Search & Browse
A manuscript collection is only as useful as the questions you can ask of it. Archetype turns a corpus into something interrogable: a scholar can move from “show me every charter in this repository” to “show me every instance of this letter-form, written in this position, dated before 1200” without writing a query or leaving the browser. Search is the entry point to the whole platform — every result links back to the image, the hand, the transcription and the annotation that produced it.
What you can search
Section titled “What you can search”An Archetype instance does not offer one flat search box over “documents.” It exposes several distinct, purpose-built collections, each with its own fields, filters and facets. In the current codebase these are:
| Collection | What a scholar searches for |
|---|---|
| Manuscripts (item parts) | Physical volumes and booklets by repository, shelfmark, date, format — and, from published TEI msDesc descriptions, material, script, decoration, seal and place of origin |
| Images | Individual folios/pages, including the script components, features and tags they carry |
| Scribes / Hands | Identified scribes and the hands attributed to manuscripts |
| Graphs | Annotated glyphs — searchable by allograph, character, component, feature and position |
| Texts / Clauses | Transcribed text and its diplomatic sub-units |
| People / Places | Named entities extracted from the transcriptions |
The Graphs collection is what distinguishes palaeographic search from ordinary catalogue search: because every annotated letter records its allograph, its constituent components, their features, and where the glyph sits, you can retrieve script evidence directly — every surviving example of a form — rather than the documents that happen to mention it. See Annotation & TEI for how that evidence is created and the Data model for the vocabulary behind these fields.
Filtering, faceting and browsing
Section titled “Filtering, faceting and browsing”Every collection is browsable through faceted filters computed live from the current result set, so each choice narrows the others and counts always reflect what is actually there. Facets cover repository and city, manuscript type and format, script components and features, named people and places, and editorial status. Palaeographic facets (component → feature) are presented as a collapsible tree, so a broad category can be opened to its specific forms.
Dates are first-class. Results carry a normalised numeric range (date_min/date_max), which powers a decade timeline view — click a bar to filter to that decade — and range filters with precision options (spans “at most” or “at least” a given number of years). A geographic distribution view plots results by repository city, and results can be shown as a table or an image grid. Filter state lives in the URL, so any search is a shareable, citable link, and searches can be saved for reuse.
Full-text and advanced query
Section titled “Full-text and advanced query”Free-text search runs across the fields that matter for each collection (transcribed content, labels, shelfmarks, names). Matching can require all terms or tolerate the last one, and an exact-phrase mode is available. Matched terms are highlighted in results. An advanced query builder composes nested AND/OR groups of conditions (is, is not, greater/less than, between, contains, starts-with, is-empty) over a collection’s filterable fields, and a global suggestion endpoint offers grouped type-ahead across every collection at once — for transcribed text, with keyword-in-context snippets cropped and highlighted around the hit.
How it works
Section titled “How it works”Search is backed by Meilisearch and is registry-driven: a single declarative registry (api/apps/search/registry.py) is the source of truth for every collection — its model, document builder, and its filterable, sortable, facetable and searchable attribute lists. Adding or changing an index is a one-entry edit; nothing about an index lives anywhere else. Indexes are built and kept current by background workers, with superuser-only management actions to reindex, clear or rebuild, plus in-sync statistics (see api/docs/search-operations.md).
Any result set can be exported for downstream research — CSV, JSON, or BibTeX for manuscripts, scribes and hands — either the current page or the full set. Manuscript results can also be staged for side-by-side viewing in the Compare view. For the standards that make exported work portable, see Export, reuse & cite.