Sync from GitHub: 12072c8faa1a2b890eeca75571a82673d6b429db

DEPLOYMENT_GUIDE.md

@@ -1,419 +0,0 @@
-# CloverShield ML API Deployment Guide
-
-Complete guide for deploying the ML Inference API to various platforms.
-
-## 📋 Prerequisites
-
-- Docker installed (for containerized deployment)
-- Model file: `Models/fraud_pipeline_final.pkl`
-- Environment variables configured
-- Platform account (Vercel/Render/Railway)
-
-## 🐳 Docker Deployment
-
-### Local Docker
-
-```bash
-# Build image
-docker build -t clovershield-ml-api .
-
-# Run container
-docker run -d \
-  --name clovershield-api \
-  -p 8000:8000 \
-  -v $(pwd)/Models:/app/Models \
-  -e GROQ_API_KEY=your-key \
-  clovershield-ml-api
-
-# Check logs
-docker logs clovershield-api
-
-# Test
-curl http://localhost:8000/health
-```
-
-### Docker Compose
-
-Create `docker-compose.yml`:
-
-```yaml
-version: '3.8'
-
-services:
-  ml-api:
-    build: .
-    ports:
-      - "8000:8000"
-    volumes:
-      - ./Models:/app/Models
-    environment:
-      - MODEL_PATH=Models/fraud_pipeline_final.pkl
-      - GROQ_API_KEY=${GROQ_API_KEY}
-    healthcheck:
-      test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
-      interval: 30s
-      timeout: 10s
-      retries: 3
-```
-
-Run:
-```bash
-docker compose up -d
-```
-
-## ☁️ Platform-Specific Deployment
-
-### Option 1: Render.com (Recommended)
-
-**Pros:**
-- Easy setup
-- Good for ML workloads
-- Free tier available
-- Supports Docker
-
-**Steps:**
-
-1. **Create Account**: Sign up at  [render.com](https://render.com)
-
-2. **Create New Web Service**:
-   - Connect GitHub repository
-   - Select "Docker" as environment
-   - Use `render.yaml` configuration
-
-3. **Configure Environment Variables**:
-   ```
-   MODEL_PATH=Models/fraud_pipeline_final.pkl
-   MODEL_THRESHOLD=0.00754482
-   GROQ_API_KEY=your-key-here
-   ```
-
-4. **Deploy**:
-   - Render will auto-detect `render.yaml`
-   - Build and deploy automatically
-   - Get public URL: `https://clovershield-ml-api.onrender.com`
-
-5. **Update Model File**:
-   - Upload model to cloud storage (S3, etc.)
-   - Modify Dockerfile to download on startup
-   - Or use Render's persistent disk
-
-**Model File Options:**
-- **Option A**: Include in Docker image (not recommended - large)
-- **Option B**: Download from S3/GCS on startup
-- **Option C**: Use Render persistent disk
-
-### Option 2: Railway.app
-
-**Pros:**
-- Simple deployment
-- Good developer experience
-- Automatic HTTPS
-
-**Steps:**
-
-1. **Create Account**: Sign up at [railway.app](https://railway.app)
-
-2. **Create New Project**:
-   - Connect GitHub repository
-   - Railway auto-detects `railway.json`
-
-3. **Configure Environment Variables**:
-   - Go to Variables tab
-   - Add required variables
-
-4. **Deploy**:
-   - Railway builds and deploys automatically
-   - Get public URL: `https://clovershield-ml-api.up.railway.app`
-
-### Option 3: Vercel (Limited)
-
-**Pros:**
-- Fast deployment
-- Great for frontend
-
-**Cons:**
-- Limited for ML workloads
-- Serverless function timeout limits
-- Model size restrictions
-
-**Steps:**
-
-1. **Install Vercel CLI**: `npm i -g vercel`
-
-2. **Deploy**:
-   ```bash
-   vercel
-   ```
-
-3. **Configure**:
-   - Set environment variables in dashboard
-   - Note: Model file must be < 50MB
-
-**Note**: Vercel is not ideal for ML APIs. Use Render or Railway instead.
-
-### Option 4: AWS/GCP/Azure
-
-For production deployments, consider:
-
-- **AWS**: ECS, EKS, or Lambda (with container support)
-- **GCP**: Cloud Run or GKE
-- **Azure**: Container Instances or AKS
-
-## 🔧 Model File Management
-
-### Option 1: Include in Docker Image
-
-**Pros**: Simple  
-**Cons**: Large image size, slower builds
-
-```dockerfile
-COPY Models/fraud_pipeline_final.pkl Models/
-```
-
-### Option 2: Download from Cloud Storage
-
-**Pros**: Smaller image, flexible  
-**Cons**: Requires cloud storage setup
-
-Modify `main.py` startup:
-
-```python
-import boto3
-import os
-
-def download_model():
-    s3 = boto3.client('s3')
-    s3.download_file('your-bucket', 'fraud_pipeline_final.pkl', 'Models/fraud_pipeline_final.pkl')
-```
-
-### Option 3: Environment Variable Path
-
-**Pros**: Flexible  
-**Cons**: Requires manual setup
-
-Set `MODEL_PATH` to point to mounted volume or cloud path.
-
-## 🔒 Production Security
-
-### 1. API Authentication
-
-Add API key middleware:
-
-```python
