Mocks

Use them from any viewer, test suite or prototype. Publish a new mock the same way when it is complex enough to reuse: a generator in docs/mocks/<name>/, its files written to public/mocks/<name>/ by npm run build:mocks, and a section on this page.

Content Search 2.0: a newspaper issue

Two pages of the Berliner Tageblatt of February 16, 1925, with their OCR and a IIIF Content Search 2.0 service. The OCR is the newspaper’s own (Fraktur, errors included), with a position for every word, so search hits and text land where the print is.

All under https://clover.iiif.library.northwestern.edu.

What is in it

• The Manifest has two Canvases (3602 × 5000 and 3536 × 4999), drawn from the IIIF Image API reference server, and a SearchService2. Each Canvas references its OCR by line in annotations.
• OCR pages are AnnotationPages of supplementing annotations with a TextualBody (language: "de") targeting the region it was read from, canvas#xywh=x,y,w,h, in the Canvas’s pixels. Each carries a textGranularity of line (about 310 a page) or word (about 2,500 a page). The Manifest references only the lines; the words are there for testing word-level highlighting.
• Search answers follow Content Search 2.0:
  • items are highlighting annotations on each match’s OCR box, a SpecificResource on the Canvas (with partOf naming the Manifest) and an xywh= FragmentSelector.
  • annotations holds a page of contextualizing annotations, one per hit, quoting it with a TextQuoteSelector: prefix, exact and suffix, six words either side.
  • Results come 25 to a page, in an AnnotationCollection (partOf with total, first and last), with startIndex, prev and next.
• Matching ignores case, accents and punctuation, and reads the Fraktur long s (ſ) as s. A word matches if it contains the query, so berlin finds “Berliner”; a phrase matches consecutive words, its box the union of theirs.

Example answers

Static files, for tests that need a fixed answer. Each is exactly what the live service returns for the query, except that page links point at the next file.

Searching it live

A static site cannot read ?q=, so the published search service URL answers nothing by itself. Two ways to search any query:

• npm run dev serves a live copy of the whole mock, with every id under the dev server: http://localhost:3000/api/mocks/content-search/manifest.json. Its service answers any q and page, with Access-Control-Allow-Origin: * so a viewer on another port can use it.

  curl "http://localhost:3000/api/mocks/content-search/search?q=moskau"

  Copy

• mockFetch, from the generator, is a fetch that answers the published mock’s URLs in the browser and passes everything else on. Hand it to anything that takes a fetch. The Viewer’s Content Search demo does:

  import { mockFetch } from "docs/mocks/content-search/index.mjs";
   
  <Clover.Viewer
    iiifContent="https://clover.iiif.library.northwestern.edu/mocks/content-search/manifest.json"
    request={{ fetch: mockFetch() }}
  />;

  Copy

  <script type="module">
    import { mockFetch } from "docs/mocks/content-search/index.mjs";
   
    document.querySelector("clover-viewer").request = { fetch: mockFetch() };
  </script>
   
  <clover-viewer
    iiif-content="https://clover.iiif.library.northwestern.edu/mocks/content-search/manifest.json"
  ></clover-viewer>

  Copy

Changing it

The generator is docs/mocks/content-search/index.mjs; the OCR is docs/mocks/content-search/ocr.json (words in reading order, [text, x, y, w, h]). After changing either, run npm run build:mocks and commit public/mocks/. The OCR pages are written compact, and Prettier leaves public/mocks/ alone.

Georeference: a township map

Northwestern University Libraries’ scan of the Map of the Township of Evanston, Cook County, Illinois, as a Manifest of one Canvas, with:

• the Georeference Extension  annotation its Meadow authoring tool stored (three control points, the whole image as its mask, a first-order polynomial transformation), on the Canvas’s annotation page;
• a navPlace  point naming Evanston, Illinois;
• the image from Northwestern’s IIIF Image API (level 2, 7337 × 9833, CORS open).

It answers no query, so it is static only. The Map page draws it, warped onto the map, and in a Viewer’s Map tab. Its generator is docs/mocks/georeference/index.mjs.

