@heroiclands/package-build 23.5.0 → 23.6.0

package/CHANGELOG.md

@@ -1,5 +1,38 @@
 # @heroiclands/package-build
 
+## 23.6.0
+
+### Minor Changes
+
+- `package-build lint` refuses a content or picture path holding a character outside printable ASCII, reporting it as an error at the path. Rename the file using ASCII only; addresses come from the shortcode, so no address changes.
+
+### Patch Changes
+
+**Dictionaries and glossaries**
+
+- A language note lists its words in `data.words`, with the class modifiers, fields and affixes they use; the compiled Item is unchanged.
+- A language names its parent in `data.family` and reads the parent's words as its own; lint reports a parent that is no language or a family that loops.
+- A language states the tag a dictionary prints for it in `data.abbreviation`, such as `Prov.`.
+- Cultures and affiliations list their own glossary terms in `data.glossary`. These terms mark words and ranks common, give respellings, let in single-word titles, and add or replace definitions.
+- A glossary term that names no language takes its affiliation's, a parent affiliation's or its culture's; lint reports a term that ends up with none.
+- Lint warns about an affiliation that belongs to no culture, since its ranks and offices reach no glossary.
+- An `sql` fence sets its rows as a dictionary of words with `layout=dictionary`, or as a glossary of terms with `layout=glossary`: bold headwords under letter headings, a key of only the abbreviations used, numbered senses and citations.
+- One query lists a language's words, including inherited ones, or its common words on any page. One query lists a culture's or an affiliation's glossary with its ranks and offices, or the whole setting's dictionary and glossary.
+- `package-build content-index` also writes the dictionary and the glossary beside the index, for reading with `jq` or an editor.
+
+**Checks and compile**
+
+- The content checks and the compile run several times faster on a large tree.
+- Two notes that pin the same `id` are reported at the second note, naming the first, instead of as dozens of items that duplicate themselves.
+
+## 23.5.1
+
+### Patch Changes
+
+`cdn publish` skips SVGs the CDN already holds, so a re-run uploads only what changed.
+
+`package-build labels check` is gone, because issue labels are defined once for the whole organisation; remove any `lint:labels` script and `.github/labels.yml`. A project created by `package-build init` no longer gets a label registry.
+
 ## 23.5.0
 
 ### Minor Changes

package/bin/content-commands.mjs

@@ -69,6 +69,7 @@ import { checkArtSlotSizes } from "../engine/art-fields.mjs";
 import { lintContentTree } from "../engine/content-lint.mjs";
 import { lintNoteStates } from "../engine/stub-lint.mjs";
 import { lintContentCharset } from "../engine/content-charset.mjs";
+import { lintPathCharset } from "../engine/path-charset.mjs";
 import { lintContentHtml } from "../engine/content-html.mjs";
 import { lintContentIcons } from "../engine/content-icons.mjs";
 import { lintContentImages } from "../engine/content-images.mjs";
@@ -1093,6 +1094,16 @@ function lintCommand() {
                     skipDirectories: config.skipDirectories,
                 });
 
+                // The spelling of every content and picture path, held to
+                // printable ASCII. Walked from the directory entries rather
+                // than the parsed notes, so a picture, which has no note,
+                // is covered by the same rule.
+                const pathCharset = lintPathCharset({
+                    contentBase: root,
+                    assetsBase: config.paths.assets,
+                    skipDirectories: config.skipDirectories,
+                });
+
                 // The names the charset check leaves room for. An icon
                 // the registry does not declare renders as its own literal
                 // text, which is visible but easy to publish, so it is reported
@@ -1175,6 +1186,7 @@ function lintCommand() {
                     ...frontmatter.findings,
                     ...schemaFindings,
                     ...charset.findings,
+                    ...pathCharset.findings,
                     ...icons.findings,
                     ...html.findings,
                     ...images.findings,
@@ -1593,16 +1605,22 @@ function contentIndexCommand() {
         handler: (argv) => {
             try {
                 const config = loadPackConfig();
-                const { file, notes, bytes, full, draft, stub } = emitContentIndex({
-                    config,
-                    ...(argv.root ? { contentBase: argv.root } : {}),
-                    ...(argv.out ? { outDir: argv.out } : {}),
-                });
+                const { file, notes, bytes, full, draft, stub, dictionary, terms } =
+                    emitContentIndex({
+                        config,
+                        ...(argv.root ? { contentBase: argv.root } : {}),
+                        ...(argv.out ? { outDir: argv.out } : {}),
+                    });
                 log.info(
                     `${config.contentPackage} → ${path.relative(process.cwd(), file)} ` +
                         `(${notes} notes — ${full} full, ${draft} draft, ` +
                         `${stub} stub — ${Math.round(bytes / 1024)} KiB)`,
                 );
+                for (const [relation, written] of Object.entries({ dictionary, terms }))
+                    log.info(
+                        `${config.contentPackage} → ${path.relative(process.cwd(), written.file)} ` +
+                            `(${written.rows} ${relation} rows)`,
+                    );
             } catch (err) {
                 reportFailure(err);
                 process.exitCode = 1;

package/bin/package-build.mjs

@@ -114,7 +114,6 @@ import { walkMarkdownTree } from "../engine/helpers.mjs";
 import { reckoningContext } from "../engine/reckoning-markers.mjs";
 import { SCHEMA_ARTIFACT_FILE } from "../engine/foreign-catalog.mjs";
 import { validateLangSource } from "../lang.mjs";
-import { checkLabelRegistry } from "../labels.mjs";
 import { lintYaml } from "../engine/yaml-lint.mjs";
 import { lintChangesetText, lintReleaseText } from "../engine/changelog-lint.mjs";
 import { groupChangelogText } from "../engine/changelog-group.mjs";
@@ -1328,77 +1327,6 @@ function changelogGroup(args) {
     console.log(`package-build: ${files.length} file(s) grouped · ${warnings} warning(s)`);
 }
 
-/**
- * `labels check` — do the machine registry and the documented table agree?
- *
- * `.github/labels.yml` is synced to GitHub and the §3 table in
- * `.github/ISSUE_REPORTING.md` is what a person reads. Neither derives from the
- * other, so either can drift, and nothing notices until an issue is filed
- * against a label that does not exist.
- *
- * Every repository wants this and only the paths ever differed, which is why it
- * is here rather than copied into each as a `utils/` script.
- *
- * @returns {object} The yargs command module.
- */
-function labelsCommand() {
-    return {
-        command: "labels <action>",
-        describe: "Issue-label registry checks",
-        builder: (y) =>
-            y
-                .positional("action", {
-                    choices: ["check"],
-                    describe: "check: the registry and the documented table name the same labels",
-                })
-                .option("registry", {
-                    type: "string",
-                    default: ".github/labels.yml",
-                    describe: "The machine registry synced to GitHub",
-                })
-                .option("doc", {
-                    type: "string",
-                    default: ".github/ISSUE_REPORTING.md",
-                    describe: "The documented reference whose §3 table lists the same labels",
-                }),
-        handler: handler(async (args) => labelsCheck(args)),
-    };
-}
-
-/**
- * Read both faces of the registry and report where they disagree.
- *
- * Findings are reported against the file each belongs to, so a missing row is
- * located in the documentation and a missing entry in the registry, rather than
- * both being attributed to whichever file was read first.
- *
- * @param {object} args - Parsed CLI arguments.
- */
-function labelsCheck(args) {
-    for (const target of [args.registry, args.doc]) {
-        if (!fs.existsSync(target)) die(`labels check: ${target} does not exist.`);
-    }
-    const registryText = fs.readFileSync(args.registry, "utf8");
-    const docText = fs.readFileSync(args.doc, "utf8");
-
-    const { registry, doc, count } = checkLabelRegistry({
-        registryText,
-        docText,
-        docPath: args.doc,
-    });
-
-    const errors =
-        reportFindings(registry, { file: args.registry }) + reportFindings(doc, { file: args.doc });
-    if (errors) {
-        console.error(
-            `\nThe registry and ${args.doc} §3 are edited together — ` +
-                `change one and change the other.`,
-        );
-        process.exit(1);
-    }
-    console.log(`package-build: registry and §3 agree (${count} labels).`);
-}
-
 /**
  * `lang <action>` — the three localization guards.
  *
@@ -1641,7 +1569,7 @@ function cdnPublishCommand() {
                 .option("set", {
                     type: "string",
                     demandOption: true,
-                    describe: "The set file's name under the package's CDN root, e.g. ai-art.json",
+                    describe: "The set file's name under the package's CDN root, e.g. alt-art.json",
                 })
                 .option("name", {
                     type: "string",
@@ -1972,7 +1900,6 @@ registerContentCommands(
         .command(typesCommand())
         .command(schemaCommand())
         .command(langCommand())
-        .command(labelsCommand())
         .command(bumpCommand())
         .command(yamlCommand())
         .command(changelogCommand())

package/docs/api.md

@@ -1052,6 +1052,15 @@ The charset authored content is held to, so a book can choose its face. The pack
 | `checkText`          | `function checkText(text, file)`           | {Array<{file: string, line: number, column: number, severity: "warning", message: string}>} What is wrong, in file order.                                                | Check one file's text against the charset and the normalization rule.      |
 | `lintContentCharset` | `function lintContentCharset(contentBase,` | `{{findings: Array<{file: string, line: number, column: number, severity: "warning", message: string}>, files: number}}` The findings, and how many files produced them. | Walk a content tree and check every authored file in it.                   |
 
+### `engine.pathCharset`
+
+The characters a content or picture path may hold: printable 7-bit ASCII, U+0020 to U+007E. A note's identity comes from its `shortcode`, so a path's spelling changes no address; a non-ASCII name changes only how the file travels between filesystems, shells, archives and CDN keys.
+
+| Export                    | Signature                                                       | Returns                                                                                                                                   | Use it when                                                                                       |
+| ------------------------- | --------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------- |
+| `offendingPathCharacters` | `function offendingPathCharacters(text)`                        | {string[]} The distinct characters outside U+0020 to U+007E, in order of first appearance.                                                | Whether a path, or one component of it, is clean.                                                 |
+| `lintPathCharset`         | `lintPathCharset({ contentBase, assetsBase, skipDirectories })` | `{{findings: Array<{file: string, severity: "error", message: string}>, files: number}}` The findings, and how many entries were checked. | Walk the content tree and the picture roots and report every entry named outside printable ASCII. |
+
 ### `engine.contentIcons`
 
 Naming an interface icon in a note, without drawing it there. The user guide describes Foundry's interface, and it did so by pasting Unicode lookalikes of the icons the sheets actually draw: `☆` for the improve flag, `✎` for the formula editor, `◆` in the success-value table, `★★★` for mastery. The system renders every one of those with **Font Awesome** — a `fa-regular fa-star`, a `fa-solid fa-pen-to-square` — so the note and the screen it describes were drawing different pictures, and drifting apart with every sheet change.

package/docs/authoring/links-and-markup.md

@@ -309,13 +309,18 @@ A column's heading is its SQL alias, so `AS "Name"` is what a reader sees. A cel
 
 ### What a query can read
 
-| Relation          | What it holds                                                                                                                        |
-| ----------------- | ------------------------------------------------------------------------------------------------------------------------------------ |
-| `notes`           | Every note except a stub. This is the one to use unless a stub is wanted.                                                            |
-| `entries`         | Every note, stubs included.                                                                                                          |
-| `events`          | One row per event a `lore`, `place` or `affiliation` note writes in `data.events`, with `whenSort` to order them by date.            |
-| `market`          | The market scale as three columns — `value`, `name`, `trade` — so a table can print `village` beside the number a note wrote.        |
-| `<package>.notes` | The same relations — `notes`, `entries` and `events` — for each package this one declares a dependency on, in a schema named for it. |
+| Relation           | What it holds                                                                                                                                                                                        |
+| ------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `notes`            | Every note except a stub. This is the one to use unless a stub is wanted.                                                                                                                            |
+| `entries`          | Every note, stubs included.                                                                                                                                                                          |
+| `events`           | One row per event a `lore`, `place` or `affiliation` note writes in `data.events`, with `whenSort` to order them by date.                                                                            |
+| `words`            | One row per word a language note writes in `data.words`, with that note as its `language`, its `position` and what it is built from.                                                                 |
+| `glossary`         | One row per entry a culture or an affiliation writes in `data.glossary`, with the `note` it belongs to and the `language` it resolves to.                                                            |
+| `language_lineage` | One row per language note and each language it inherits words from through `data.family`, itself included at `depth` 0. Join it to `words` for a language's words with inherited ones.               |
+| `dictionary`       | One row per word per language it is listed under, inherited words included, with its `depth`. See [dictionaries and glossaries](#dictionaries-and-glossaries).                                       |
+| `terms`            | One row per glossary entry, rank or office per note that owns it, with the `cultures` it reaches. See [dictionaries and glossaries](#dictionaries-and-glossaries).                                   |
+| `market`           | The market scale as three columns — `value`, `name`, `trade` — so a table can print `village` beside the number a note wrote.                                                                        |
+| `<package>.notes`  | The same relations — `notes`, `entries`, `events`, `words`, `glossary`, `language_lineage`, `dictionary` and `terms` — for each package this one declares a dependency on, in a schema named for it. |
 