-from fastapi import Header, HTTPException
-
-API_KEY = os.getenv("API_KEY")
-
-@app.middleware("http")
-async def verify_api_key(request: Request, call_next):
-    if request.url.path.startswith("/docs"):
-        return await call_next(request)
-    
-    api_key = request.headers.get("X-API-Key")
-    if api_key != API_KEY:
-        raise HTTPException(status_code=401, detail="Invalid API key")
-    
-    return await call_next(request)
-```
-
-### 2. Rate Limiting
-
-Install: `pip install slowapi`
-
-```python
-from slowapi import Limiter, _rate_limit_exceeded_handler
-from slowapi.util import get_remote_address
-
-limiter = Limiter(key_func=get_remote_address)
-app.state.limiter = limiter
-app.add_exception_handler(RateLimitExceeded, _rate_limit_exceeded_handler)
-
-@app.post("/predict")
-@limiter.limit("10/minute")
-async def predict(...):
-    ...
-```
-
-### 3. HTTPS
-
-- Use platform's built-in HTTPS (Render, Railway)
-- Or configure reverse proxy (nginx, Traefik)
-
-### 4. Monitoring
-
-Add logging and monitoring:
-
-```python
-import logging
-from prometheus_client import Counter, Histogram
-
-request_count = Counter('requests_total', 'Total requests')
-request_duration = Histogram('request_duration_seconds', 'Request duration')
-
-@app.middleware("http")
-async def log_requests(request: Request, call_next):
-    start_time = time.time()
-    response = await call_next(request)
-    duration = time.time() - start_time
-    request_duration.observe(duration)
-    request_count.inc()
-    return response
-```
-
-## 📊 Health Checks
-
-All platforms support health checks:
-
-```bash
-# Test health endpoint
-curl https://your-api-url.com/health
-
-# Expected response
-{
-  "status": "healthy",
-  "model_loaded": true,
-  "model_version": "1.0.0"
-}
-```
-
-## 🧪 Testing Deployment
-
-### 1. Health Check
-
-```bash
-curl https://your-api-url.com/health
-```
-
-### 2. Prediction Test
-
-```bash
-curl -X POST https://your-api-url.com/predict \
-  -H "Content-Type: application/json" \
-  -d '{
-    "transaction": {
-      "step": 1,
-      "type": "TRANSFER",
-      "amount": 5000.00,
-      "nameOrig": "C123456789",
-      "oldBalanceOrig": 50000.00,
-      "newBalanceOrig": 45000.00,
-      "nameDest": "C234567890",
-      "oldBalanceDest": 25000.00,
-      "newBalanceDest": 30000.00
-    }
-  }'
-```
-
-### 3. Load Testing
-
-Use `ab` (Apache Bench) or `wrk`:
-
-```bash
-# Install
-brew install wrk  # macOS
-# or
-apt-get install apache2-utils  # Linux
-
-# Test
-wrk -t4 -c100 -d30s --script=test.lua https://your-api-url.com/predict
-```
-
-## 🐛 Troubleshooting
-
-### Model Not Loading
-
-**Symptoms**: `503 Model not loaded` error
-
-**Solutions**:
-1. Check `MODEL_PATH` environment variable
-2. Verify model file exists in container
-3. Check file permissions
-4. Review container logs
-
-### Out of Memory
-
-**Symptoms**: Container crashes, OOM errors
-
-**Solutions**:
-1. Increase container memory limit
-2. Use model quantization
-3. Implement model caching
-4. Scale horizontally
-
-### Slow Response Times
-
-**Symptoms**: High latency
-
-**Solutions**:
-1. Optimize model (quantization, pruning)
-2. Use faster hardware (GPU if available)
-3. Implement caching
-4. Optimize feature engineering
-
-### CORS Errors
-
-**Symptoms**: Frontend can't call API
-
-**Solutions**:
-1. Configure CORS in `main.py`
-2. Add frontend domain to allowed origins
-3. Check platform CORS settings
-
-## 📈 Scaling
-
-### Horizontal Scaling
-
-- Use load balancer
-- Deploy multiple instances
-- Use platform auto-scaling
-
-### Vertical Scaling
-
-- Increase container resources
-- Use faster CPU/GPU
-- Optimize model
-
-## ✅ Deployment Checklist
-
-- [ ] Model file accessible
-- [ ] Environment variables set
-- [ ] Health check working
-- [ ] API endpoints responding
-- [ ] CORS configured
-- [ ] Authentication enabled (production)
-- [ ] Rate limiting configured
-- [ ] Monitoring set up
-- [ ] Logs accessible
-- [ ] HTTPS enabled
-- [ ] Backup strategy in place
-
-## 🚀 Quick Deploy Commands
-
-### Render
-
-```bash
-# Using Render CLI
-render deploy
-```
-
-### Railway
-
-```bash
-# Using Railway CLI
-railway up
-```
-
-### Docker Hub + Any Platform
-
-```bash
-# Build and push
-docker build -t your-username/clovershield-ml-api .
-docker push your-username/clovershield-ml-api
-
-# Deploy to platform using image
-```
-
----
-
-**Status**: Ready for deployment! 🚀
-