Tagging: a photogravure

Northwestern University Libraries’ record of Edward S. Curtis’s The Vanishing Race – Navaho (1904), plate 1 of The North American Indian, as a Manifest of one Canvas, with:

• the four bits of text printed on the plate tagged where they sit, as Cookbook recipe 0021  tags a fountain: an Annotation with the motivation tagging, its body the words (a TextualBody in English), its target a box on the Canvas;
• the record’s Caption as the Canvas’s summary, no longer in the Manifest’s metadata;
• everything else as Northwestern’s Digital Collections API published it on 14 October 2026, its Cultural Context note included, apart from the ids of the Manifest and its Canvas, which are the mock’s own;
• the image from Northwestern’s IIIF Image API (level 2, 4290 × 3528, CORS open).

It answers no query, so it is static only. The homepage opens on it. Its generator is docs/mocks/tagging/index.mjs, and the record as fetched is docs/mocks/tagging/source.json: to bring the record up to date, fetch https://api.dc.library.northwestern.edu/api/v2/works/46dc28d1-9d59-4160-a511-72418f70cffc?as=iiif into source.json again and run npm run build:mocks.

navPlace: a photograph’s corner

Northwestern University Libraries’ record of a Ronald J. Sullivan slide, Chicago bus 516 on route 151 at Michigan Avenue and Washington Street (July 3, 1955), as a Manifest of one Canvas, with:

• a navPlace  point at that corner, [-87.6245, 41.8833], labeled “Michigan Avenue and Washington Street, Chicago, Illinois”. The record carries no coordinates: the place is read from its summary, and the photograph agrees (it looks north up Michigan Avenue, the new Prudential Building on the right, the Coca-Cola clock and the Blatz sign on the left);
• everything else as Northwestern’s Digital Collections API published it on 14 October 2026, apart from the ids of the Manifest and its Canvas, which are the mock’s own;
• the image from Northwestern’s IIIF Image API (level 2, 3968 × 2685, CORS open).

It answers no query, so it is static only. The homepage’s playground shows it as Photo, its place in a Viewer’s Map tab. Its generator is docs/mocks/navplace/index.mjs, and the record as fetched is docs/mocks/navplace/source.json: to bring the record up to date, fetch https://api.dc.library.northwestern.edu/api/v2/works/24157482-0c2c-4907-8d25-7905adc3cfd9?as=iiif into source.json again and run npm run build:mocks.

Hindenburg: a broadcast and its box

Northwestern University Libraries’ record of WLS-AM’s broadcast of Herbert Morrison and Charlie Nehlsen’s recording of the Hindenburg disaster (May 1937), from the Charles Deering McCormick Library of Special Collections: a Manifest of a Canvas of sound, the reel’s, then four photographs, the front and back of the reel’s enclosure and of its carrier, with:

• an accompanyingCanvas on the Canvas of sound, which the record does not carry: the enclosure’s back, the box the reel came in, where its label is written (the one the record’s Notes quote), painted with the record’s own image of it, so a player shows the box while the broadcast plays;
• no width and height on the Canvas of sound and its sound, and no duration on the photographs: the API gives them stand-ins (100 × 100, and 1 second), which sound and stills do not have, so the Canvas of sound is a Timeline once upgraded to Presentation 4;
• English captions, a supplementing WebVTT file (captions.vtt). The record has no transcript, so the words were transcribed from the broadcast by speech recognition (Whisper) and corrected by hand: the names from the record and the broadcast itself (Herbert Morrison, Charles Nehlsen, WLS, the Prairie Farmer station, Lakehurst), and those of the airship’s commanders and survivors as history gives them. A word that could not be told is [unclear], and so is a passenger’s answer (10:25–10:54). They have not yet been checked word for word by a listener;
• nine chapters (structures), one for each part of the broadcast (the studio’s introduction, Morrison asked about the scene, the landing, the fire, his return to the wreck, the survivors he meets, the count of the saved, the studio’s close), each titled with the broadcast’s own words where it begins: “It burst into flames!”;
• a page of annotations, “The broadcast”, eight describing notes, each on a stretch of the sound (t=start,end) and saying what happens there;
• everything else as Northwestern’s Digital Collections API published it on October 24, 2026, apart from the ids of the Manifest and its Canvases, which are the mock’s own;
• the sound from Northwestern’s HLS stream (14 minutes), and the images from its IIIF Image API (level 2, CORS open).

