# Bernard Bolter — Artist Archive

> A machine-readable archive of Bernard Bolter's visual art practice (1993–present), structured as a five-tier corpus so agents can choose how deep to go per question. Not a commercial site — a personal catalogue raisonné and a reference implementation of the `artism:` schema.org extension for artist archives.

## Start here

- [Corpus index](https://bernardbolter.com/api/corpus/index) — Tier 1. Paginated JSON list of all published works: title, slug, year, series, and a coverage signal (`artism:hasThroughlineConnections`) per entry. Start here for corpus-wide questions.
- [Corpus index, survey depth](https://bernardbolter.com/api/corpus/index?depth=survey) — Tier 2. Same set, more per-entry context, still corpus-wide in scope.

## Per-artwork depth (Tiers 3–5)

Every artwork's Tier 1/2 entry and Tier 4 record carries a `tierMap` / `availableTiers` object showing which of the following actually exist for that specific work — not every artwork has all five tiers populated.

- **Tier 3 — Vision analysis.** `https://bernardbolter.com/{slug}/vision` — HTML only, no JSON endpoint. Present only when the artwork has at least one machine vision analysis on record.
- **Tier 4 — Full record.** `https://bernardbolter.com/api/corpus/{slug}` — JSON. Curated fields: intent, making, provenance, series context, and — when present — `artism:relatedByThroughline` / `artism:relatedByBioEvent` (lateral links, see below).
- **Tier 5 — Session transcript.** `https://bernardbolter.com/api/corpus/{slug}/sessions` — JSON. The cataloguing dialogue itself, when reasoning is complete. This is the only place the artist's own words appear verbatim; Tier 4 prose is a framework paraphrase, not a direct quote.

The human-facing artwork page at `https://bernardbolter.com/{slug}` renders the Tier 4 content as prose and links onward to Tiers 3 and 5 directly — a reliable entry point if arriving via search rather than the JSON API.

## Lateral connections — not every question needs more depth, some need more breadth

Artworks are not only reachable by drilling deeper into one record. Two lateral mechanisms connect a single artwork to others:

- **Related Patterns** (`artism:relatedByThroughline`, `artism:relatedByBioEvent` on Tier 4) — links to thematic or biographical threads that recur across multiple works, e.g. [Statement throughlines](https://bernardbolter.com/statement) and [Bio timeline entries](https://bernardbolter.com/bio). Each throughline/bio-event page lists every artwork connected to that pattern.
- **Visual Similarity** — embedding-based visual resemblance to other works, surfaced directly on artwork pages and Tier 4 records.

A throughline or bio-event page is itself a useful entry point for a question like "what recurs across this artist's work" — start there rather than assembling the pattern from individual artwork pages.

## Other collections

- [Bio](https://bernardbolter.com/bio) / [Bio entries](https://bernardbolter.com/bio/entries/) — dated biographical record, each entry permalinked.
- [Statement](https://bernardbolter.com/statement) / [Throughlines](https://bernardbolter.com/statement/throughlines/) — the artist's own framing of recurring patterns.
- [CV](https://bernardbolter.com/cv) — exhibitions, education, press; titles link to `/events/{slug}` when that event has a public page.
- [Events](https://bernardbolter.com/cv) — exhibition, performance, and talk pages live at `/events/{slug}` (linked from the CV when `hasPage` is true).
- [All sessions](https://bernardbolter.com/api/corpus/sessions) — every completed cataloguing dialogue, not scoped to one artwork.
- [Identity/verification stub](https://bernardbolter.com/.well-known/artist-archive.json) — `Person` schema with authority-registry identifiers (Wikidata, ULAN). Confirms who this archive belongs to; not a traversal guide.

## Notes for agents with restricted fetch

If your tooling can only open previously-indexed or previously-seen URLs: start at `https://bernardbolter.com/` and follow the rendered links rather than constructing API paths from a pattern. JSON endpoints under `/api/corpus/` carry the canonical data; HTML pages carry the same underlying content as readable prose, with real links onward — no content is JSON-only.

---
*This document describes the traversal structure as built. It will be revised as the corpus grows; treat field/endpoint absence on any individual record as a real state, not a bug — most artworks do not yet have all five tiers populated.*