HUGGINGFACE_DEPLOYMENT.md

@@ -1,123 +0,0 @@
-# Deploying CloverShield ML API to Hugging Face Spaces
-
-This guide outlines the steps to migrate the backend ML API from Render to Hugging Face Spaces using the Docker SDK.
-
-## Prerequisites
-
-- A [Hugging Face account](https://huggingface.co/join).
-- The `ml-api` folder ready with the updated `Dockerfile` and `README.md`.
-
-## Step 1: Create a New Space
-
-1. Log in to your Hugging Face account.
-2. Click on your profile picture in the top right and select **"New Space"**.
-3. Fill in the details:
-   - **Space Name:** `clovershield-ml-api` (or similar).
-   - **License:** MIT or Apache 2.0 (matches your project).
-   - **Space SDK:** Select **Docker**.
-   - **Space Hardware:** **CPU Basic (Free)** is usually sufficient for this model size (2MB). If performance is slow, you can upgrade later.
-   - **Visibility:** **Public** or **Private** (Private is safer for internal testing, but you might need a token to access it from the frontend). For the hackathon demo, **Public** is easiest but be careful with keys. **Private** is recommended if you can handle authentication headers in your frontend.
-
-4. Click **"Create Space"**.
-
-## Step 2: Upload Files
-
-You have two options to upload the code:
-
-### Option A: Web Interface (Easiest)
-
-1. Once the Space is created, you will see the instructions page. Click on the **"Files"** tab.
-2. Click **"Add file"** > **"Upload files"**.
-3. Drag and drop **ALL** files and folders from your local `ml-api` directory into the upload area.
-   - **Important:** Ensure you include:
-     - `Dockerfile`
-     - `requirements.txt`
-     - `main.py`
-     - `inference.py`
-     - `feature_engineering.py`
-     - `entrypoint.sh`
-     - `README.md`
-     - `Models/fraud_pipeline_final.pkl` (The folder structure `Models/` must be preserved).
-     - `dataset/test_dataset.csv` (If used).
-4. In the "Commit changes" box, type "Initial commit" and click **"Commit changes to main"**.
-
-### Option B: Git Command Line (Manual)
-
-1. Clone the Space repository locally.
-2. Copy files and push manually.
-
-### Option C: Automatic Sync with GitHub (Recommended)
-
-This method automatically updates your Space whenever you push code to GitHub.
-
-1.  **Get a Hugging Face Access Token:**
-    *   Go to [Hugging Face Settings > Access Tokens](https://huggingface.co/settings/tokens).
-    *   Click "New token".
-    *   Name it `GITHUB_ACTION` (or similar).
-    *   **Permissions:** Select **"Write"** access.
-    *   Copy the token (starts with `hf_...`).
-
-2.  **Add Token to GitHub Secrets:**
-    *   Go to your GitHub repository.
-    *   Navigate to **Settings** > **Secrets and variables** > **Actions**.
-    *   Click **"New repository secret"**.
-    *   **Name:** `HF_TOKEN`
-    *   **Value:** (Paste your Hugging Face token here).
-    *   Click "Add secret".
-
-3.  **Configure the Workflow:**
-    *   Open the file `.github/workflows/deploy-ml-api.yml` in your project.
-    *   Edit the `env` section to match your details:
-        ```yaml
-        HF_SPACE_USERNAME: "your-hf-username"
-        HF_SPACE_NAME: "clovershield-ml-api"
-        ```
-    *   Commit and push this change to GitHub.
-
-Now, whenever you push changes to the `ml-api` folder in GitHub, the Action will run and update your Hugging Face Space automatically.
-
-## Step 3: Configure Environment Variables (Secrets)
-
-Your API needs the `GROQ_API_KEY` to function if LLM explanations are enabled.
-
-1. Go to the **"Settings"** tab of your Space.
-2. Scroll down to the **"Variables and secrets"** section.
-3. Click **"New secret"**.
-   - **Name:** `GROQ_API_KEY`
-   - **Value:** (Paste your Groq API key here)
-4. (Optional) Add other variables if you changed defaults:
-   - `MODEL_THRESHOLD`
-   - `PAGERANK_LIMIT`
-
-**Note:** Hugging Face automatically sets `PORT` to `7860`. Our updated `Dockerfile` and `entrypoint.sh` are configured to respect this.
-
-## Step 4: Build and Verify
-
-1. Go to the **"App"** tab.
-2. You will see a "Building" status. Click "Logs" to watch the build process.
-3. Once "Running", you should see the Swagger UI or the JSON response from the root endpoint in the preview window.
-4. Your API URL will be: `https://<your-username>-clovershield-ml-api.hf.space`
-
-## Step 5: Update Frontend Connection (Vercel)
-
-Now that your backend is on HF Spaces, you need to update your Vercel deployment to point to the new URL.
-
-1.  **Go to Vercel Dashboard:**
-    *   Navigate to your project on [Vercel](https://vercel.com).
-    *   Go to **Settings** > **Environment Variables**.
-
-2.  **Update the Variable:**
-    *   Find the existing `NEXT_PUBLIC_ML_API_URL` or create a new one.
-    *   **Value:** `https://<your-username>-<your-space-name>.hf.space`
-    *   *Important:* Do NOT include a trailing slash (/) at the end.
-
-3.  **Redeploy:**
-    *   Go to the **Deployments** tab.
-    *   Click on the three dots (...) of your latest deployment and select **Redeploy**. This is necessary for Vercel to pick up the new environment variable value in the client-side code.
-
-## Troubleshooting
-
-- **Build Fails:** Check the "Logs". Common issues are missing dependencies in `requirements.txt` or file path errors.
-- **"403 Forbidden":** If your Space is **Private**, you cannot access it directly from a public frontend without passing an authorization header (Bearer token). For this hackathon/demo, switch the Space Visibility to **Public** in Settings > General.
-- **Model Load Error:** Ensure `Models/fraud_pipeline_final.pkl` was actually uploaded. Check the "Files" tab.
-