It answers no query. In npm run dev it is live under /api/mocks/hindenburg/, so its captions answer before they are deployed. The homepage’s playground shows it as Recording, and Timeline’s page plays it. Its generator is docs/mocks/hindenburg/index.mjs, its captions’ cues docs/mocks/hindenburg/captions.json, and the record as fetched is docs/mocks/hindenburg/source.json: to bring the record up to date, fetch https://api.dc.library.northwestern.edu/api/v2/works/f184ccf3-4ee6-4b64-bb7b-a87bb9d15254?as=iiif into source.json again and run npm run build:mocks.

Paged: a catalog in spreads

Northwestern University Libraries’ record of the White Sewing Machine Company’s Portfolio of illustrations showing the White bicycles : then and now (1896), a paged Manifest of 31 Canvases (its covers and 29 pages), with:

• a summary, which the record has none of, written from its own fields (Publisher, Location, Date, the illustrators its Alternate Title names, Materials) and its pages, in under 30 words: “The White Sewing Machine Company of Cleveland issued this bicycle catalog in 1896. Color plates by Thulstrup, Mente, Veenfliet, and Clarke pair “Then” with “Now”; specifications describe four models.”;
• behavior: ["facing-pages"] on Page 15. It is one image of a whole opening, a color plate across both pages (9423 × 4694, twice a page’s width), so it is shown alone and the pages after it pair as the book’s openings do. The spreads run Front Cover, 1–2 … 13–14, 15, 16–17 … 28–29, Back Cover. As published, the plate pairs with Page 16, and every opening after it is split;
• structures, which the record has none of: a Range for each of the catalog’s sections, titled exactly as the catalog has it, by the heading printed on its page (its words, without the closing period), or, for a cover, by the Canvas’s own label. Front Cover; Introduction; Specifications “White” 1896 Model A, B, C and D, each with its engraving; “White” Bearings (Crank Bearing, The Hubs); Important Features of the “White”, six headings, where those that share a page each name their region of it (#xywh=); Guaranty on “White” Bicycles; Tire Guarantees; Terms of Sale; Back Cover;
• on each plate that pairs a “Then” with a “Now” (Pages 4, 6, 18, 20, 22, 24 and 26, and the Back Cover), the two tagged where they are, as Cookbook recipe 0021  tags a fountain: “Then” around the inset scene and the word printed by it, “Now” around most of the picture, clear of its edges (on the Back Cover, the round scene below), each after the printed word described briefly, as a cataloger would (“Now – Postman on bicycle greets woman in period style white dress on bicycle.”). Page 15 prints “Now” with no “Then”, so it has none;
• its words, read by machine with Apple’s Vision framework and not corrected by hand (docs/mocks/paged/ocr.json): each page’s OCR by line and by word, supplementing text on the region each was read from, with a textGranularity, the lines referenced from each Canvas that has words (“OCR (lines)”). Running text reads well; ornamental lettering (the covers’ script, a plate’s signature) and some fractions do not;
• a IIIF Content Search 2.0 service (SearchService2), searchable by word or phrase, as the newspaper’s is: each hit a highlighting annotation on its words’ box, with the words around it, 25 to a page of results;
• everything else as Northwestern’s Digital Collections API published it on 14 October 2026, apart from the ids of the Manifest and its Canvases, which are the mock’s own;
• the images from Northwestern’s IIIF Image API (level 2, CORS open).

The published copy cannot read a query, so the docs answer the search in the browser (docs/lib/mock-search.ts), and npm run dev answers any query live. The homepage’s playground shows it as Catalog. To read its words again, run npm run build:paged-ocr (macOS: scripts/ocr-vision.swift), which reads each page at a size its info.json lists and keeps no image, then npm run build:mocks. Its generator is docs/mocks/paged/index.mjs, and the record as fetched is docs/mocks/paged/source.json: to bring the record up to date, fetch https://api.dc.library.northwestern.edu/api/v2/works/386ff30f-1975-4d9c-9905-ec76381d6992?as=iiif into source.json again and run npm run build:mocks. The package’s tests use a copy, in packages/clover/src/fixtures/mocks/paged/; copy the published file there again after a change.

Transcripts: a film’s captions and its transcript

The IIIF Cookbook’s recipe 0219 , Lunchroom Manners, a Manifest of one video Canvas (9 min 32 s), with:

• its English captions as the recipe publishes them, a supplementing WebVTT file labeled “Captions in WebVTT format”;
• the film’s transcript as plain text beside them, as the Canvas’s rendering, which is how recipe 0017  gives a transcript. The file is the one recipe 0253  (Presentation 4) provides for the same film, with that recipe’s label, “Transcript in plain text format”, and its language, en;
• everything else as recipe 0219 publishes it, apart from the ids of the Manifest, its Canvas and their annotations, which are the mock’s own;
• the video, the captions and the transcript from the Cookbook’s own fixtures, at fixtures.iiif.io (CORS open).

So a viewer has both kinds of transcript to offer, the timed captions and the untimed text: Clover’s Transcript lists them as “EN” and “TXT”.

It answers no query, so it is static only. Its generator is docs/mocks/transcripts/index.mjs, and recipe 0219’s Manifest as fetched is docs/mocks/transcripts/source.json: to bring it up to date, fetch https://iiif.io/api/cookbook/recipe/0219-using-caption-file/manifest.json into source.json again and run npm run build:mocks. The package’s tests use a copy, in packages/clover/src/fixtures/mocks/transcripts/; copy the published file there again after a change.

Rudolf Hess at Nuremberg: people followed through a film

Sentencing of Nazi Leaders at Nuremberg, a reel of the International Military Tribunal’s judgment from the US National Archives (National Archives Identifier 23769, local identifier 111-ADC-10005), as Wikimedia Commons  publishes it, made to show Rudolf Hess’s distress in the dock beside the other defendants: Hess is marked wherever he is in view, and so are the defendants around him at that end of the dock, wherever each can be told: Hermann Göring and Joachim von Ribbentrop either side of him and Wilhelm Keitel beyond Ribbentrop, in the front row, and Erich Raeder, Baldur von Schirach and Fritz Sauckel in the row behind. Commons is not a IIIF publisher, so the Manifest is the mock’s own, made from the file’s Commons record (read on October 19, 2026):