 Nested frontmatter is addressed exactly as a note writes it: `name.full`, `sohl.weight`, `data.born`. Every Address is a string — the full canonical Address, as in `list_contains(data.parents, 'thalorna-note-place-xerathia')` — whichever package published it. Every note carries a computed `state` column of `full`, `draft` or `stub`. Notes tagged `gm` are absent from the relations on public surfaces.
 
@@ -414,12 +419,13 @@ An event with no `where.locus` keeps its row with an empty **Where** cell.
 
 ### Fence attributes
 
-A braced list may follow the language on the opening fence. Two attributes are accepted, and anything else is reported by name.
+A braced list may follow the language on the opening fence. Three attributes are accepted, and anything else is reported by name.
 
-| Attribute       | Value                       | Default | What it does                                                                                                                                                                            |
-| --------------- | --------------------------- | ------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| `allow-empty`   | `true` or `false`           | `false` | Whether a query selecting no rows is allowed. A zero-row result is a finding otherwise, because a table that silently prints nothing is the commonest way a query goes wrong unnoticed. |
-| `section-level` | an integer from 1 through 6 | `2`     | The heading level given to each group a `_section` column creates.                                                                                                                      |
+| Attribute       | Value                               | Default | What it does                                                                                                                                                                            |
+| --------------- | ----------------------------------- | ------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `allow-empty`   | `true` or `false`                   | `false` | Whether a query selecting no rows is allowed. A zero-row result is a finding otherwise, because a table that silently prints nothing is the commonest way a query goes wrong unnoticed. |
+| `section-level` | an integer from 1 through 6         | `2`     | The heading level given to each group a `_section` column creates, and to each letter of a dictionary or a glossary.                                                                    |
+| `layout`        | `table`, `dictionary` or `glossary` | `table` | How the result is set: as a table, as a dictionary of words, or as a glossary of terms — see [dictionaries and glossaries](#dictionaries-and-glossaries).                               |
 
 ````markdown
 ```sql {allow-empty=true section-level=3}
@@ -445,6 +451,65 @@ FROM notes WHERE type IN ('lore', 'place') ORDER BY type, name.full
 
 See [SQL details](../reference/format-details.md#content-tables) for query options.
 
