AI Compliance Risk Copilot

AI Compliance Risk Copilot is an enterprise-grade Governance, Risk, and Compliance (GRC) platform that parses regulatory files, contracts, and legal agreements to extract clauses, audit regulatory alignment, calculate predictive risk indices, and generate C-suite executive summaries.

---

🖥️ Application UI

---

Architecture Overview

graph TD
  User([Compliance Officer]) -->|Upload PDF / Chat| Frontend[Next.js 15 App Portal]
  Frontend -->|REST APIs / JWT| API[FastAPI Gateway]
  
  subgraph Vector RAG Layer
    API -->|Local Storage| Qdrant[(Qdrant Vector DB)]
    API -->|Text Embeddings| Embedder[BAAI/bge-small-en-v1.5]
  end

  subgraph Database Layer
    API -->|SQLAlchemy| DB[(MySQL Database)]
  end

  subgraph LangGraph Flow [LangGraph Multi-Agent Pipeline]
    API -->|Triggers| Workflow[Sequential Workflow]
    Workflow --> A1[Doc Classifier]
    A1 --> A2[Clause Extractor]
    A2 --> A3[Compliance Mapper]
    A3 --> A4[Risk Analyst]
    A4 --> A5[ML Predictor]
    A5 --> A6[SHAP Explainer]
    A6 --> A7[Recommendation Agent]
    A7 --> A8[Executive Reporter]
    
    A1 & A2 & A3 & A4 & A5 & A6 & A7 & A8 -.->|Audit Trails| DB
  end

  subgraph Machine Learning Layer
    A5 -->|Tabular Inference| MLModel[XGBoost / RandomForest Regressor]
    A6 -->|Inference Explanation| SHAP[SHAP Explainer Engine]
  end
  
  MLModel & SHAP -.->|Auto-Train/Inference Assets| MLDir[ml/models/]

Loading

Relational Database Schema

erDiagram
    users ||--o{ documents : uploads
    documents ||--o{ document_chunks : partitions
    documents ||--o{ clauses : extracts
    documents ||--o{ compliance_results : audits
    documents ||--o| risk_assessments : scores
    documents ||--o{ recommendations : mitigates
    documents ||--o| executive_reports : briefs
    documents ||--o{ audit_logs : logs

    users {
        int id PK
        string name
        string email UK
        string password_hash
        string role
        datetime created_at
    }
    documents {
        int id PK
        int user_id FK
        string filename
        string document_type
        string status
        string file_path
        datetime uploaded_at
    }
    document_chunks {
        int id PK
        int document_id FK
        text chunk_text
        int chunk_index
    }
    clauses {
        int id PK
        int document_id FK
        string clause_type
        text clause_text
        string risk_level
    }
    compliance_results {
        int id PK
        int document_id FK
        string framework
        string requirement
        string status
        text gap_description
    }
    risk_assessments {
        int id PK
        int document_id FK
        float risk_score
        string risk_level
        string prediction_model
    }
    recommendations {
        int id PK
        int document_id FK
        text recommendation
    }
    executive_reports {
        int id PK
        int document_id FK
        text summary
    }
    audit_logs {
        int id PK
        int document_id FK
        string agent_name
        string action
        text input_data
        text output_data
        datetime created_at
    }

Loading

---

Tech Stack

• Frontend: Next.js 15, TypeScript, Tailwind CSS, Zustand, React Query, Lucide Icons, Recharts.
• Backend API: FastAPI, Uvicorn, Pydantic Settings, SQLAlchemy ORM, PyMySQL.
• Relational DB: MySQL (local schema ai_compliance).
• Vector DB: Qdrant (client-embedded storage in backend/qdrant_db).
• Embeddings: BAAI/bge-small-en-v1.5 loaded locally via sentence-transformers on CPU/GPU.
• LLM Engine: Ollama running Llama 3.1 (llama3.1:latest).
• AI Agent Framework: LangGraph, LangChain.
• Machine Learning: Scikit-Learn, XGBoost, SHAP.

---

Environment Setup & Installation

Prerequisite Checklist

1. Ollama installed and running on localhost:11434.
   • Download the model: ollama pull llama3.1
2. MySQL Server installed and running on localhost:3306.
   • Ensure you have a root account or configured user matching the connection details.

1. Backend Setup

1. Open a terminal in the backend/ directory:

   cd backend

2. Activate the virtual environment:
   • Windows PowerShell:

     .\venv\Scripts\Activate.ps1

   • Windows Command Prompt:

     .\venv\Scripts\activate.bat

3. Verify that packages are installed. (They are auto-installed in the virtual environment).
4. Run the database creation and seeding script:

   python seed.py

   This automatically creates the MySQL database ai_compliance, generates all table schemas, and seeds default credentials:
   • Admin User: admin@compliance.com / Password: admin123
   • Analyst User: analyst@compliance.com / Password: analyst123
5. Run the ML Model Training pipeline:

   $env:PYTHONPATH=".."
   python ml/training/train.py

   This generates 10,000 synthetic GRC risk rows, evaluates RandomForest, XGBoost, and SVR regressors, and saves the best model and SHAP explainability assets.

2. Frontend Setup

1. Open a terminal in the frontend/ directory:

   cd frontend

2. Install npm dependencies:

   npm install

3. Start the Next.js local development server:

   npm run dev

   The client portal will be listening at http://localhost:3000.

---

Running the Application

1. Launch Backend API

In the backend/ directory with the virtual environment activated, start the Uvicorn server:

uvicorn app.main:app --reload --host 127.0.0.1 --port 8000

API docs will be available at http://localhost:8000/docs.

2. Launch Frontend Portal

In the frontend/ directory, start Next.js dev server:

npm run dev

Open http://localhost:3000 in your browser. Log in with:

• Email: analyst@compliance.com
• Password: analyst123

---

Verification & Testing

To run the automated test suite verifying MySQL connection, XGBoost risk predictions, SHAP calculations, local Qdrant collection vectors, and LangGraph agent runs:

# In the workspace root directory:
$env:PYTHONPATH="D:\ML,NLP,DL\AI Compliance Risk Copilot"
backend\venv\Scripts\python verify_system.py