feat(dedup): per-kind duplicate-report floors — notes/tasks default 0.93

At chunk grain (#280) a note-pair's similarity is its closest chunk pair,
so the shared 0.82 floor saturated the note/task reports with related
families (38 note / 155 task groups against the 200-pair cap, measured
2026-08-09). Split kb_duplicate_threshold into per-kind settings keys
with per-kind defaults: snippet 0.82 (single-chunk, scale unchanged),
note/task 0.93 (points the report at genuinely-alike records). Settings
UI grows the two new knobs; report entrypoints inherit the change via
get_duplicate_threshold(user_id, kind).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaYUaouG9jjhATyuxCKrQs

frontend/src/views/SettingsView.vue

@@ -27,10 +27,14 @@ const kbWritePathEnabled = ref(true);
 // code embeddings sit on a much higher similarity floor than prose, so 0.55 let
 // unrelated code through (#2223). Shares top-k, not the threshold.
 const kbWritePathThreshold = ref("0.68");
-// Near-duplicate report floor. Deliberately looser than the 0.90 write-time
-// gate: that one BLOCKS a create and must be unforgiving of noise, this one only
-// suggests a merge the operator reviews (services/dedup.py).
-const kbDuplicateThreshold = ref("0.82");
+// Near-duplicate report floors, one per record kind (services/dedup.py).
+// Snippets are single-chunk, so their floor sits below the 0.90 write-time
+// gate and catches what it lets through. Notes/tasks are scored at chunk
+// grain (#280) — related families clear 0.90 easily — so their floor sits
+// above the gate to keep the report pointed at genuinely-alike records.
+const kbDupThresholdSnippet = ref("0.82");
+const kbDupThresholdNote = ref("0.93");
+const kbDupThresholdTask = ref("0.93");
 const savingKbInject = ref(false);
 const kbInjectSaved = ref(false);
 
@@ -76,16 +80,20 @@ async function saveRetention() {
 async function saveKbInject() {
   const t = Math.min(1, Math.max(0, Number(kbInjectThreshold.value) || 0));
   const k = Math.min(10, Math.max(1, Math.floor(Number(kbInjectTopK.value) || 1)));
-  // `|| 0.82` not `|| 0`: an unparseable value here should fall back to the
-  // default, not to 0 — a 0 floor would report every snippet as a duplicate of
-  // every other one.
-  const dupT = Math.min(1, Math.max(0, Number(kbDuplicateThreshold.value) || 0.82));
-  // Same `|| default` reasoning as dupT: falling back to 0 would surface every
+  // `|| default` not `|| 0`: an unparseable value here should fall back to the
+  // per-kind default, not to 0 — a 0 floor would report every record as a
+  // duplicate of every other one.
+  const dupSnip = Math.min(1, Math.max(0, Number(kbDupThresholdSnippet.value) || 0.82));
+  const dupNote = Math.min(1, Math.max(0, Number(kbDupThresholdNote.value) || 0.93));
+  const dupTask = Math.min(1, Math.max(0, Number(kbDupThresholdTask.value) || 0.93));
+  // Same `|| default` reasoning: falling back to 0 would surface every
   // snippet in the corpus on every edit, which is the failure this knob fixes.
   const wpT = Math.min(1, Math.max(0, Number(kbWritePathThreshold.value) || 0.68));
   kbInjectThreshold.value = String(t);
   kbInjectTopK.value = String(k);
-  kbDuplicateThreshold.value = String(dupT);
+  kbDupThresholdSnippet.value = String(dupSnip);
+  kbDupThresholdNote.value = String(dupNote);
+  kbDupThresholdTask.value = String(dupTask);
   kbWritePathThreshold.value = String(wpT);
   savingKbInject.value = true;
   kbInjectSaved.value = false;
@@ -99,7 +107,9 @@ async function saveKbInject() {
       // measurements that split them.
       kb_writepath_enabled: kbWritePathEnabled.value ? 'true' : 'false',
       kb_writepath_threshold: String(wpT),
-      kb_duplicate_threshold: String(dupT),
+      kb_duplicate_threshold_snippet: String(dupSnip),
+      kb_duplicate_threshold_note: String(dupNote),
+      kb_duplicate_threshold_task: String(dupTask),
     });
     kbInjectSaved.value = true;
     setTimeout(() => (kbInjectSaved.value = false), 2000);
@@ -487,8 +497,14 @@ onMounted(async () => {
   if (allSettings.kb_writepath_threshold !== undefined) {
     kbWritePathThreshold.value = allSettings.kb_writepath_threshold;
   }
-  if (allSettings.kb_duplicate_threshold !== undefined) {
-    kbDuplicateThreshold.value = allSettings.kb_duplicate_threshold;
+  if (allSettings.kb_duplicate_threshold_snippet !== undefined) {
+    kbDupThresholdSnippet.value = allSettings.kb_duplicate_threshold_snippet;
+  }
+  if (allSettings.kb_duplicate_threshold_note !== undefined) {
+    kbDupThresholdNote.value = allSettings.kb_duplicate_threshold_note;
+  }
+  if (allSettings.kb_duplicate_threshold_task !== undefined) {
+    kbDupThresholdTask.value = allSettings.kb_duplicate_threshold_task;
   }
   if (allSettings.notify_task_reminders !== undefined) {
     notifyTaskReminders.value = allSettings.notify_task_reminders !== "false";
@@ -1267,10 +1283,10 @@ function formatUserDate(iso: string): string {
              is for (#274). -->
 
         <div class="field">
-          <label for="kb-duplicate-threshold">Near-duplicate report threshold</label>
+          <label for="kb-duplicate-threshold-snippet">Near-duplicate report threshold — snippets</label>
           <input
-            id="kb-duplicate-threshold"
-            v-model="kbDuplicateThreshold"
+            id="kb-duplicate-threshold-snippet"
+            v-model="kbDupThresholdSnippet"
             type="number"
             min="0"
             max="1"
@@ -1285,6 +1301,46 @@ function formatUserDate(iso: string): string {
             you review — it never acts on its own.
           </p>
         </div>
+
+        <div class="field">
+          <label for="kb-duplicate-threshold-note">Near-duplicate report threshold — notes</label>
+          <input
+            id="kb-duplicate-threshold-note"
+            v-model="kbDupThresholdNote"
+            type="number"
+            min="0"
+            max="1"
+            step="0.01"
+            class="input"
+            style="max-width: 8rem"
+          />
+          <p class="field-hint">
+            The floor for the Knowledge page's note report. Notes are compared
+            section by section, so related records — a run of dev-logs, notes on
+            one topic — score high without being duplicates. Stricter than the
+            snippet floor on purpose; lower it to browse related families rather
+            than hunt true duplicates.
+          </p>
+        </div>
+
+        <div class="field">
+          <label for="kb-duplicate-threshold-task">Near-duplicate report threshold — tasks</label>
+          <input
+            id="kb-duplicate-threshold-task"
+            v-model="kbDupThresholdTask"
+            type="number"
+            min="0"
+            max="1"
+            step="0.01"
+            class="input"
+            style="max-width: 8rem"
+          />
+          <p class="field-hint">
+            Same as the note floor, for the task report. Step tasks from different
+            milestones ("Verify on CI") legitimately resemble each other, so this
+            stays strict to keep the report pointed at work opened twice.
+          </p>
+        </div>
         <div class="actions">
           <button class="btn-primary" @click="saveKbInject" :disabled="savingKbInject">
             {{ savingKbInject ? 'Saving…' : 'Save' }}