• The Manifest (Presentation 3) is labeled “Rudolf Hess at Nuremberg Sentencing” (its dates are in its metadata), and its summary says what it shows of him. Its metadata are the record’s: the National Archives’ title (the reel’s, which the Canvas keeps as its label) and description among them. Its rights are the public domain, the film being the US Army Signal Corps’. Its homepages are the Commons page and the National Archives Catalog’s record.
• Its date is the two days of the judgment the film shows, as its navDate (30 September 1946) and a Date line. The Commons record’s only date is 1947-10, a year after the trial, so the days are read from the film against the trial’s record (30 September , 1 October ). The soundtrack carries the reading of 30 September up to its last words, on the SS and the General Staff, and the Tribunal adjourning “until 1 October 1946 at 0930 hours” (5:21); from 5:46 it carries the verdicts of 1 October, over the courtroom and then the dock. Hess is taken ill and led out of the dock on 30 September, during the reading on the murder and ill-treatment of the civilian population, if the film’s sound was recorded with its pictures (the record does not say when). Despite its title, the reel shows no sentencing: the sentences were pronounced on the afternoon of 1 October.
• Its one Canvas is 654 × 480 and 671.164 s, the film as shown (stored 720 × 480, at a display aspect of 654:480). It paints the original WebM file (VP8 and Vorbis, 133 MB), and its placeholderCanvas is the still Commons cuts from the file at 14 s (500 × 367, Commons cuts stills at its own widths only).
• Two AnnotationPages. “Rudolf Hess” holds three annotations, one for each part of what the film shows of him, each identifying and describing: “Hess shows distress” (0:12–1:25), “Hess is led out of the dock” (1:25–1:40) and “Hess returns to the dock” (3:18 to the end). “Other Defendants” holds an identifying annotation for each of the others followed, named by him, in the dock’s order: Göring, Ribbentrop and Keitel, then Raeder, Schirach and Sauckel. Each was told by his face, against a photograph of the dock, and by where he sat. Karl Dönitz, who sat behind Göring, is not marked: the close shots leave him out of frame, and in the wide shot from 1:25.7 and the pan from 3:18 the back row is too small and soft to tell one man from another, so none of the back row is marked there (nor Keitel in the wide shot). Nor is Keitel in the close shot from 0:45, where only half his face is in frame and the man behind him is hard to tell from him.
• Its English captions, a supplementing WebVTT file (captions.vtt, 99 cues), as Cookbook recipe 0219  gives one. The soundtrack is the courtroom’s interpretation system, which cuts between its channels, so the captions follow: the judgment read in English, in the published judgment’s words, checked against the soundtrack (where the record’s two printings differ, as on Göring’s “16th July, 1941” and “desettling”, the one the soundtrack agrees with); the Russian interpretation (1:57–3:16, the euthanasia killings, Bach-Zelewski’s testimony and the persecution of the Jews; 7:02–7:25, the end of the verdict on Hess) and the French (9:17–9:40, the end of Keitel’s) in the record’s English, marked “[In Russian]” or “[In French]”; and a voice over the sound system between the days (“Hello, sound recording…”), as heard. An ellipsis marks speech cut by the film’s editing. The soundtrack was transcribed with Whisper (small and medium.en) only to place and check the words; none of its own words are used where the record has them.
• Its transcript, the same words as plain text (transcript.txt), the Canvas’s rendering, as Cookbook recipe 0017  gives one, labeled “Transcript in plain text format” (recipe 0253’s words for a film’s). It has a paragraph for each stretch of the reading the film keeps whole: one starts after a pause, at a cut, where the voice or language changes, and where the editing cuts the speech. Each paragraph of a stretch in Russian or French opens with its note, which the captions give once.
• Chapters, a Range for each part of the film the National Archives’ description names, titled with its words (each cut to the clause that names the shot) and timed to the film’s cuts: “Sitting in prisoners’ dock listening to testimony” (0:12.5), “Hess taken ill leaves prisoners’ dock” (1:25.6), “Shots of English and US prosecution tables” (1:57.2), “Scenes of Justice Kravechenko, Russian Justice, reading statement from bench” (2:38.1), “Pan of prisoners’ dock and courtroom” (3:17.8), “Court adjourns” (4:57.5), “MS of Erich Raeder, Fritz Sauckel, Hess, Goering, von Ribbentrop, in dock” (6:19.7), “Justice Francis Biddle in speaking of von Ribbentrop” (7:53.4) and “Camera on prisoners’ dock as Justice Lawrence speaks about the defendant, Ernst Kaltenbrunner” (8:48.2, to 10:00.4). Each is a SpecificResource on the Canvas with a t=start,end FragmentSelector.
• Each annotation’s targets are the stretches of the film he is in view, each a SpecificResource on the Canvas with a FragmentSelector of both Media Fragment dimensions, xywh=x,y,w,h&t=start,end: the region of his head and shoulders (Hess standing, down to the rail) through that stretch, in the Canvas’s pixels. Stretches that meet in time are one movement; a gap between two is time he is not in view. There are 1,187 in all: Hess 230, Göring 207, Ribbentrop 239, Keitel 71, Raeder 89, Schirach 145 and Sauckel 206.

The regions were found with OpenCV’s CSRT tracker on Commons’ 480p transcode (the same 654 × 480 frame), each shot begun from a region placed by eye, each box kept at its size, and the tracker’s movement held to regions read by eye where it wandered: every 2.5 s for Hess in the close shot from 0:45, where its boxes grew onto Ribbentrop, and every few seconds for Göring from 6:28, when a man walking past the camera threw it. For Hess in the wide shot, where it lost him as he stood, the regions were read by eye every half second; Ribbentrop there was followed from the frame Hess uncovers him, checked every quarter second as Hess passes him again on his way out. The others were followed the same way, each stretch ending where he leaves the frame or the tracker could no longer hold him; where it lost Schirach and Sauckel as the officer leans across the dock (0:32–0:35), their regions were read by eye. The regions were then thinned to the fewest stretches that keep within 3 px of every frame, and each shot’s frames were checked with its regions drawn on them. They are as good as the film allows, which is soft and dark: about a head’s width either way.