QUICK_START.md

@@ -1,92 +0,0 @@
-# ML API Quick Start
-
-Quick reference for deploying CloverShield ML Inference API.
-
-## ✅ What's Included
-
-- ✅ FastAPI service with `/predict` endpoint
-- ✅ Docker containerization
-- ✅ Deployment configs for Render/Railway/Vercel
-- ✅ Complete documentation
-
-## 🚀 Deploy in 5 Minutes
-
-### Option 1: Render.com (Recommended)
-
-1. **Sign up** at [render.com](https://render.com)
-2. **Create Web Service** → Connect GitHub repo
-3. **Configure**:
-   - Environment: Docker
-   - Build Command: (auto-detected)
-   - Start Command: (auto-detected)
-4. **Set Environment Variables**:
-   ```
-   MODEL_PATH=Models/fraud_pipeline_final.pkl
-   GROQ_API_KEY=your-key (optional)
-   ```
-5. **Deploy** → Get URL: `https://your-app.onrender.com`
-
-### Option 2: Railway.app
-
-1. **Sign up** at [railway.app](https://railway.app)
-2. **New Project** → Connect GitHub
-3. **Railway auto-detects** `railway.json`
-4. **Set environment variables** in dashboard
-5. **Deploy** → Get URL automatically
-
-### Option 3: Local Docker
-
-```bash
-# Build
-docker build -t clovershield-ml-api .
-
-# Run
-docker run -p 8000:8000 \
-  -v $(pwd)/Models:/app/Models \
-  clovershield-ml-api
-
-# Test
-curl http://localhost:8000/health
-```
-
-## 📝 API Usage
-
-### Health Check
-```bash
-curl https://your-api-url.com/health
-```
-
-### Predict Fraud
-```bash
-curl -X POST https://your-api-url.com/predict \
-  -H "Content-Type: application/json" \
-  -d '{
-    "transaction": {
-      "step": 1,
-      "type": "TRANSFER",
-      "amount": 5000.00,
-      "nameOrig": "C123456789",
-      "oldBalanceOrig": 50000.00,
-      "newBalanceOrig": 45000.00,
-      "nameDest": "C234567890",
-      "oldBalanceDest": 25000.00,
-      "newBalanceDest": 30000.00
-    }
-  }'
-```
-
-## 📚 Full Documentation
-
-- [README.md](./README.md) - Complete API documentation
-- [DEPLOYMENT_GUIDE.md](./DEPLOYMENT_GUIDE.md) - Detailed deployment guide
-
-## ✅ Checklist
-
-- [ ] Model file available (`Models/fraud_pipeline_final.pkl`)
-- [ ] Environment variables configured
-- [ ] API deployed and accessible
-- [ ] Health check passing
-- [ ] Test prediction working
-
-**Ready to deploy!** 🚀
-

