Overview
The Document Data Node lets you upload and process one or more document files and extract content for use in workflows. You can choose how PDFs and images are processed (OCR, vision-language model, or hybrid) and optionally include XML-formatted metadata for citations. Useful for research, reporting, and feeding content into LLM or extraction nodes.Configuration Parameters
-
Upload Files:
Upload one or more documents to process. Supported types include
.pdf,.docx,.txt,.md, images, and other formats supported by the loaders. -
Method:
How to process PDFs and images:
- OCR with layout detection: Fast; best for text-heavy documents.
- Vision-language models (VLMs): Best for complex layouts, images, and diagrams.
- Hybrid: Combines OCR and VLM; good default when you are unsure.
- Include Document Metadata for Citations: When enabled, output includes XML-formatted metadata for citations.
Advanced Settings
Advanced Settings
- [OCR] Strategy: For OCR method—Auto, Fast, or High resolution. Use High resolution for small text or dense layouts.
- [VLM] Vision model: Vision model used when Method is VLM or Hybrid (e.g. GPT-4o).
- Temperature: Only affects VLM-based extraction. Recommendation: use 0 or low (e.g. 0.2) for consistent, repeatable extraction.
- Max Tokens: Maximum output length per page when using VLM (1–100000).
Expected Inputs and Outputs
-
Inputs:
- This node does not require any specific inputs from other nodes in the workflow.
-
Outputs:
- The output will be a string containing the content of the uploaded document. If the metadata option is selected, the output will also include formatted citations.
Use Case Examples
- Research Document Processing: If you are conducting research and have a document that contains valuable information, you can upload this document using the Document Data Node. The extracted content can then be passed to LLM nodes for analysis and reports generation.
- Content Creation: For content creators, this node can be used to load articles or papers and extract relevant sections for reuse in new content. The option to include metadata ensures that sources are properly credited.
- Copy Writing: When working on copywriting projects, you can upload reference documents written by you to this node and feed its output to LLM nodes for generating new content using your writing style.
Error Handling and Troubleshooting
- Missing Document: If you receive an error indicating that no document data was provided, ensure that you have uploaded a file before executing the node.
- Unsupported File Type: If the uploaded file type is not supported, check the allowed file types in the configuration and upload a compatible document.