@aspan-corporation/ac-shared 1.2.36 → 1.2.38

package/lib/utils/diary.d.ts

@@ -15,8 +15,21 @@ export declare const TAG_DIARY_TITLE = "ac:diary:title";
 export declare const TAG_DIARY_PREVIEW = "ac:diary:preview";
 /** One per embedded photo: value is the referenced media key. */
 export declare const TAG_DIARY_PHOTO = "ac:diary:photo";
+/** One per embedded audio recording: value is the referenced diary key. */
+export declare const TAG_DIARY_AUDIO = "ac:diary:audio";
 /** Namespace for body word tokens: `ac:text:<word>`. */
 export declare const TEXT_TOKEN_PREFIX = "ac:text:";
+/**
+ * Audio formats a browser MediaRecorder can produce: Safari records
+ * audio/mp4 (.m4a), Chromium/Firefox audio/webm or audio/ogg (opus).
+ * Deliberately kept SEPARATE from the image-only `isAllowedExtension`
+ * (thumbsKey.ts) — that allowlist is what keeps diary audio out of the
+ * meta-extractor/resizer dispatch on save, so recordings stay entry-only
+ * (no meta item, never in browse/search).
+ */
+export declare const AUDIO_EXTENSIONS: string[];
+/** True when a key's extension is a supported diary audio format. */
+export declare const isAllowedAudioExtension: (key: string) => boolean;
 /** Max body characters surfaced in the preview tag. */
 export declare const DIARY_PREVIEW_LENGTH = 160;
 /** Upper bound on distinct word tokens written per entry (bounds item size). */
@@ -49,6 +62,14 @@ export declare const parseDiaryKeyDate: (key: string) => {
  * words. Strips markdown syntax (code fences, image/link targets, formatting),
  * drops stop-words and tokens shorter than 2 chars, and caps the result at
  * `MAX_TEXT_TOKENS` to bound the meta-item size. No stemming (exact-word match).
+ *
+ * Word boundaries and lowercasing are Unicode-aware: tokens are split on any
+ * run of characters that are neither a Unicode letter (`\p{L}`) nor number
+ * (`\p{N}`), so non-Latin scripts (e.g. Cyrillic) survive as searchable words
+ * rather than being discarded. `toLowerCase()` is likewise Unicode-aware; we
+ * deliberately avoid `toLocaleLowerCase()` so tokenisation is deterministic and
+ * independent of the host locale (the indexer lambda and the browser that
+ * builds search queries must produce identical tokens).
  */
 export declare const tokenizeText: (markdown: string) => string[];
 /** Plain-text preview: strip markdown, collapse whitespace, truncate. */

package/lib/utils/diary.js

@@ -5,6 +5,7 @@
  * searchable by tokenising it into `ac:text:<word>` tags that ride the existing
  * inverted-index search table.
  */
+import { getKeyExtension } from "./thumbsKey.js";
 /** Key prefix for every diary object (and the `folder`/id namespace). */
 export const DIARY_PREFIX = "diary/";
 /** Marks a meta item as a diary entry (value "true"). */
@@ -15,8 +16,21 @@ export const TAG_DIARY_TITLE = "ac:diary:title";
 export const TAG_DIARY_PREVIEW = "ac:diary:preview";
 /** One per embedded photo: value is the referenced media key. */
 export const TAG_DIARY_PHOTO = "ac:diary:photo";
+/** One per embedded audio recording: value is the referenced diary key. */
+export const TAG_DIARY_AUDIO = "ac:diary:audio";
 /** Namespace for body word tokens: `ac:text:<word>`. */
 export const TEXT_TOKEN_PREFIX = "ac:text:";
+/**
+ * Audio formats a browser MediaRecorder can produce: Safari records
+ * audio/mp4 (.m4a), Chromium/Firefox audio/webm or audio/ogg (opus).
+ * Deliberately kept SEPARATE from the image-only `isAllowedExtension`
+ * (thumbsKey.ts) — that allowlist is what keeps diary audio out of the
+ * meta-extractor/resizer dispatch on save, so recordings stay entry-only
+ * (no meta item, never in browse/search).
+ */
+export const AUDIO_EXTENSIONS = ["m4a", "webm", "ogg"];
+/** True when a key's extension is a supported diary audio format. */
+export const isAllowedAudioExtension = (key) => AUDIO_EXTENSIONS.includes(getKeyExtension(key));
 /** Max body characters surfaced in the preview tag. */
 export const DIARY_PREVIEW_LENGTH = 160;
 /** Upper bound on distinct word tokens written per entry (bounds item size). */
@@ -91,6 +105,14 @@ const STOP_WORDS = new Set([
  * words. Strips markdown syntax (code fences, image/link targets, formatting),
  * drops stop-words and tokens shorter than 2 chars, and caps the result at
  * `MAX_TEXT_TOKENS` to bound the meta-item size. No stemming (exact-word match).
+ *
+ * Word boundaries and lowercasing are Unicode-aware: tokens are split on any
+ * run of characters that are neither a Unicode letter (`\p{L}`) nor number
+ * (`\p{N}`), so non-Latin scripts (e.g. Cyrillic) survive as searchable words
+ * rather than being discarded. `toLowerCase()` is likewise Unicode-aware; we
+ * deliberately avoid `toLocaleLowerCase()` so tokenisation is deterministic and
+ * independent of the host locale (the indexer lambda and the browser that
+ * builds search queries must produce identical tokens).
  */
 export const tokenizeText = (markdown) => {
     if (!markdown)
@@ -109,11 +131,12 @@ export const tokenizeText = (markdown) => {
         // leftover markdown punctuation
         .replace(/[#>*_~\-]+/g, " ");
     const seen = new Set();
-    for (const raw of stripped.toLowerCase().split(/[^a-z0-9]+/)) {
+    for (const raw of stripped.toLowerCase().split(/[^\p{L}\p{N}]+/u)) {
         if (raw.length < 2)
             continue;
-        // Drop pure numbers (dates/counts add noise); keep alphanumerics like "v2".
-        if (/^\d+$/.test(raw))
+        // Drop pure-number tokens (dates/counts add noise) in any script; keep
+        // alphanumerics like "v2".
+        if (/^\p{N}+$/u.test(raw))
             continue;
         if (STOP_WORDS.has(raw))
             continue;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aspan-corporation/ac-shared",
-  "version": "1.2.36",
+  "version": "1.2.38",
   "description": "",
   "keywords": [],
   "exports": {