README.md

@@ -1,310 +0,0 @@
----
-title: CloverShield ML API
-emoji: 🛡️
-colorFrom: green
-colorTo: blue
-sdk: docker
-app_port: 7860
----
-
-# CloverShield ML Inference API
-
-FastAPI microservice for fraud detection predictions. This service replaces the local model loading mechanism in the Streamlit app.
-
-## 🚀 Quick Start
-
-### Local Development
-
-1. **Install dependencies:**
-   ```bash
-   pip install -r requirements.txt
-   ```
-
-2. **Copy model file:**
-   ```bash
-   cp ../Models/fraud_pipeline_final.pkl Models/
-   ```
-
-3. **Set environment variables:**
-   ```bash
-   cp env.template .env
-   # Edit .env with your values
-   ```
-
-4. **Run the API:**
-   ```bash
-   python main.py
-   # Or
-   uvicorn main:app --reload
-   ```
-
-5. **Test the API:**
-   ```bash
-   curl http://localhost:8000/health
-   ```
-
-### Docker
-
-1. **Build the image:**
-   ```bash
-   docker build -t clovershield-ml-api . 
-   ```
-
-2. **Run the container:**
-   ```bash
-   docker run -p 8000:8000 \
-     -v $(pwd)/Models:/app/Models \
-     -e GROQ_API_KEY=your-key \
-     clovershield-ml-api
-   ```
-
-## 📚 API Endpoints
-
-### `GET /`
-Root endpoint - service information
-
-### `GET /health`
-Health check endpoint
-
-**Response:**
-```json
-{
-  "status": "healthy",
-  "model_loaded": true,
-  "model_version": "1.0.0",
-  "shap_available": true,
-  "llm_available": true
-}
-```
-
-### `GET /model/info`
-Get model metadata
-
-### `POST /predict`
-Predict fraud probability for a single transaction
-
-**Request:**
-```json
-{
-  "transaction": {
-    "step": 1,
-    "type": "CASH_OUT",
-    "amount": 5000.00,
-    "nameOrig": "C123456789",
-    "oldBalanceOrig": 50000.00,
-    "newBalanceOrig": 45000.00,
-    "nameDest": "C234567890",
-    "oldBalanceDest": 25000.00,
-    "newBalanceDest": 30000.00
-  },
-  "options": {
-    "include_shap": true,
-    "include_llm_explanation": false,
-    "language": "en",
-    "topk": 10
-  }
-}
-```
-
-**Response:**
-```json
-{
-  "transaction_id": "550e8400-e29b-41d4-a716-446655440000",
-  "prediction": {
-    "fraud_probability": 0.2345,
-    "decision": "pass",
-    "risk_level": "low",
-    "confidence": 0.85
-  },
-  "shap_explanations": [
-    {
-      "feature": "amount_over_oldBalanceOrig",
-      "value": 0.1,
-      "shap": 0.0234,
-      "shap_abs": 0.0234,
-      "rank": 1
-    }
-  ],
-  "llm_explanation": null,
-  "processing_time_ms": 145,
-  "model_version": "1.0.0",
-  "timestamp": "2024-01-15T10:30:00Z"
-}
-```
-
-### `POST /predict/batch`
-Predict fraud probability for multiple transactions
-
-## 🐳 Docker Deployment
-
-### Build and Push to Registry
-
-```bash
-# Build
-docker build -t clovershield-ml-api:latest . 
-
-# Tag for registry
-docker tag clovershield-ml-api:latest your-registry/clovershield-ml-api:latest
-
-# Push
-docker push your-registry/clovershield-ml-api:latest
-```
-
-## ☁️ Platform Deployment
-
-### Hugging Face Spaces (Recommended)
-
-1. Create a new Space on Hugging Face.
-2. Select "Docker" as the SDK.
-3. Upload the contents of this directory to the Space.
-4. Set the `GROQ_API_KEY` secret in the Space settings.
-
-
-### Vercel
-
-1. Install Vercel CLI: `npm i -g vercel`
-2. Deploy: `vercel`
-3. Set environment variables in Vercel dashboard
-
-**Note:** Vercel has limitations for ML models. Consider Render or Railway for better performance.
-
-### Render
-
-1. Connect your GitHub repository
-2. Create new Web Service
-3. Use `render.yaml` configuration
-4. Set environment variables in dashboard
-5. Deploy
-
-### Railway
-
