Option to merge image descriptions into content for single-field consumers (RAG pipelines)

[khibye]

Many downstream RAG and indexing platforms (for example, Open WebUI's "External" content extraction engine) expect a simple, flat text field per document to embed and index.

Currently, ExtractedDocument cleanly separates the main text into content and pushes image data/descriptions into a separate images list. While this is great for clean data structures, any consumer or pipeline that only reads the primary text field silently loses all the valuable visual context provided by the image descriptions.

Current Workaround: Writing an external adapter/middleware that manually iterates through the document and concatenates document.content with each image's description.

Proposed Solution
I would love to see a native way to get everything as a single, well-formatted string. Two potential approaches come to mind:

merge_image_descriptions: bool in ExtractionConfig
A flag that, when enabled, appends or injects the image descriptions directly into the main text body.

Enhanced output_format="markdown"
Ensure that when Markdown is selected, image captions/descriptions are reliably embedded inline at the exact position the image appeared (e.g., ), rather than just stripping them or isolating them.

Benefits
Out-of-the-box RAG compatibility: Seamless integration with standard indexing engines without requiring custom wrapper code.

Preserved Context: Prevents silent data loss of visual information during ingestion.