Kinds of material

Every site shows photographs, scans and other images. Three other kinds of material are options, chosen when the site is set up, so that a collection installs only the packages it needs:

npm run setup                      # asks what the collection holds besides images
npm run setup -- --with=pdf,3d     # or says so
npm run setup -- --with=none       # images only

Options combine: list as many as the collection holds, separated by commas. Setup writes the choice to config.json as "media", installs the packages of the options chosen, builds the interface and prints the Python packages to install. Run it again to change the choice.

Option	For	Python packages
3d	3D models as .glb files	pip install -r requirements-3d.txt and python -m playwright install chromium
pdf	Born-digital and scanned PDFs	pip install -r requirements-pdf.txt
zoom	Maps and other large images served by IIIF image servers	none

3D models

Models are drawn before indexing. python -m src.render3d render takes a folder of .glb files, or a manifest with a model_url column, and draws views of each model from several sides. It writes a manifest of those views, which python -m src.ingest then indexes with the photographs:

python -m src.render3d render /path/to/models --data-dir data/collections/my-collection
python -m src.ingest data/collections/my-collection/models3d/views.parquet --data-dir data/collections/my-collection

Rendering can be stopped and run again, and it ends by printing the command that indexes the views.

The views become images of their object, so a model is found by what it looks like, and the object’s page opens the model, which visitors can turn and zoom. A shape index of the views also lists other objects of a similar shape. The viewer is loaded only on the page of an object that has a model.

A model kept on the site must be a .glb file. A .gltf file refers to other files, so it can be served only from a web address.

PDFs

python -m src.ingest reads the PDFs in a folder, bucket or manifest page by page. Each page is searched by what it looks like, without OCR, so handwriting, diagrams and forms are found too. The pages of one PDF are shown as one object, and its page opens the whole document in the browser.

An install without the PDF reader indexes the images and leaves the PDFs out, saying how many and what to install.

Deep zoom

An image whose catalog record names an IIIF image service, in a column such as iiif, iiif_url or iiif_service, opens in a deep zoom viewer that loads only the part of the image in view. The column may hold the service’s address, its info.json or any image request to it. If the server does not answer, the plain image is shown instead.

Images kept on an IIIF server need not be copied to the site at all. Visitors’ browsers load them straight from the server, and an index that names only the service lets the site ask it for each size. What the site keeps of each image is chosen when indexing.

In a container

The published image, ghcr.io/l4cch/digital-collections-explorer, shows every kind of material, since it cannot know what a collection holds; the code of each viewer loads only on a page that needs it. An image of your own built with docker build --build-arg MEDIA=pdf . holds only what you list.

Collections set up before version 2

config.json files from version 1 name a collection_type instead. Setup reads it as photographs for images only, maps for zoom and documents for pdf, and writes the new setting.