Images in Open Notebook: ingestion, display, and visual understanding

[dalennear]

I installed the docker file and now upon installation I am trying to upload images and it sees them as text instead of images and will not allow me to view them

How do I fix that so that I can go ahead and upload images. This happens also with any URL it strips out the images... need some help please.

[lfnovo]

Moving this to Ideas and broadening the title so it can serve as the canonical product Discussion for images in Open Notebook.

The original report surfaces several distinct needs that should be designed together but not confused:

1. Ingestion and extraction — accepting standalone image files and extracting useful text or structure from them.
2. Preservation and display — retaining images from uploaded documents and web pages, then showing the relevant visual material in the source experience instead of reducing everything to plain text.
3. Visual understanding — using image-capable LLMs or VLMs to reason about diagrams, screenshots, charts, photographs, and other content that OCR alone cannot understand.

We have exploratory work around Docling, but its image path has not yet been sufficiently tested or released as a supported solution. Docling may help with extraction and OCR; it does not by itself settle how images should be preserved, displayed, cited, embedded, or passed to multimodal models.

The technical and architectural questions are also tracked in #331. Please use this Discussion for concrete image workflows, expected behavior, model/provider constraints, privacy concerns, and examples where text extraction is insufficient. Once the product direction is clear, executable slices will be tracked in Issues.

[lfnovo]

Evidence entry from #1236 (fabricefrebel): pasting an image from the clipboard into a note. The flow described — copy a screenshot, paste into the editor, get prompted for a name, the file is saved under the notebook's data directory and ![name](path) is inserted at the cursor — is how they do it today in Notepad++ via an AutoHotkey macro.

What it implies for this thread: notes would need to carry attachments (today notes are plain markdown in the database, with no storage for note-owned files), and the rendered note view would need to resolve those paths. Same storage/display questions as image ingestion, from the authoring side.

[lfnovo]

Status refresh: images sit in two places in #1315: as evidence (a citation should be able to point at a figure or a page, not only text — question 1 on locators) and as generated output (question 6). The image-paste-into-notes need logged here last week fits the second. Locators are the concrete input: which ones matter in your material?