+### Dictionaries and glossaries
+
+Two layouts set a result as paragraphs under letter headings rather than as a table. **`layout=dictionary` sets words**, read from the `dictionary` relation: each with its class, its etymology, its senses and the notes it is cited from. **`layout=glossary` sets terms**, read from the `terms` relation: a culture's or an affiliation's own glossary entries and an affiliation's ranks and offices, each sense tagged with the body or culture it comes from. Each needs the `name`, `senses` and `usage` columns, and a query that leaves one out is reported, naming the column.
+
+**The query decides what appears**, and any note may carry either fence. These are the common ones:
+
+| What                                      | Layout       | Query                                                                             |
+| ----------------------------------------- | ------------ | --------------------------------------------------------------------------------- |
+| A language's dictionary                   | `dictionary` | `SELECT * FROM dictionary WHERE language = '<the language's Address>'`            |
+| A language's common words, on any page    | `dictionary` | `SELECT * FROM dictionary WHERE language = '<the language's Address>' AND common` |
+| A culture's glossary                      | `glossary`   | `SELECT * FROM terms WHERE list_contains(cultures, '<the culture's Address>')`    |
+| An affiliation's glossary                 | `glossary`   | `SELECT * FROM terms WHERE owner = '<the affiliation's Address>'`                 |
+| The setting's _Dictionary of the Tongues_ | `dictionary` | `SELECT * EXCLUDE (common) FROM dictionary WHERE depth = 0`                       |
+| The setting's _Glossary_                  | `glossary`   | `SELECT * EXCLUDE (common) FROM terms`                                            |
+
+A language's dictionary lists the words it inherits through `data.family` beside its own; `depth = 0` lists each word once, under the language that states it. A culture's glossary holds the culture's own terms and the ranks, offices and glossary terms of every affiliation that reaches it. The two setting pages leave out `common`, so they mark no entry: which words are common is a culture's judgement.
+
+The most common Vylari words, on the Vylarian culture's page:
+
+````markdown
+## Words Everyone Knows {#common-words}
+
+```sql {layout=dictionary section-level=3}
+SELECT * FROM dictionary WHERE language = 'thalorna-note-skill-vylarilng' AND common
+```
+````
+
+A culture's glossary:
+
+````markdown
+## Glossary {#glossary}
+
+```sql {layout=glossary section-level=3}
+SELECT * FROM terms WHERE list_contains(cultures, 'thalorna-note-lore-provenzianclt')
+```
+````
+
+Either layout prints, in order:
+
+- **a key**, once, listing only the abbreviations the result uses — a language's tag, and for words a class or a modifier — each with what it stands for, and the common mark where an entry carries it;
+- **a heading per letter**, at `section-level`, accents taken off and apostrophes skipped, so `'ostal` files under O and `Écuyer` under E;
+- **one paragraph per entry**: `•` for a common entry, the headword in bold, its pronunciation between slashes, its language tag in italics, the senses, and after a dash the notes it is cited from, as links.
+
+A word also carries its class beside the tag and its etymology in brackets:
+
+```markdown
+• **solmère** /sol-MEH-reh/ _n. m._ [fr. Élav. _sólmerë_] Longing; the ache for a time that is gone. — [[thalorna-note-lore-lastshrine|The Song of the Last Shrine]]
+```
+
+A term carries no class, and each of its senses names its source. A glossary sets the rows sharing a headword and a language as one entry, numbering their senses in the order the rows arrive and printing a sense several bodies give once, with all of them:
+
+```markdown
+**Sworn Knight** **1** (_Duchy of Aubrac_) Holds land of the duke. **2** (_Kingdom of Provenzia_) Holds land of the crown. **3** (_Provenzia_) A knight bound by oath to a lord. — [[thalorna-note-affiliation-duchy|Duchy of Aubrac]], [[thalorna-note-affiliation-kingdom|Kingdom of Provenzia]], [[thalorna-note-lore-provenzianclt|Provenzia]]
+```
+
+The entries are ordered by the layout, not by the query: by letter, then by `name_sort`. **The language tag prints where the entry's language is not the page's own**: a language note's own language, a culture's `data.language`, or the language an affiliation's glossary resolves to. It is the language note's `data.abbreviation`, or its full name where it states none. The etymology never repeats the entry's own language: a borrowed word prints `[fr. Élav. _sólmerë_]`, a built word its pieces each with its gloss, `[_am-_ "love" + _-ère_ "one who"]`, an inherited word its older form, `[*_hur_]` or `[Hel. *_thalass_]`, and a root word its root and pattern. A citation of a note with no published page prints as its name.
+
+A culture's or an affiliation's `data.glossary` shapes both: an entry without a definition marks a word or a rank common, gives it a sort key or a respelling, and lets in a rank whose title is a single word; an entry with a definition replaces the senses of the owner's own rank or office, or is a term of the owner's own. The [format reference](../reference/format-details.md#dictionaries-and-glossaries) gives every column and rule.
+
 ## Page lists
 
 A fence marked `pagelist` is replaced by a list of the pages carrying a tag. Where a content table answers "what do these notes say?", a page list answers "which pages belong to this group?" — and the group is expressed by intent rather than enumerated, so a page joins it by being tagged and leaves it by being untagged.

package/docs/commands.md

@@ -789,50 +789,6 @@ package-build: 1 template(s) fully localized and compiling.
 `package-build lang check`, `package-build lang coverage`,
 [Diagnostics](diagnostics.md), [Configuration](configuration.md).
 
