#70-Debug: Diar Weak-Fold (7->4 Sprecher), LLM-Robustheit + Whole-Pass-Modellgrenze, Thermal-Cap fuer analyzeRecording

- diar.js: Stufe-3 Weak-Fold (DIAR_MIN_TOTAL_SEC 20s) gegen Phantom-Sprecher
- llm.js: End-Instruktion, JSON-Repair, Sanity+Fallback, Chunk-Retry, Quote-Sanitizer, gelabelter/gewichteter Reduce; WHOLE_MAX 20000 (Gemma-4B-Grenze)
- rkllm_server.cpp: max_context_len 8192->12288
- server.js: analyzeRecording unter A76-Takt-Cap (95-97 Grad ungedeckelt)
- Doku Kap. 03/04/05 + Changelog

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0113vDKdSNtD3Dwcw4QMu9yV

app/server.js

@@ -82,7 +82,11 @@ async function identifySpeakers(recData) {
 
 // Run LLM analysis for a recording and persist title/summary/todos/participants/
 // clean dialogue. Background (NPU ~30-90s) so record/stop stays fast.
+// A76 clock cap for the whole run: rkllm burns ~1.7 CPU cores alongside the NPU
+// and a long Gemma generation hit 95-97°C uncapped on the fanless board (#70,
+// 16-min run) — same shutdown risk transcribeWav/rediarize already guard against.
 async function analyzeRecording(id) {
+  await pexec('sudo', ['-n', '/opt/callrec/aime-cpucap.sh', '1608000']).catch(() => {});
   try {
     const rec = await db.getRecording(id);
     if (!rec) return;
@@ -122,6 +126,8 @@ async function analyzeRecording(id) {
   } catch (e) {
     console.error(`analysis failed for #${id}:`, e.message);
     await db.updateAnalysis(id, { title: null, summary: null, todos: null, status: 'error' }).catch(() => {});
+  } finally {
+    await pexec('sudo', ['-n', '/opt/callrec/aime-cpucap.sh', '2256000']).catch(() => {});
   }
 }