@chatpanel/events 0.18.0 → 0.18.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@chatpanel/events",
-  "version": "0.18.0",
+  "version": "0.18.1",
   "description": "The canonical ChatPanel event-log and capability contracts \u2014 typed durable facts, clock-free deterministic linearization, schema upcasting, and the invariants the replay harness asserts. Pure, dependency-free ESM shared by the ChatPanel extension, gateway and bridge.",
   "type": "module",
   "main": "index.js",

package/schedule.js

@@ -152,6 +152,32 @@ export function createTriggerRegistry(triggers = []) {
 const words = (s) => String(s || '').toLowerCase().match(/[\p{L}\p{N}']+/gu) || [];
 const norm = (s) => String(s || '').trim().toLowerCase();
 
+/** Shorter than this and ordinary speech contains it constantly. */
+export const MIN_PHRASE_CHARS = 3;
+
+const escapeRe = (s) => s.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+
+/**
+ * Whole-word containment, not `includes`.
+ *
+ * A substring match turns "in" into a trigger that fires on interview, thing, going and
+ * finding — which is indistinguishable, from the user's chair, from a trigger that ignores
+ * its phrase entirely. That is exactly how it was reported: "for every utterance I say, it
+ * runs". Boundaries are only required at ends that are word characters, so "action item"
+ * still matches "an action item," and ":shipped" still matches ":shipped".
+ */
+export function saidIn(text, phrase) {
+  const p = norm(phrase);
+  if (p.length < MIN_PHRASE_CHARS) return false;
+  const left = /[\p{L}\p{N}]/u.test(p[0]) ? '\\b' : '';
+  const right = /[\p{L}\p{N}]/u.test(p[p.length - 1]) ? '\\b' : '';
+  try {
+    return new RegExp(`${left}${escapeRe(p)}${right}`, 'iu').test(String(text || ''));
+  } catch {
+    return norm(text).includes(p); // a phrase that will not compile still gets a plain match
+  }
+}
+
 // Whose speech a meeting trigger cares about. The default is 'anyone' because a phrase
 // trigger is usually watching for what OTHERS say — unlike a spoken command, which is only
 // ever the owner's (see voice-intents.js).
@@ -208,12 +234,14 @@ export const phraseTrigger = defineTrigger({
   kind: 'meeting',
   watches: ['meeting.transcript.delta'],
   matches: (event, params = {}, ctx = {}) => {
-    const any = (params.any || []).map(norm).filter(Boolean);
-    if (!any.length) return null; // a phrase trigger with no phrase would fire on every word
+    // Both guards exist because their absence looks the same to a user: a trigger that fires
+    // on everything. An empty list would match every line; a one- or two-letter phrase
+    // effectively does too.
+    const any = (params.any || []).map(norm).filter((p) => p.length >= MIN_PHRASE_CHARS);
+    if (!any.length) return null;
     for (const seg of event.segments || []) {
       if (!speakerAllowed(params.speaker, seg.speaker, ctx)) continue;
-      const text = norm(seg.text);
-      const hit = any.find((p) => text.includes(p));
+      const hit = any.find((p) => saidIn(seg.text, p));
       if (hit) return { why: `“${hit}” said by ${seg.speaker || 'someone'}`, segment: seg, phrase: hit };
     }
     return null;
@@ -253,7 +281,11 @@ export const questionTrigger = defineTrigger({
       // Other people by default: this exists for reacting to what someone ELSE asks (an
       // interview, a customer call). The job form offers the choice, so 'anyone' and 'me' are
       // one dropdown away rather than an invisible default nobody can reach.
-      if (!speakerAllowed(params.speaker || 'others', seg.speaker, ctx)) continue;
+      // ANYONE by default, including you. 'others' read well — this exists for reacting to
+      // what someone ELSE asks — but it made the first thing anybody does (test it alone in a
+      // call, ask a question, wait) match nothing, so the feature looked dead. "Only what
+      // other people ask" is still one dropdown away in the job form.
+      if (!speakerAllowed(params.speaker || 'anyone', seg.speaker, ctx)) continue;
       const text = String(seg.text || '').trim();
       if (text.length < 8) continue; // "what?" is not a question worth waking a model for
       if (text.includes('?') || QUESTION.test(text)) return { why: `question from ${seg.speaker || 'someone'}`, segment: seg };