LaraCMS v1.47.2

Documentation

Comment moderation (summary)

CMS registers CommentModerationContextBuilder on Core’s ModerationContextBuilderRegistry; AI resolves it at runtime without importing CMS classes.

Bulk imports

CMS exposes the module-owned cms:import command while Core supplies its common command mechanics. The command accepts only importers implementing the CMS marker contract, so selecting the command also fixes the destination domain. Source access and mapping remain in external plugins; CMS retains its content DTOs, pipeline, upserters, and post-processing.

See docs/IMPORTS.md for options, Naxos examples, dry-run guarantees, and extension rules.

Search indexing

CMS content search indexes relation data for contributors, categories, tags, and locations. Public filters may use the schema-declared dot paths exposed by Core search, for example tags.id, categories.slug, or locations.country.

These filters target indexed relation fields, not arbitrary Eloquent traversal. Core translates them to Elasticsearch nested queries, Typesense nested-field filters, or database whereHas / whereDoesntHave depending on the active search driver.

A content's references (bibliography) are indexed in the content's search document as the references text field: each label and the URL host with the leading www. stripped, so searching a cited source (Nature, nature.com) finds the contents that cite it. Adding, changing or removing a reference re-sends its content. Existing indexes need a rebuild to gain the field: run php artisan scout:sync-mapping then php artisan scout:reindex for Modules\CMS\Models\Content.

Public API access (anonymous reader)

The CMS /api/v1 surface is read-only for the anonymous user once the Core expose_api switch is on. The guest role of the api guard holds select on exactly these CMS entities (the guest role of web holds no permission, so anonymous /app reads are refused):

CMSDatabaseSeeder::defaultGuestApiGrants() grants these and creates the ACLs only when none exists for the role and permission, so a re-seed never overwrites an ACL an operator changed. The categories ACL names scopes (Core's scope filters, Modules/Core/docs/ACL_SYSTEM.md#scope-filters), so the window follows HasValidity::valid() instead of copying it. Contents, references and media stay on column conditions on purpose: search engines never receive a scope node, so a scope in the anonymous contents ACL would be applied only when the hits are reloaded and would skew the totals and page sizes of every anonymous content search, whereas column conditions go to the engine and keep them exact. An installation seeded before scopes keeps its stored column ACL for categories (the seeder never overwrites an ACL), which keeps working. Staff roles keep their web permissions and read contents without a publication window. The publisher role (on web) also reads core_settings (public settings only) and core_entities (CMS types only), because related records obey their own permission and the content screens load them with a content.

A content's cover (the first media of the cover collection) is an appended attribute loaded once per result set under the media permission and ACL of the caller: on the api guard it is the safe media projection (URLs, dimensions, alt text; no storage details), and it is null when the caller may not read media or the ACL hides it. A role that reads contents and holds no vend_media permission sees contents without their media. Anonymous responses are cached per user and shared; CMS contents do not use the response cache, so a content that stops being valid disappears at once. See Modules/Core/docs/rag/API_ACCESS_CONTROL.md.

Graph provider

CMS registers Modules\CMS\Graph\CmsGraphProvider as a Core Graph provider. Core still owns graph routes, traversal, authorization, request validation, and response shape. CMS only contributes content-oriented defaults: contents expand to tags, categories, contributors, and locations when no relations[] are requested; summary fields prefer editorial identifiers such as title, slug, path, status, type, and timestamps; edge labels map CMS relations to names such as tagged_as, categorized_as, contributed_by, and located_at.

Graph relation loading follows Core rules: explicit relations[] win, provider defaults apply only when relations are omitted, and excluded CMS implementation relations such as translations, history, modifications, locks, and media are not graph-traversable.

Insights (map & tag graph)

The map and tag-graph surfaces (dashboard widgets and the facet alternative-selectors) get their data two different ways.

Map — generic CRUD select, no bespoke endpoint. The locations map is served by the ordinary list endpoint (POST app/crud/select/cms/locations) with a contents count aggregate: a column { "name": "contents", "type": "count" } yields a contents_count on each row (via Core's dotless main-model relation-count aggregate), and the place-derived latitude/longitude accessors give the coordinates. The viewport is a between filter on place.latitude/place.longitude; "used" locations are those with a non-zero contents_count (filtered client-side). The count honours the related entity's read ACL, so it never counts contents the viewer cannot see. This is why there is no insights/map/* route — the CRUD layer already does it.

Tag graph — a real endpoint. GET cms/insights/graph/tags (cms.insights.tag-graph) returns the tag co-occurrence adjacency graph as { nodes, edges }: nodes are tags weighted by how many contents use them; edges join two tags (undirected, source < target) weighted by how many contents they share (Modules\CMS\Services\Graph\TagCoOccurrenceService). This cannot be a select — the edges are a self-join over the cms_taggables pivot producing tag pairs, not a per-row relation count. Optional query params minCoOccurrence (default 1) drops weak edges and maxNodes (default 200) caps the busiest tags. Labels resolve from cms_tags_translations, preferring the active locale then the fallback locale, so a tag hidden by the model locale scope still gets a name.

Graph runtime benchmark

CMS includes an opt-in benchmark for Core Graph runtime traversal over realistic content relations. The benchmark is intentionally outside the normal PHPUnit suites and is skipped unless explicitly enabled. Run it when changing Core Graph traversal/search behavior, CMS graph provider defaults, or before reopening the question of materialized graph edges, which were evaluated and not built (see Modules/Core/docs/GRAPH_SYSTEM.md).

CMS_GRAPH_BENCHMARK_ENABLED=true rtk php artisan test --compact Modules/CMS/tests/Benchmark/CmsGraphRuntimeBenchmarkTest.php

Environment variables:

The benchmark reports per-scenario duration, query count, peak memory, node count, edge count, truncation, and ACL filtering. Treat it as performance evidence, not as a normal pass/fail feature test.

License

CMS Module is open-sourced software licensed under the GNU AGPL v3.

Copyright (c) 2025 Marco P. The full licence text is in LICENSE.