In npm run dev it is live too, at /api/mocks/rudolf-hess-nuremberg/manifest.json, with ids under the dev server, so the captions and the transcript it names are files the dev server answers (the published ones exist only once the site is deployed). The dev server gives that Manifest at the published path too, /mocks/rudolf-hess-nuremberg/manifest.json, as it does every mock with a live copy, so the homepage and the demos open it. Its generator is docs/mocks/rudolf-hess-nuremberg/index.mjs; the captions are docs/mocks/rudolf-hess-nuremberg/captions.json, one row a cue, [start, end, text], which captions() writes as WebVTT and transcript() as plain text, and the regions are docs/mocks/rudolf-hess-nuremberg/tracks.json: each page, its annotations (with their motivations), and each annotation’s stretches, one row a stretch, [start, end, x, y, w, h]. To follow another defendant, add an annotation to “Other Defendants” there and run npm run build:mocks. The video and the still are Wikimedia’s (CORS open). The package’s tests use a copy, in packages/clover/src/fixtures/mocks/rudolf-hess-nuremberg/; copy the published files there again after a change.

Page transcript: a newspaper’s pages as text

The newspaper issue of the Content Search mock (the Berliner Tageblatt of February 16, 1925, two pages), with each page’s transcript as plain text: a text/plain file in the Canvas’s rendering, labeled “Text (OCR)” and in German (de), as libraries publish a page’s OCR text beside its image. It has no search service and no OCR annotations, so the text files are the one way to the words.

Each file is the page’s own OCR, as the Content Search mock reads it: its lines in reading order, a line of the page a line of the file, with a blank line where the next line jumps up a column or steps down more than 1.6 times the page’s usual line pitch (a new block). Nothing in it is typed by hand, so it reads as the OCR does, long s (ſ) and misreadings included.

In npm run dev it is live too, at /api/mocks/page-transcript/manifest.json, with ids under the dev server, so the text files it names are ones the dev server answers (the published ones exist only once the site is deployed). Its generator is docs/mocks/page-transcript/index.mjs, which reads the Content Search mock’s OCR (docs/mocks/content-search/ocr.json): run npm run build:mocks after changing either. The package’s tests use a copy, in packages/clover/src/fixtures/mocks/page-transcript/; copy the published files there again after a change.

Standalone canvas: a map on its own

A Canvas published on its own, not in a Manifest: the one Canvas of Northwestern University Libraries’ record of Bartolomeo Borghi’s map of the British Isles, L’Isole Britanniche, as Clover 3.x’s site publishes it (standalone-canvas.json, fetched on October 19, 2026), with:

• ids under the mock’s own address for the Canvas, its painting page and annotation; its partOf still names the record’s Manifest, from which a viewer can open it;
• @context for its context key, a slip in the published file;
• a metadata line it does not carry, its Contributor, as the record’s Manifest publishes it: “Borghi, Bartolomeo, 1750-1821 (Cartographer)”, “Costa, Agostino (Engraver)”, “Borghi, Bartolomeo, 1750-1821 (Contributor)”;
• an annotation page of its own, tagging Ireland where the map draws it (xywh=600,4160,2180,2800, read from the image), by the name the map prints there, “Irlanda”, in Italian, as the map is;
• everything else as published, its label, summary, thumbnail and placeholder included;
• the image from Northwestern’s IIIF Image API (level 2, CORS open).

It answers no query, so it is static only. The Canvas page shows it, loaded by a Canvas<clover-canvas> on its own. Its generator is docs/mocks/standalone-canvas/index.mjs, and the Canvas as fetched is docs/mocks/standalone-canvas/source.json: to bring it up to date, fetch the published file into source.json again and run npm run build:mocks.

Cookbook: the IIIF Cookbook’s recipes, curated

A Collection of the IIIF Cookbook ’s recipes, each a member by the resource the Cookbook publishes for it (a Manifest, or a Collection), in the Cookbook’s order, with:

• its label, the recipe’s title as the Cookbook writes it, and its summary, the name of its folder on iiif.io (“0047-homepage”), in no language (none);
• its thumbnail: the resource’s own, else its first Canvas’s (or a Collection’s first member’s), else its first image by its image service, at a size the service’s info.json allows. A recipe that publishes no image has none;
• its homepage, the recipe’s page on iiif.io.

A resource is listed once, under the recipe it belongs to: 0466 and 0599, which open 0001’s and 0006’s Manifests, are left out. The Collection’s own words, and what it leaves out and why (LEFT_OUT), are curated in its generator.

