HTTP API for pdfparse.

GET  /ping                     health check
POST /parse?format=json|markdown|text&pages=1-3&tables=0
POST /summarize?pages=1-3      summary via Amazon Bedrock

Request bodies are raw PDF bytes, e.g.
  curl --data-binary @file.pdf -H 'Content-Type: application/pdf' $URL/parse

AWS credentials come from the environment's default chain (the ECS task role
in production); nothing is configured in code.