fix(vibe): warm CLAP at server startup + honest UI feedback/timeout

The first /vibe hit cold-loaded CLAP (~16s) with the UI stuck on 'searching…',
reading as hung. Now serve.py warms the model in a daemon thread at startup (when
semantics_embeds.npz is present) so the first query is instant; the client shows
'first search loads the model, ~15s' on a cold start and aborts after 90s with a
retry hint instead of spinning forever.

armada/serve.py

@@ -173,4 +173,16 @@ if __name__ == "__main__":
     print(f"serving {os.getcwd()}")
     print(f"  local : http://127.0.0.1:{a.port}/")
     print(f"  LAN   : http://{ip}:{a.port}/   ← open on phone (same wifi)")
+
+    # Warm CLAP in the background so the FIRST /vibe query isn't a 15s cold-load.
+    if os.path.exists("semantics_embeds.npz"):
+        def _warm():
+            try:
+                print("  vibe  : warming CLAP…", flush=True)
+                _vibe_load()
+                print("  vibe  : ready (/vibe, /similar)", flush=True)
+            except Exception as e:
+                print(f"  vibe  : disabled — {e}", flush=True)
+        threading.Thread(target=_warm, daemon=True).start()
+
     httpd.serve_forever()

armada/tide-table/unwrapped.html

@@ -384,12 +384,19 @@ function vibeRamp(sim){const t=state.vibe?state.vibe.norm(sim):0.5;
   // faint violet → brand magenta as similarity rises
   const r=Math.round(90+t*(217-90)),g=Math.round(58+t*(0-58)),b=Math.round(106+t*(255-106));
   return `rgb(${r},${g},${b})`;}
+let _vibeWarm=false;
 async function runVibe(url,label){
-  const banner=$("#vibebanner");banner.innerHTML=`<span class="mono">searching…</span>`;
-  banner.classList.add("on");
+  const banner=$("#vibebanner");banner.classList.add("on");
+  banner.innerHTML=`<span class="mono">searching…${_vibeWarm?"":" (first search loads the model, ~15s)"}</span>`;
   let data;
-  try{ data=await fetch(url).then(r=>r.json()); }
-  catch(_){ data={error:"no /vibe endpoint — start serve.py (python3 ../serve.py --dir .)"}; }
+  const ac=new AbortController();const to=setTimeout(()=>ac.abort(),90000);
+  try{ const r=await fetch(url,{signal:ac.signal});
+    if(!r.ok){data=await r.json().catch(()=>({error:`endpoint ${r.status}`}));}
+    else{data=await r.json();_vibeWarm=true;} }
+  catch(e){ data={error: e.name==="AbortError"
+    ? "timed out (90s) — is CLAP still loading? retry in a moment"
+    : "no /vibe endpoint — run: python3 armada/serve.py --dir armada/tide-table --port 8731"}; }
+  finally{ clearTimeout(to); }
   if(data.error){banner.innerHTML=`<span class="err">⚠ ${data.error}</span>`
     +`<button class="x" id="vclr">clear</button>`;$("#vclr").onclick=clearVibe;
     state.vibe=null;return;}