It answers no query, so it is static only. The homepage reads it twice: its Viewing field lists the members under “IIIF Cookbook recipes”, and its Natural story’s select (the same field) lists them alone, each opening in the playground’s Viewer. Its generator is docs/mocks/cookbook/index.mjs, and what it lists is docs/mocks/cookbook/recipes.json, written by npm run build:cookbook, which lists only the recipes on the Cookbook’s viewer matrix , reads each one’s resource from iiif.io (leaving out, and saying, one that no longer answers) and says which recipes iiif.io’s index or matrix lists that the docs’ own list (cookbookSections.ts) does not. When the Cookbook changes, run npm run build:cookbook, then npm run build:mocks.

Esu Shrine: a 3D model

Northwestern University’s Material History Lab’s scan of an Esu shrine, an altar piece sculpture of Esu collected in Ibadan, Oyo State, Nigeria, in 2012 (catalog number IBD12-AD-2-1), as a Presentation 4 Manifest of one Scene. The Lab publishes no IIIF for it: its object page  shows the model in Smithsonian Voyager, from a Voyager scene that names the model, a GLB. The Manifest has:

• its label, its metadata and its homepage from the object page as published on October 23, 2026: its title, and its Object Data word for word (Catalog Number, Collection Location, Year Collected, Object Type, Country, Region, Place, Culture, Time Period, Material(s), Height, Width, Length);
• a summary written from those fields, as the page has none;
• its attribution the Voyager scene’s copyright, ”© 2025 Northwestern University”;
• its thumbnail the first of the object page’s photographs of the shrine, as the page links it (602 by 1024 pixels), linked, never copied: the Voyager scene’s renderings of the model show its editor’s guides, a box drawn around it;
• one Scene painting the model at its origin, turned −39.52° about its upright axis (a RotateTransform), so a viewer’s default camera, which looks from the front, opens on the view the Lab’s Voyager scene opens on, and no camera or light, which the viewer provides;
• the model, the Lab’s GLB (esushrine_11.17.glb there; 34.5 MB, glTF binary), kept beside the Manifest as model.glb, as it came: the Lab’s server sends no CORS headers, so a viewer on another site could not fetch it from there.

The homepage’s playground shows it as Model. Its generator is docs/mocks/esu-shrine/index.mjs, and the Voyager scene as fetched is docs/mocks/esu-shrine/source.json: to bring it up to date, fetch https://materialhistorylab.mtiapp.northwestern.edu/voyager/scenes/IBD12-AD-2-1/voyager-scene/scene.svx.json into source.json again, delete public/mocks/esu-shrine/model.glb, and run npm run build:mocks, which fetches the model from the Lab when it is missing.

Rights

The newspaper’s scans and text, in the Content Search and Page transcript mocks, are the Berliner Tageblatt, Staatsbibliothek zu Berlin – Preußischer Kulturbesitz, via Europeana, under the Public Domain Mark . The Evanston map is Northwestern University Libraries’. The Curtis plate is in the public domain in the United States (No Copyright – United States); its record and image are Northwestern University Libraries’, from the Charles Deering McCormick Library of Special Collections, and keep their attribution. The Sullivan slide is In Copyright, its rights-holders unlocatable or unidentifiable; its record and image are from Northwestern University Libraries’ Transportation Library, with the attribution the record gives. The White catalog is in the public domain in the United States (No Copyright – United States); its record and images are Northwestern University Libraries’, from the Transportation Library, and keep their attribution. Lunchroom Manners, its captions and its transcript are the IIIF Cookbook’s fixtures, as recipes 0219 and 0253 publish them. Borghi’s map is in the public domain in the United States (No Copyright – United States); its record and image are Northwestern University Libraries’, from the Charles Deering McCormick Library of Special Collections, and keep their attribution. The Esu shrine’s model and its photograph are Northwestern University’s, from its Material History Lab: the model (© 2025 Northwestern University) kept here as the Lab serves it, the photograph (on a page © 2026 Northwestern University) linked from there, never copied. The Cookbook Collection’s members are the IIIF Cookbook’s own resources, linked, never copied, and their thumbnails are asked of the servers they name. The mocks’ structure is Clover’s, under its license.