Refactor: Agent-Hirn-Info zeigt real genutztes Modell (NousResearch-Update-Logik raus)

hermes_brain_info() suchte das role==hermes-Modell + verglich gegen NousResearch-Hermes-
Releases. Da das Hirn jetzt ein beliebiges Modell ist (fast = Qwen3.6 via Alias), war das
irrefuehrend. Neu: _active_brain_name() liest Hermes' model.default und loest den Alias/Namen
auf das installierte Modell auf; recommended/update_available entfallen; Budget-Check bleibt.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

backend/services/agent.py

@@ -25,63 +25,43 @@ def _hermes_version(name: str) -> float | None:
     return float(m.group(1)) if m else None
 
 
+def _active_brain_name() -> str:
+    """Aktives Agent-Hirn aus Hermes' Config: model.default (sonst model.model)."""
+    try:
+        from ruamel.yaml import YAML
+        p = HERMES_HOME / "config.yaml"
+        if p.exists():
+            with p.open(encoding="utf-8") as f:
+                cfg = YAML().load(f) or {}
+            m = (cfg.get("model") or {}) if isinstance(cfg, dict) else {}
+            return str(m.get("default") or m.get("model") or "auto")
+    except Exception:
+        log.debug("_active_brain_name: Lesefehler", exc_info=True)
+    return "auto"
+
+
 def hermes_brain_info() -> dict:
-    """Aktuelles Agent-Hirn (hermes-Rolle) + bestes verfügbares NousResearch-Hermes-Modell,
+    """Aktuelles Agent-Hirn = Modell/Alias, das Hermes laut Config nutzt (model.default),
-    das auf diese Hardware passt. Für den Modell-Manager: Brain sichtbar + updatebar,
+    plus Budget-Check. Zeigt das REAL genutzte Hirn — unabhängig von einer 'hermes'-Rolle."""
-    sobald NousResearch eine neuere Hermes-Generation veröffentlicht."""
+    from services import llamaswap
-    from services import discover, llamaswap
-    from services.fit import evaluate_fit, extract_params_b
 
     models = llamaswap.list_models()
-    cur = next((m for m in models if m.get("role") == "hermes"), None)
+    brain = _active_brain_name()              # z.B. "fast" (Alias) oder ein Modellname
-    cur_ver = _hermes_version(cur["name"]) if cur else None
+    bl = brain.lower()
+    cur = next((m for m in models if (m.get("role") or "").lower() == bl), None) \
+        or next((m for m in models if bl in (m["name"] or "").lower()), None)
     cur_params = (cur.get("capabilities") or {}).get("params_b") if cur else None
     current = None
     if cur:
-        current = {"name": cur["name"], "filename": cur.get("filename"),
+        current = {"name": cur["name"], "alias": brain, "filename": cur.get("filename"),
                    "params_b": cur_params, "quant": cur.get("quant"),
-                   "size_bytes": cur.get("size_bytes"), "version": cur_ver,
+                   "size_bytes": cur.get("size_bytes"),
                    "gguf_path": cur.get("gguf_path"), "incomplete": cur.get("incomplete")}
 
-    ram = psutil.virtual_memory().total / (1024 ** 3)
+    # Fit-Check: passt das (immer warme) Hirn + das größte on-demand-Modell zusammen ins Budget?
-    best = None
-    try:
-        cands = []
-        for r in discover._fetch_author_models("NousResearch"):
-            rid = r.get("id", "")
-            if "hermes" not in rid.lower():
-                continue
-            pb = extract_params_b(rid)
-            fit = evaluate_fit(pb, "Q4_K_M", 8192, ram, name=rid)
-            if fit["level"] == "too_tight":
-                continue
-            cands.append({"repo": rid, "name": rid.split("/")[-1],
-                          "version": _hermes_version(rid) or 0.0, "params_b": pb,
-                          "downloads": int(r.get("downloads") or 0), "fit": fit})
-        # neueste Hermes-Version zuerst, dann größer/fähiger, dann beliebter
-        cands.sort(key=lambda c: (c["version"], c["params_b"], c["downloads"]), reverse=True)
-        best = cands[0] if cands else None
-    except Exception:
-        log.debug("hermes_brain_info: HF-Abfrage fehlgeschlagen", exc_info=True)
-
-    update = False
-    if best is not None:
-        if cur_ver is None:
-            update = True
-        elif best["version"] > cur_ver:
-            update = True
-        elif best["version"] == cur_ver and best["params_b"] > (cur_params or 0) * 1.05:
-            update = True
-        # gleiche Datei schon installiert? dann kein Update
-        if current and best["repo"].split("/")[-1].lower() in (current["name"] or "").lower():
-            update = False
-
-    # Fit-Check: passt das EMPFOHLENE Brain als Always-On noch ins Budget, sodass das
-    # größte on-demand-Modell daneben lädt? (Brain muss immer resident sein.)
     budget = None
     try:
         from services.budget import footprint_gb, gtt_budget_gb
-        from services.fit import estimate_memory_gb
         groups = llamaswap.list_groups()
         persist = set()
         for g in groups.values():
@@ -89,13 +69,7 @@ def hermes_brain_info() -> dict:
                 persist.update(g.get("members") or [])
 
         cur_name = cur["name"] if cur else None
-        brain_ctx = int((cur.get("ctx") if cur else None) or 32768)
+        brain_gb = footprint_gb(cur) if cur else 0.0
-        if best:
-            brain_gb = estimate_memory_gb(float(best["params_b"]), "Q4_K_M", brain_ctx)
-        elif cur:
-            brain_gb = footprint_gb(cur)
-        else:
-            brain_gb = 0.0
         # voller Always-Warm-Footprint (alle persist, Brain=Empfehlung) — nur Info
         warm = brain_gb + sum(footprint_gb(m) for m in models
                               if m["name"] in persist and m["name"] != cur_name)
@@ -114,7 +88,7 @@ def hermes_brain_info() -> dict:
     except Exception:
         log.debug("hermes_brain_info: Budget-Berechnung fehlgeschlagen", exc_info=True)
 
-    return {"current": current, "recommended": best, "update_available": update, "budget": budget}
+    return {"current": current, "recommended": None, "update_available": False, "budget": budget}
 
 
 def _reach(url: str, path: str = "") -> bool: