How a model sees a page

A model cannot see a page, and Dinoer does not give it an image to look at. The two views below are what a call returns, captured against https://example.com, unedited.

One: as structure

/opt/dinoer/venv/bin/python3 /opt/dinoer/shot.py \
  --url https://example.com --a11y --guide-version 1.6

- heading "Example Domain" [level=1]
- paragraph: This domain is for use in documentation examples without needing permission. Avoid use in operations.
- paragraph:
  - link "Learn more":
    - /url: https://iana.org/domains/example

This is the accessibility tree that Chromium builds for screen readers; Dinoer does not invent it. It carries roles (heading, link), destinations (/url:) and nesting. Chromium loads and lays out the page to build it, but no screenshot is taken.

Two: as text

{"type": "extraire_texte"}

{"extraction_texte": {
  "titre": "Example Domain",
  "texte": "Example Domain\nExample Domain\nThis domain is for use in documentation examples without needing permission. Avoid use in operations.\nLearn more",
  "url": "https://example.com/",
  "date_capture": "2026-08-14T13:23:35+00:00"}}

Same page, same session. The tree says what the page is: one heading, one paragraph, one link. The extracted text says what it says, without markup or noise tags. On an article or a documentation page, it is usually the view to read first.

What is missing

Dinoer returns no screenshot and no numbered overlay: the code that produced them was removed. Why →

If the structure of a page says nothing about an element, such as a canvas or an image without a label, there is nothing else to fall back on. Where perception stops →

Reproduce it

--a11y returns the tree; the extraire_texte action, in the same call, returns the text:

/opt/dinoer/venv/bin/python3 /opt/dinoer/shot.py --url https://example.com --a11y \
  --action '{"type":"extraire_texte"}' --guide-version 1.6