
Custom @tool decorators Descriptive tool docstrings JSON-only returns Error status fields
Domain 3
Claude Code Configuration
CLAUDE.md hierarchy Dev conventions Tool routing rules
Domain 4
Prompt Engineering & Structured Output
Explicit JSON schema Few-shot examples Per-agent system prompts Validation-retry loop
Domain 5
Context Management & Reliability
Progressive summarization Context positioning Error propagation Status fields vs exceptions

Setup & Run Locally

You need an Anthropic API key. No deployment costs — runs entirely on your machine.

Clone the repository

git clone https://github.com/PritamSonawane7798/india-stock-ai.git
cd india-stock-ai

Create a virtual environment and install dependencies

python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt

Add your Anthropic API key

cp .env.example .env
# Edit .env and set:
ANTHROPIC_API_KEY=sk-ant-...

Get your API key at console.anthropic.com. The app uses Claude Sonnet — check current pricing there.

Launch the Streamlit app

streamlit run src/app.py

Opens at http://localhost:8501

Or run the demo script (CLI)

python examples/demo_analysis.py