🛡️ Autonomous SEBI Compliance Agent

An AI-powered agent that monitors SEBI regulatory circulars 24/7, extracts compliance obligations using LLM, and auto-assigns tasks to teams — built to understand what funded Indian fintech startups like OnFinance AI ($4.2M raised) are solving.

---

🎯 What Problem Does This Solve?

Every time SEBI (India's stock market regulator) releases a new circular, compliance teams at banks, mutual funds, and brokerages must:

• Manually check the SEBI website
• Read and interpret dense legal PDFs
• Figure out what the company must do, by when, and who is responsible
• Assign work to legal, compliance, and operations teams

This takes hours. Every single time. For every circular.

This agent does it all automatically in seconds.

---

✨ Features

• 🔄 Autonomous Monitoring — Wakes up every hour, checks SEBI website for new circulars
• 🧠 AI-Powered Extraction — Llama 3.3 70B reads actual circular PDFs and extracts obligations, deadlines, penalties
• 📋 Auto Task Creation — Creates tasks for compliance, legal, and ops teams with priorities
• 💾 Memory — Remembers processed circulars, never duplicates work
• 📊 Live Dashboard — Streamlit UI with metrics, filters, and export to JSON/CSV
• 💰 Zero Cost — Built entirely on free APIs and free hosting

---

🏗️ Architecture & Workflow

SEBI Website (sebi.gov.in)
         ↓
  SEBIScraper fetches new circulars
         ↓
  Filter — already processed? Skip
         ↓
  Fetch actual PDF/HTML content
         ↓
  Groq AI (Llama 3.3 70B) reads & extracts:
  • Key obligations
  • Deadlines
  • Penalties
  • Applicable entities
  • Affected departments
  • Impact level (HIGH/MEDIUM/LOW)
         ↓
  Auto-create 3 tasks per circular
  (Compliance Team / Legal Team / Ops Team)
         ↓
  Save state + Update Streamlit Dashboard
         ↓
  Wait 1 hour → Repeat

---

🛠️ Tech Stack

Component	Tool	Cost
LLM Brain	Groq API — Llama 3.3 70B	₹0 Free
Web Scraping	BeautifulSoup + Requests	₹0 Free
PDF Extraction	PyPDF2	₹0 Free
UI Dashboard	Streamlit	₹0 Free
Hosting	Streamlit Community Cloud	₹0 Free

Total infrastructure cost: ₹0

---

📁 Project Structure

compliance-agent/
├── autonomous_agent.py      # Main agent loop & orchestration
├── sebi_scraper.py          # SEBI website scraper & PDF fetcher
├── groq_extractor.py        # Groq LLM integration & requirement extraction
├── streamlit_app.py         # Streamlit dashboard UI
├── requirements.txt         # Python dependencies
├── .env.example             # Environment variables template
└── README.md                # This file

---

🚀 Quick Start

Prerequisites

• Python 3.10+
• Free Groq API key from console.groq.com

Installation

# Clone the repo
git clone https://github.com/aditya8975/compilance-agent.git
cd compilance-agent

# Install dependencies
pip install -r requirements.txt

# Set up environment
cp .env.example .env
# Add your Groq API key to .env

Run Locally

# Run the Streamlit dashboard
streamlit run streamlit_app.py

# Or run the agent directly
python autonomous_agent.py

Environment Variables

GROQ_API_KEY=your_groq_api_key_here
MONITORING_INTERVAL=3600
LOG_LEVEL=INFO

---

📸 Demo

Dashboard Overview

• Real-time compliance metrics
• Latest SEBI circulars with AI-extracted obligations
• Task management with priority levels
• Export reports as JSON or CSV

What the AI Extracts from Each Circular

{
  "circular_id": "SEBI/HO/OIAE/2024/001",
  "title": "Guidelines on Disclosure of Material Information",
  "key_obligations": [
    "Disclose material information within 30 minutes of board decision",
    "Appoint a dedicated compliance officer for disclosures"
  ],
  "deadline": "01-04-2024",
  "applicable_to": ["All Listed Entities"],
  "penalties": ["Fine up to ₹25 Crore", "Trading suspension"],
  "impact_level": "HIGH",
  "affected_departments": ["Compliance", "Legal", "Board Secretariat"]
}

---

🧠 Why This Is a Real AI Agent

Three things separate this from a plain script:

1. Autonomous — Runs continuously without human triggers
2. Memory — Tracks processed circulars in .agent_state.json, never duplicates work
3. Action — Doesn't just answer questions, actively creates tasks and reports

---

📈 Inspired By

This project was built to understand the problem space that funded Indian AI startups are solving:

• OnFinance AI — Raised $4.2M for BFSI compliance automation, built NeoGPT (fine-tuned on SEBI/RBI/IRDAI data) and 70+ compliance agents
• ClearTax — Tax and GST compliance automation
• Setu — Financial data APIs for Indian market

The key insight: the real moat for production systems is domain-specific fine-tuned models trained on decades of regulatory data — not general-purpose LLMs. This project is a starting point that demonstrates the agent architecture.

---

🔮 Roadmap

• Add RBI and IRDAI circular monitoring
• Fine-tune model on historical SEBI circulars
• Email/Slack notifications for high priority circulars
• Multi-language support (Hindi + English)
• Compliance scoring dashboard
• Integration with HRMS for task assignment

---

🤝 Contributing

Feel free to open issues and pull requests. Areas where contribution is welcome:

• Better SEBI scraping logic
• More regulatory sources
• Improved AI prompts for extraction accuracy
• UI enhancements

---

📝 License

MIT License — use freely for personal and commercial projects.

---

👨‍💻 Author

Aditya — Building in public, learning by doing.

Open to opportunities in AI/ML engineering at Indian startups working on agentic workflows, fintech, or compliance automation.

LinkedIn • GitHub

---