-1. Connect your GitHub repository
-2. Create new project
-3. Railway will auto-detect `railway.json`
-4. Set environment variables
-5. Deploy
-
-## 🔧 Configuration
-
-### Environment Variables
-
-- `MODEL_PATH` - Path to model file (default: `Models/fraud_pipeline_final.pkl`)
-- `TEST_DATASET_PATH` - Path to test dataset CSV for feature engineering (optional, auto-detected if not provided)
-- `MODEL_THRESHOLD` - Decision threshold (default: `0.00754482` - optimized for 99% recall)
-- `PORT` - Server port (default: `8000`)
-- `HOST` - Server host (default: `0.0.0.0`)
-- `GROQ_API_KEY` - Optional Groq API key for LLM explanations
-- `PAGERANK_LIMIT` - Optional limit on nodes for PageRank computation (for memory optimization)
-- `MAX_FIT_ROWS` - Maximum rows to use for feature engineering fitting (default: `50000`)
-
-### Model File
-
-The model file (`fraud_pipeline_final.pkl`) must be available at runtime. Options:
-
-1. **Include in Docker image** (not recommended for large files)
-2. **Mount as volume** (Docker)
-3. **Download from cloud storage** (S3, GCS, etc.)
-4. **Use environment variable** to specify path
-
-## 📊 API Documentation
-
-Interactive API documentation available at:
-- Swagger UI: `http://localhost:8000/docs`
-- ReDoc: `http://localhost:8000/redoc`
-
-## 🧪 Testing
-
-### Test with curl
-
-```bash
-# Health check
-curl http://localhost:8000/health
-
-# Prediction
-curl -X POST http://localhost:8000/predict \
-  -H "Content-Type: application/json" \
-  -d '{
-    "transaction": {
-      "step": 1,
-      "type": "TRANSFER",
-      "amount": 5000.00,
-      "nameOrig": "C123456789",
-      "oldBalanceOrig": 50000.00,
-      "newBalanceOrig": 45000.00,
-      "nameDest": "C234567890",
-      "oldBalanceDest": 25000.00,
-      "newBalanceDest": 30000.00
-    }
-  }'
-```
-
-### Test with Python
-
-```python
-import requests
-
-response = requests.post(
-    "http://localhost:8000/predict",
-    json={
-        "transaction": {
-            "step": 1,
-            "type": "TRANSFER",
-            "amount": 5000.00,
-            "nameOrig": "C123456789",
-            "oldBalanceOrig": 50000.00,
-            "newBalanceOrig": 45000.00,
-            "nameDest": "C234567890",
-            "oldBalanceDest": 25000.00,
-            "newBalanceDest": 30000.00
-        }
-    }
-)
-
-print(response.json())
-```
-
-## 🔒 Security
-
-**Production Recommendations:**
-- Add API key authentication for production
-- Use HTTPS in production
-- Implement rate limiting
-- Validate all inputs
-- Monitor for abuse
-- Restrict CORS origins (currently set to `["*"` for development)
-- Use environment variables for sensitive data
-
-## 📈 Performance
-
-**Optimization Tips:**
-- Model loads once on startup (lazy loading supported for serverless)
-- Single worker recommended (ML models are not thread-safe)
-- Consider caching for frequently accessed data
-- Monitor memory usage (XGBoost models can be large)
-- Feature engineering uses sampled dataset (configurable via `MAX_FIT_ROWS`)
-- SHAP background uses minimal sample size (100 rows) for efficiency
-
-## 🐛 Troubleshooting
-
-### Model not loading
-
-- Check `MODEL_PATH` environment variable
-- Verify model file exists
-- Check file permissions
-- Review logs for error messages
-
-### SHAP not working
-
-- Ensure XGBoost is installed
-- Check model type compatibility
-- Review SHAP initialization logs
-
-### LLM explanations failing
-
-- Verify `GROQ_API_KEY` is set
-- Check API key validity
-- Review Groq API status
-
-## 📝 License
-
-See main project LICENSE file.