-### `package-build labels check`
-
-**NAME**
-
-Compare the machine label registry against the documented reference table
-(`check` is the only action).
-
-**SYNOPSIS**
-
-```
-package-build labels check [--registry <path>] [--doc <path>]
-```
-
-**DESCRIPTION**
-
-Compares the machine label registry synced to GitHub against the documented
-reference table, and reports where they disagree — a label the registry has
-that the table does not, or the reverse. Neither derives from the other, so
-nothing else notices when they drift. Reads both files named below; writes
-nothing.
-
-**OPTIONS**
-
-| Option       | Type   | Default                      | Description                                                    |
-| ------------ | ------ | ---------------------------- | -------------------------------------------------------------- |
-| `--registry` | string | `.github/labels.yml`         | The machine registry synced to GitHub.                         |
-| `--doc`      | string | `.github/ISSUE_REPORTING.md` | The documented reference whose §3 table lists the same labels. |
-
-**EXIT STATUS**
-
-1 if either file does not exist. 1 if the registry and the table disagree on
-any label. Otherwise 0.
-
-**EXAMPLES**
-
-```
-$ package-build labels check --registry .github/labels.yml --doc .github/ISSUE_REPORTING.md
-package-build: registry and §3 agree (11 labels).
-```
-
-**SEE ALSO**
-
-None.
-
 ### `package-build bump [packages..]`
 
 **NAME**
@@ -1348,7 +1304,7 @@ A workflow calls the command the same way, passing the four values as named
 secrets in the step's environment:
 
 ```yaml
-- run: npx package-build cdn publish --set ai-art.json --name "Thalorna — illustrated"
+- run: npx package-build cdn publish --set alt-art.json --name "Thalorna — illustrated"
   env:
     CDN_ENDPOINT: ${{ secrets.CDN_ENDPOINT }}
     CDN_BUCKET: ${{ secrets.CDN_BUCKET }}
@@ -1374,12 +1330,12 @@ it. Otherwise 0.
 **EXAMPLES**
 
 ```
-$ package-build cdn publish --set ai-art.json --name "Thalorna — illustrated" --dry-run
+$ package-build cdn publish --set alt-art.json --name "Thalorna — illustrated" --dry-run
 Dry run: comparing 582 pictures against https://cdn.heroiclands.org, read-only.
   would upload thalorna/icons/anvilbdg.svg (969 bytes)
   …
-  would upload thalorna/ai-art.json (582 assets)
-Would upload 582, skipped 0, failed 0; https://cdn.heroiclands.org/thalorna/ai-art.json lists 582 left-out pictures (would change).
+  would upload thalorna/alt-art.json (582 assets)
+Would upload 582, skipped 0, failed 0; https://cdn.heroiclands.org/thalorna/alt-art.json lists 582 left-out pictures (would change).
 
 $ package-build cdn publish --set default.json --name Shipped
 package-build: `--set default.json` is refused: `default.json` is the shipped set's name, the fallback every other set is read against. Name the set for its art

package/docs/content-format.md

@@ -91,6 +91,18 @@ errors, reported at the `ranks` key.
 
 An Item's `{#description}` page is its readable description. The note Address `<package>-note-<type>-<shortcode>#description` resolves to a JournalEntryPage UUID. The build writes the same `@UUID[...]` pointer to SoHL Item `system.docHtml` and HM3 Item `system.description`. An Item note without that anchor uses its first page as the description. Other pages and anchors remain available for links.
 
+## File and folder names
+
+Every character in the path of a content file or a picture is printable 7-bit ASCII, U+0020 to U+007E. The rule covers each file and folder name under the content tree and under the package's `icons/` and `images/` roots. Letters, digits, spaces, hyphens, underscores, dots and apostrophes are all inside it, so `Knight's_Shield.md` is a valid name. An accented letter, any other non-ASCII character, and a control character are outside it.
+
+`package-build lint` reports each offending name as an error located at the path, with no line or column, and names the characters by code point:
+
+```text
+assets/content/Misc_Gear/Scribe/A_Sílhari_Manual.md: error: the path holds U+00ED (í), outside printable ASCII; rename the file using ASCII only
+```
+
+The fix is to rename the file, `A_Sílhari_Manual.md` to `A_Silhari_Manual.md`. A note's identity comes from its `shortcode`, and a picture's from its basename without the extension, so the rename changes no address and breaks no link. The accented spelling belongs in the note's `name`, where it displays as written.
+
 ## A complete minimal note
 
 Place this at `assets/content/Places/Harbor.md`:
@@ -129,4 +141,8 @@ for a complete example, parent ordering and findings.
 
