Unstructured for Dify

Use Unstructured to turn PDFs, DOCX, PPTX, HTML, images, and scanned documents into clean Markdown, JSON, HTML, or text. The Transform tool can also prepare documents for retrieval-augmented generation with enrichment, chunking, and embeddings.

Tools

Transform document

The recommended tool for new workflows. It connects to the hosted Unstructured Transform MCP server, accepts either a Dify file or a public HTTP(S) URL, waits for the asynchronous job to finish, and returns the transformed output as both content and a downloadable file.

Set the provider fields as follows:

• Server Type:
• API URL:
• API Key: A key from Unstructured Transform

The tool supports , , , and parsing, optional enrichments, chunking, and embedding. See the Transform quickstart for account and usage details.

Local files and inline results are limited to 50 MB. JSON results omit embedded base64 images to keep Dify workflows within the plugin's memory allocation. When a result exceeds the inline limit, use the job ID and reported in the error to retrieve the full output from Unstructured Transform.

Partition

The existing Partition tool remains available for legacy Partition API and self-hosted deployments. Configure its API URL and choose either Local Deployment or Unstructured Official API.

Development

Install dependencies and run the standard Dify plugin entrypoint:

Package the plugin from its parent directory:

Source: langgenius/dify-official-plugins