Create `plate.zip` that is self contained

[nicpottier]

With the work that @amoghbanta has been doing on ADT Studio it is clear that having an easy way to pass around all the assets that go into packaging an ADT would be useful. Currently we use the plate.json as the source for web generation, but this doesn't include all the images that are also part of the final ADT.

Instead of just having a plate.json we could instead have a plate.zip which contains not just the JSON of the plate but also the other assets the plate references. (namely images right now)

This could be in a similar directory structure as webpubs:

  /plate.zip
    - plate.json
    - cover.png
    - /images
        - img_p1_r0.png
        - img_p2_r1.png
        - ...

This could then be passed to ADT Studio as a self contained description of what was extracted from the PDF. After edits, ADT Studio could repackage the zip in the same format. adt-press could then "package" the ADT from the plate zip into the desired languages with TTS generated in all the package format we support (ADT, webpub, epub?)

Thoughts @amoghbanta ?

[amoghbanta]

Hey @nicpottier , totally aligned with this. The latest ADT Studio already treats everything as a single archive, so we’re effectively in the workflow you describe:

• Uploads: Studio accepts the same plate.zip bundle you outlined (root folder + plate.json, images/…, cover, audio, config). We hydrate from that archive client-side (no server calls) and keep the path structure intact.
• Editing: Images/text/glossary edits all happen locally. Image replacement swaps the actual file in the archive (PNG, ≤ 3 MB), so there’s no drift between JSON and assets.
• Downloads: The Download updated ZIP CTA repackages the exact structure we loaded (or, for JSON-only uploads, creates the slugged folder once). The ZIP filename carries a timestamp, but the contents mirror the original layout, so you can drop it straight back into ADT Press for packaging/TTS without any manual reshuffling.

So yes — ADT Studio already treats plate.zip as the canonical interchange format between extraction and Press.

For bulk asset regeneration back in press, I’m leaning towards signaling this directly in plate.json, e.g.:

{
  "title": "ChessProject",
  "language_code": "en",
  "sections": [
    {
      "regenerate": "complete_section",
      "section_id": "sec_p1_s0",
      "section_type": "separator",
      "page_image_upath": "output/openChess/images/img_p1.png",
      "part_ids": [
        "txt_p1_t0",
        "txt_p1_t1",
        "img_p1_r2"
      ],
      "explanation_id": "sec_p1_s0_eli5",
      "explanation": "some explanation goes here",
      "easy_read": "These boxes show the main ideas.",
      "glossary": [
        {
          "word": "concept box",
          "variations": [
            "concept box",
            "concept boxes"
          ],
          "definition": "A special box in a textbook that highlights and explains an important idea.",
          "emojis": [
            "💡",
            "📦",
            "📝"
          ]
        }
      ],
      "background_color": "#000000",
      "text_color": "#ffffff",
      "layout_type": "single_column"
    },
    {
      "regenerate": "tts",
      "section_id": "sec_p2_s0",
      "section_type": "text_and_images",
      "page_image_upath": "output/openChess/images/img_p2.png",
      "part_ids": [
        "txt_p2_t0",
        "img_p2_r1",
        "txt_p2_t1",
        "txt_p2_t7",
        "txt_p2_t8"
      ],
      "explanation_id": "sec_p2_s0_eli5",
      "explanation": "some explanation goes here",
      "easy_read": "Let's build together!",
      "glossary": [
        {
          "word": "concept",
          "variations": [
            "concept",
            "concepts"
          ],
          "definition": "An idea or way of thinking about something.",
          "emojis": [
            "💡",
            "🧠"
          ]
        }
      ],
      "background_color": "#000000",
      "text_color": "#ffffff",
      "layout_type": "single_column"
    }
  ]
}