 Places, affiliations, lore and beings may name their culture in `data.culture`, and a `doc` of `subType: settingguide` must. A place or affiliation that names none inherits the culture of its parents. See [culture in the index](reference/format-details.md#culture-in-the-index).
 
+An affiliation may name the trade, temple or order tongue it keeps in its own `data.language`. An affiliation that reaches no culture is a warning, in a tree holding a culture note. A language needs no culture and no affiliation. See [type: affiliation](reference/format-details.md#type-affiliation).
+
+A language note (`type: skill`, `subType: language`) states its words in frontmatter: `data.words` states each word as a map of its form, class, field and gloss, with its source, the pieces it is built from and the notes it is attested in, beside `data.classModifiers`, `data.fields` and `data.affixes`. The language of a word is the note it sits on, and none of it reaches the compiled Item. A daughter language names its parent in `data.family` and reads the parent's words as its own; the `language_lineage` relation joins a language to its ancestors. Classes come from the [master word classes](reference/format-details.md#word-classes). A culture or an affiliation states its own glossary terms in `data.glossary`, and every entry resolves to a language — its own `language`, else the affiliation's, its parents' or its culture's, else the culture's — or is an error at the entry. The `words`, `glossary` and `language_lineage` relations hold them for `sql` fences, the `dictionary` relation lists each word under every language that has it, inherited words included, and the `terms` relation each glossary entry, rank and office with the cultures it reaches; a fence sets them with `layout=dictionary` or `layout=glossary`. A language note names the tag a dictionary prints for it in `data.abbreviation`. See [a language's words](reference/format-details.md#a-languages-words), [glossary terms](reference/format-details.md#glossary-terms) and [dictionaries and glossaries](reference/format-details.md#dictionaries-and-glossaries).
+
 Blockquote alerts use `[!NOTE]`, `[!TIP]`, `[!IMPORTANT]`, `[!WARNING]`, or `[!]`, with optional attributes after the marker. They replace the former info and warn div fences; secret blocks remain supported. See [alerts](authoring/links-and-markup.md#alerts).

package/docs/diagnostics.md

@@ -284,7 +284,7 @@ when it reports at least one `error`-severity diagnostic. A `warning` alone
 never fails a run.** `bin/report.mjs`'s `reportFindings` draws this line once
 for the commands built from the package's pure rule functions (`package-build
 lint`, `package-build lang check`, `package-build lang coverage`,
-`package-build yaml`, `package-build labels check`): it emits every finding
+`package-build yaml`): it emits every finding
 and returns only the count whose `severity` is `error`, and the command sets
 `process.exitCode = 1` exactly when that count is nonzero. Commands whose
 findings are hand-assembled — `package-build links`, `package-build
@@ -335,7 +335,6 @@ over one would teach people to stop reading the report.
 | `package-build lang check`                       | any localization finding is `error`-severity                                                                                                                                                                                                                       |
 | `package-build lang coverage`                    | a referenced key is missing (`error`-severity); an unreferenced key is always advisory and never fails the run                                                                                                                                                     |
 | `package-build yaml [paths..]`                   | any YAML finding is `error`-severity                                                                                                                                                                                                                               |
-| `package-build labels check`                     | the label registry and the documented table disagree on any label                                                                                                                                                                                                  |
 
 Every command also fails on a thrown error unrelated to a
 diagnostic — a missing configuration, an unreadable file, a network failure —

package/docs/guides/module-project.md

@@ -861,7 +861,7 @@ everything that reads the tree runs before everything that writes the stage.
 
 [`project-setup.md`](../project-setup.md) turns this into npm scripts, and covers
 everything a repository carries beyond the build itself: the git hooks, the
-changeset directory, the label registry, and what each script in the chain is
+changeset directory, and what each script in the chain is
 for.
 
 ## What comes next

package/docs/guides/reference-pages.md

@@ -163,6 +163,95 @@ frontmatter remains queryable. An Item or Actor note can have both a readable
 Consumers can use the index's own addresses and file paths rather than
 reconstructing them from filenames.
 
