Satellite Data Viewer Backend

FastAPI service for downloading satellite imagery from Microsoft Planetary Computer. Deployed serverless on AWS Lambda with Docker.

Note: Download functionality operates in controlled mode - the backend is enabled during demonstrations to manage AWS costs. Files are automatically deleted from S3 after 10 minutes for security and cost optimization.

Quick Start

# Local development
uv sync --link-mode=copy
uv run python run_local.py

# Visit http://localhost:8000/docs for API documentation

Features

• Download satellite tiles as GeoTIFF
• Support for Sentinel-2, Landsat, HLS, MODIS, DEM
• Cloudflare Turnstile bot protection
• Rate limiting (10 req/min per IP)
• Download quotas (5 GB/hour)
• AWS Lambda with up to 1.5GB storage
• Auto-deletion of files (10 min via EventBridge, 1 day S3 lifecycle backup)

Supported Collections

Collection	Bands Available
Sentinel-2 L2A	TCI (True Color), B04 (Red), B03 (Green), B02 (Blue), B08 (NIR), B05 (Red Edge), SCL (Scene Classification)
Landsat Collection 2	Red, Green, Blue, NIR, SWIR16, SWIR22, Thermal, QA
Sentinel-1 GRD	VV, VH (downloads disabled due to file size)
HLS (Harmonized Landsat Sentinel-2)	Red, Green, Blue, NIR, SWIR
MODIS	Surface Reflectance bands
Copernicus DEM	Elevation

API

POST /download

Download satellite tiles as GeoTIFF.

{
  "collection": "sentinel-2-l2a",
  "item_id": "S2A_MSIL2A_20240101T104441_R008_T32UNF",
  "asset_key": "B04",
  "turnstile_token": "<cloudflare_turnstile_token>"
}

GET /health

Health check returning {"status":"healthy","version":"0.1.0"}.

GET /collections

List all supported collections and available bands.

Protection

• Bot Protection: Cloudflare Turnstile verification
• Rate Limiting: 10 requests/minute per IP
• Download Quotas: 5 GB/hour per IP
• File Size Limit: 1.5 GB max
• Auto-Cleanup: Files deleted after 10 minutes (S3 lifecycle backup at 1 day)
• CORS: Restricted to authorized frontend domains only

Local Development

# Install dependencies with UV
uv sync --link-mode=copy

# Run locally
uv run python run_local.py

Or using traditional pip:

# Create virtual environment
python -m venv .venv
.venv\Scripts\activate     # Windows
source .venv/bin/activate  # Linux/Mac

# Install dependencies
pip install -e .

# Run locally
python run_local.py

Deployment

AWS Lambda

# Build and tag
docker build -t sat-data-viewer-backend .
docker tag sat-data-viewer-backend:latest <account>.dkr.ecr.<region>.amazonaws.com/sat-data-viewer-backend:latest

# Push to ECR
aws ecr get-login-password --region <region> | docker login --username AWS --password-stdin <account>.dkr.ecr.<region>.amazonaws.com
docker push <account>.dkr.ecr.<region>.amazonaws.com/sat-data-viewer-backend:latest

# Get platform-specific digest
aws ecr batch-get-image --repository-name sat-data-viewer-backend --region <region> \
  --image-ids imageTag=latest --accepted-media-types "application/vnd.oci.image.index.v1+json" \
  --query 'images[0].imageManifest' --output text

Use the amd64 digest when updating Lambda (not the multi-platform manifest).

Lambda Config:

• Memory: 3008 MB
• Timeout: 180s
• Ephemeral storage: 1536 MB
• Environment: TURNSTILE_SECRET_KEY

Function URL CORS:

• Origin: https://kongstad.github.io
• Methods: GET, POST
• Headers: *

Environment Variables

Variable	Default	Description
TURNSTILE_SECRET_KEY	""	Cloudflare Turnstile secret
ALLOWED_ORIGINS	*	CORS origins (comma-separated)
MAX_FILE_SIZE_MB	1500	Max file size limit

Tech Stack

• FastAPI - Web framework with async support
• Mangum - ASGI adapter for Lambda
• Rasterio - Geospatial data processing (GDAL)
• Boto3 - AWS S3 uploads and EventBridge scheduling
• HTTPX - Async STAC API client

License

MIT - See LICENSE