DoclingDocling for IBM watsonx
Guides

Overview

Task-oriented guides for Docling for IBM watsonx

These guides show how to do one thing at a time with Docling for IBM watsonx. All processing runs on the service, so you don't need models or GPUs on your side.

You can call the service in several ways:

  • HTTP requests to the API endpoints, from any language
  • Python SDK: pip install docling-client (recommended for Python)
  • Java SDK: ai.docling:docling-serve-client
  • Command line: docling convert-remote

Setup

Regardless of which method you decide on, you must have a Service URL and an API key, personal to your Docling for IBM watsonx account. Please refer to the Quick Start guide for guidance on how to access these.

All examples read your credentials from these environment variables:

export DOCLING_SERVICE_URL="https://api.example.dcls.saas.ibm.com/your-instance-id"
export DOCLING_SERVICE_API_KEY="your-api-key"

{TASK_ID} in the HTTP examples stands for the task ID returned by a conversion request.

Choosing a conversion method

MethodUse whenSourcesResults
Single conversion (convert/source, convert/file)Converting one documentOne web URL or one uploaded fileDownload link
Batch conversion (convert/source/batch)Converting more than one document, or ingesting from storage and content systemsWeb URLs and any source connector (S3, Azure Blob, Google Cloud Storage, SharePoint, Box, …)Download links, a storage target, or chunks streamed to Kafka or a vector database

In short: use single conversion for one document, and the batch endpoint for everything else — even just a handful of documents.

Available guides

On this page