railway.json

@@ -1,13 +0,0 @@
-{
-  "$schema": "https://railway.app/railway.schema.json",
-  "build": {
-    "builder": "DOCKERFILE",
-    "dockerfilePath": "Dockerfile"
-  },
-  "deploy": {
-    "startCommand": "uvicorn main:app --host 0.0.0.0 --port $PORT",
-    "restartPolicyType": "ON_FAILURE",
-    "restartPolicyMaxRetries": 10
-  }
-}
-

render.yaml

@@ -1,26 +0,0 @@
-# Render.com deployment configuration for CloverShield ML API
-
-services:
-  - type: web
-    name: clovershield-ml-api
-    runtime: docker
-    dockerfilePath: ./Dockerfile
-    dockerContext: .
-    envVars:
-      - key: MODEL_PATH
-        value: Models/fraud_pipeline_final.pkl
-      - key: MODEL_THRESHOLD
-        value: 0.0793
-      - key: PORT
-        value: 8000
-      - key: HOST
-        value: 0.0.0.0
-      - key: MAX_FIT_ROWS
-        value: "30000"  # Reduced for Render free tier (512MB RAM)
-      - key: PAGERANK_LIMIT
-        value: "5000"  # Reduced for memory efficiency
-      - key: GROQ_API_KEY
-        sync: false  # Set in Render dashboard
-    healthCheckPath: /health
-    plan: starter  # Upgrade for production
-

vercel.json

@@ -1,20 +0,0 @@
-{
-  "version": 2,
-  "builds": [
-    {
-      "src": "main.py",
-      "use": "@vercel/python"
-    }
-  ],
-  "routes": [
-    {
-      "src": "/(.*)",
-      "dest": "main.py"
-    }
-  ],
-  "env": {
-    "MODEL_PATH": "Models/fraud_pipeline_final.pkl",
-    "MODEL_THRESHOLD": "0.0793"
-  }
-}
-