AML Policy FAQ Bot

AI-powered AML (Anti-Money Laundering) compliance chatbot using RAG with LangChain, LangGraph, NVIDIA LLM, and Qdrant Cloud.

---

Project Structure

AML_Policy_FAQ_BOT/
├── backend/                 # FastAPI + Lambda
│   ├── app/                 # Application code
│   ├── handler.py           # Lambda entry point
│   └── pyproject.toml       # Dependencies (uv)
├── frontend/                # Vite + React
│   ├── src/                 # React components
│   └── package.json
├── infrastructure/
│   ├── bootstrap/           # Terraform state (run first)
│   ├── backend/             # Lambda, API Gateway, S3
│   └── frontend/            # S3, CloudFront
└── .github/workflows/
    ├── infrastructure.yml         # Backend infra
    ├── infrastructure-frontend.yml # Frontend infra
    ├── deploy-backend.yml         # Deploy Lambda
    └── deploy-frontend.yml        # Deploy React

---

Quick Start (Local Development)

Backend

cd backend
uv sync

# Create .env
cat > .env << EOF
NVIDIA_API_KEY=your_nvidia_api_key
NVIDIA_MODEL_NAME=meta/llama-3.1-70b-instruct
NVIDIA_EMBEDDING_MODEL_NAME=nvidia/nv-embedqa-e5-v5
QDRANT_URL=https://your-cluster.qdrant.io:6333
QDRANT_API_KEY=your_qdrant_api_key
EOF

# Run
uv run uvicorn app.main:app --reload --port 8000

Frontend

cd frontend
npm install

# Create .env
echo 'VITE_API_URL=http://localhost:8000' > .env

# Run
npm run dev

Open http://localhost:5173

---

AWS Deployment

Prerequisites

1. AWS Account with admin access
2. NVIDIA API Key from build.nvidia.com
3. Qdrant Cloud account from cloud.qdrant.io

Step 1: Bootstrap (Run Once Locally)

cd infrastructure/bootstrap
terraform init
terraform apply

Note the outputs: S3 bucket and DynamoDB table names.

Step 2: Configure GitHub Secrets

Add these secrets to your GitHub repository:

Secret	Description
AWS_ACCESS_KEY_ID	AWS access key
AWS_SECRET_ACCESS_KEY	AWS secret key
NVIDIA_API_KEY	NVIDIA API key
NVIDIA_MODEL_NAME	meta/llama-3.1-70b-instruct
NVIDIA_EMBEDDING_MODEL_NAME	nvidia/nv-embedqa-e5-v5
QDRANT_URL	Qdrant Cloud URL
QDRANT_API_KEY	Qdrant Cloud API key

Step 3: Deploy Infrastructure

Run workflows in this order:

1. Infrastructure (Backend)     → apply → Creates Lambda, API Gateway, S3
2. Infrastructure (Frontend)    → apply → Creates S3, CloudFront
3. Deploy Backend               → Deploys Lambda code
4. Deploy Frontend              → Builds React, uploads to S3

Step 4: Access Your App

After deployment, you'll see URLs in the workflow summary:

• Frontend: https://d1234567890.cloudfront.net
• Backend API: https://xxxx.execute-api.ap-south-1.amazonaws.com

---

API Endpoints

Method	Endpoint	Description
GET	/api/v1/health	Health check
POST	/api/v1/ingest	Upload documents
POST	/api/v1/query	Query (sync)
WS	/api/v1/ws/query	Query (streaming)
GET	/docs	Swagger UI

---

Architecture

                    ┌─────────────────┐
                    │   CloudFront    │
                    │  (Frontend CDN) │
                    └────────┬────────┘
                             │
┌────────────────────────────┼────────────────────────────────┐
│                            │                                │
│   ┌────────────────┐       │       ┌────────────────┐       │
│   │   S3 Bucket    │◀──────┘       │  API Gateway   │       │
│   │  (React App)   │               │   (HTTP API)   │       │
│   └────────────────┘               └───────┬────────┘       │
│                                            │                │
│                                    ┌───────▼────────┐       │
│                                    │     Lambda     │       │
│                                    │   (FastAPI)    │       │
│                                    └───┬────────┬───┘       │
│                                        │        │           │
│                            ┌───────────▼──┐  ┌──▼──────────┐|
│                            │ Qdrant Cloud │  │ NVIDIA LLM  │|
│                            │ (Vectors)    │  │ (Inference) │|
│                            └──────────────┘  └─────────────┘|
└─────────────────────────────────────────────────────────────┘

---

Environment Variables

Variable	Required	Description
NVIDIA_API_KEY	✅	NVIDIA API key for LLM
NVIDIA_MODEL_NAME	✅	LLM model name
NVIDIA_EMBEDDING_MODEL_NAME	✅	Embedding model name
QDRANT_URL	✅	Qdrant Cloud cluster URL
QDRANT_API_KEY	✅	Qdrant Cloud API key
S3_BUCKET	Lambda	S3 bucket for documents
VITE_API_URL	Frontend	Backend API URL

---

Destroy Infrastructure

To tear down AWS resources:

# Via GitHub Actions
1. Run "Infrastructure (Frontend)" → destroy
2. Run "Infrastructure (Backend)" → destroy

# Or locally
cd infrastructure/frontend && terraform destroy
cd infrastructure/backend && terraform destroy

---

Tech Stack

• Backend: FastAPI, LangChain, LangGraph, NVIDIA AI Endpoints
• Frontend: React, Vite
• Vector Store: Qdrant Cloud
• Infrastructure: Terraform, AWS Lambda, API Gateway, CloudFront
• CI/CD: GitHub Actions