Support for UI display of figures/images from ingested documents?

[k3llyliuuuuu]

hihi, im currently using onyx v0.17.1 and I couldn’t find support for displaying figures/images from ingested documents in the UI. As far as I can tell, this version doesn’t render images/figures when answering — it only shows text content. Is that correct?

im developing a chatbot and would really like to include inline figure/image rendering (e.g., diagrams or charts) from uploaded documents as part of the bot’s responses.

Does any current or upcoming version support figure/image display from ingested documents?

Thanks in advance!

[pauldev-hub]

Hey @k3llyliuuuuu ,

Your observation is correct — Onyx v0.17.1 does not render figures/images from ingested documents inline in chat responses.
Here's the full picture.
Current state (confirmed)

Onyx's RAG pipeline extracts and indexes text content only from ingested documents. When a document containing charts, diagrams or figures is ingested, the image data is not stored or passed to the LLM response — only the surrounding text context is retrieved and shown.

This is an architectural limitation, not a UI bug. The indexing pipeline at the backend level doesn't extract and store image blobs from document chunks for later retrieval and rendering.

What Onyx does support for images

Two distinct things that are NOT what you're asking about:

Image Generation — generating new images from prompts via image gen models

Code Execution — rendering graphs/charts from data via code interpreter

Neither of these retrieves and displays figures embedded in your uploaded documents.
Workaround for your chatbot use case
If you need inline figure rendering from ingested documents, the practical options right now are:

Option 1 — Use Unstructured.io for ingestion preprocessing

Run documents through Unstructured's API before ingesting into Onyx. It extracts images as separate elements with coordinates/captions, which you can store separately and reference by URL in your chatbot response layer.

Option 2 — Build a custom retrieval layer
Use Onyx's /api/search endpoint to get document chunk references, then fetch the original document server-side, extract the image at the referenced page/position, and render it in your custom chatbot UI.

Option 3 — Switch to a multimodal RAG framework for this use case
If inline figure rendering is core to your chatbot, tools like RagFlow or Kotaemon natively support multimodal document indexing with image extraction and inline display.

Feature request recommendation

This is a valid gap worth opening as a GitHub issue on the Onyx repo — the indexing pipeline would need to: extract images during ingestion → store them in the blob store (MinIO is already there) → return image URLs alongside text chunks in retrieval → render in the chat UI.

👍 If this helped you, please mark it as the answer — it helps others in the community who run into the same issue find the solution faster!