+## Set words as a dictionary and terms as a glossary
+
+A language's words and a people's terms are each one `sql` fence, and nothing
+is copied into the note: the words come from the language notes' `data.words`,
+the terms from the `data.glossary` entries of cultures and affiliations and
+from the affiliations' ranks and offices, and a section changes when they do.
+`layout=dictionary` sets rows of the `dictionary` relation and
+`layout=glossary` rows of the `terms` relation. The query decides what
+appears, so any note may carry either.
+
+A language note lists its words, with the words it inherits from its parent
+language:
+
+````markdown
+# Dictionary {#dictionary}
+
+```sql {layout=dictionary section-level=2}
+SELECT * FROM dictionary WHERE language = 'thalorna-note-skill-provnzlng'
+```
+````
+
+A culture note lists the words of its tongue a player is expected to know, and
+its glossary: its own terms and the ranks, offices and terms of every
+affiliation that reaches it.
+
+````markdown
+# Words Everyone Knows {#common-words}
+
+```sql {layout=dictionary section-level=2}
+SELECT * FROM dictionary WHERE language = 'thalorna-note-skill-provnzlng' AND common
+```
+
+# Glossary {#glossary}
+
+```sql {layout=glossary section-level=2}
+SELECT * FROM terms WHERE list_contains(cultures, 'thalorna-note-lore-provenzianclt')
+```
+````
+
+The glossary renders as a key, a heading per letter and one paragraph per
+term, each sense tagged with the body or culture it comes from and each term
+in another tongue tagged with it:
+
+```markdown
+**Key:** _Lit._ Liturgic Language; • a common word, which a player is expected to know by heart.
+
+## K
+
+**Knight-Captain** **1** (_Kingdom of Provenzia_) A knight who leads a company. **2** (_Kingdom of Provenzia_) Commands the guard in the field. — [[thalorna-note-affiliation-kingdom|Kingdom of Provenzia]]
+
+## O
+
+**Ordained Priest** _Lit._ (_Church of the Rite_) Ordained to serve a parish. — [[thalorna-note-affiliation-church|Church of the Rite]]
+```
+
+The culture shapes both through its own `data.glossary`.
+`{name: solmère, common: true, pronunciation: sol-MEH-reh}` marks a word of its
+tongue common, so the first fence lists it, and gives it a respelling;
+`{name: Subject, common: true}` marks a rank and lets in a title of one word,
+which no glossary holds unless an entry names it; and
+`{name: blood-field, definition: …}` is a term of the culture's own.
+
+An affiliation lists its own terms with `WHERE owner = '<its Address>'`. The
+setting's two pages list everything, each word once and each term under every
+owner that holds it:
+
+````markdown
+```sql {layout=dictionary}
+SELECT * EXCLUDE (common) FROM dictionary WHERE depth = 0
+```
+````
+
+````markdown
+```sql {layout=glossary}
+SELECT * EXCLUDE (common) FROM terms
+```
+````
+
+`package-build content-index` also writes both relations beside the index, for
+reading with `jq`:
+
+```bash
+jq -r 'select(.cultures | index("thalorna-note-lore-provenzianclt")) | .name' \
+  build/content-index/thalorna-terms.jsonl
+```
+
+The [format reference](../reference/format-details.md#dictionaries-and-glossaries)
+gives every column and every rule.
+
 ## Resolve another package
 
 In a Foundry package, declare a content-bearing dependency under

package/docs/guides/release-and-deploy.md

@@ -139,13 +139,13 @@ reaches a world from there as a named asset set. `package-build cdn publish`
 is the one command that writes there:
 
 ```bash
-package-build cdn publish --set ai-art.json --name "Thalorna — illustrated" --dry-run
-package-build cdn publish --set ai-art.json --name "Thalorna — illustrated"
+package-build cdn publish --set alt-art.json --name "Thalorna — illustrated" --dry-run
+package-build cdn publish --set alt-art.json --name "Thalorna — illustrated"
 ```
 
 It uploads every icon and image flat, transformed as the stage transforms
 them, skips what the CDN already holds at the same size, and writes
-`<site.assets>/<contentPackage>/ai-art.json` listing the left-out pictures.
+`<site.assets>/<contentPackage>/alt-art.json` listing the left-out pictures.
 Run it after the pictures or their provenance change; it is not part of a
 build or a release.