From aa077f60e64dcef84349cb3674578dee7c3f89d0 Mon Sep 17 00:00:00 2001
From: duffyduck <info@hacker-net.de>
Date: Mon, 11 May 2026 22:36:14 +0200
Subject: [PATCH] fix(diagnostic+brain): Sprachmodell-Einstellung auf
 runtime.json umgestellt
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

War kaputt nach OpenClaw-Abriss: handleGetModel/handleSetModel haben gegen
aria-core (dockerExec + node-script in den Container) gearbeitet, der gibt's
nicht mehr.

diagnostic/server.js
  - handleGetModel/handleSetModel lesen/schreiben jetzt brainModel in
    /shared/config/runtime.json
  - RUNTIME_CONFIG_FIELDS um "brainModel" erweitert
  - Tote Variante (findSettingsFile + base64-node-script) komplett raus

aria-brain/proxy_client.py
  - Liest brainModel aus runtime.json beim Container-Start
  - Fallback: BRAIN_MODEL env → "claude-sonnet-4" Default
  - Bei Aenderung in Diagnostic: aria-brain restarten damit's greift
    (Hinweis steht in der UI)

diagnostic/index.html
  - Section "Model" → "Sprachmodell (Brain)"
  - Hinweis-Block mit Default-Erklaerung und Restart-Hinweis
  - Modelle: claude-sonnet-4 (default), claude-opus-4, claude-haiku-4-5

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---
 aria-brain/proxy_client.py | 21 ++++++++++++++++++-
 diagnostic/index.html      | 14 ++++++++++---
 diagnostic/server.js       | 43 +++++++++++++++++++-------------------
 3 files changed, 53 insertions(+), 25 deletions(-)

diff --git a/aria-brain/proxy_client.py b/aria-brain/proxy_client.py
index 8bf7a8e..d747a07 100644
--- a/aria-brain/proxy_client.py
+++ b/aria-brain/proxy_client.py
@@ -9,8 +9,10 @@ neuen CLI-Prozess (Cold-Start), das dauert.
 
 from __future__ import annotations
 
+import json
 import logging
 import os
+from pathlib import Path
 from typing import List, Optional
 
 import httpx
@@ -18,11 +20,28 @@ from pydantic import BaseModel
 
 logger = logging.getLogger(__name__)
 
-DEFAULT_MODEL = os.environ.get("BRAIN_MODEL", "claude-sonnet-4")
+RUNTIME_CONFIG_FILE = Path("/shared/config/runtime.json")
+ENV_MODEL = os.environ.get("BRAIN_MODEL", "claude-sonnet-4")
 PROXY_URL = os.environ.get("PROXY_URL", "http://proxy:3456")
 PROXY_TIMEOUT_SEC = float(os.environ.get("PROXY_TIMEOUT_SEC", "300"))
 
 
+def _read_model_from_runtime() -> str:
+    """Liest brainModel aus runtime.json. Fallback: ENV BRAIN_MODEL."""
+    try:
+        if RUNTIME_CONFIG_FILE.exists():
+            data = json.loads(RUNTIME_CONFIG_FILE.read_text(encoding="utf-8"))
+            m = (data.get("brainModel") or "").strip()
+            if m:
+                return m
+    except Exception as exc:
+        logger.warning("runtime.json lesen fehlgeschlagen: %s", exc)
+    return ENV_MODEL
+
+
+DEFAULT_MODEL = _read_model_from_runtime()
+
+
 class Message(BaseModel):
     role: str  # "system" | "user" | "assistant" | "tool"
     content: Optional[str] = None
diff --git a/diagnostic/index.html b/diagnostic/index.html
index 8d514eb..7272334 100644
--- a/diagnostic/index.html
+++ b/diagnostic/index.html
@@ -659,13 +659,21 @@
          selbst (Skills + skill_create Meta-Tool). Es gibt keine granulare
          Permission-Maske, Brain weiss zur Laufzeit welche Tools es hat. -->
 
-    <!-- Model-Einstellungen -->
+    <!-- Brain-Model-Einstellung -->
     <div class="settings-section">
-      <h2>Model</h2>
+      <h2>Sprachmodell (Brain)</h2>
       <div class="card" style="max-width:500px;">
+        <div style="font-size:11px;color:#8888AA;margin-bottom:10px;line-height:1.5;">
+          Welches Claude-Model nutzt das Brain pro Anfrage. Wert wird in
+          <code>/shared/config/runtime.json</code> als <code>brainModel</code> persistiert.
+          Bei Aenderung: <strong>aria-brain restarten</strong> (Reparatur-Section oben), damit's greift.
+          <br><br>
+          Verfuegbar via Proxy: <code>claude-sonnet-4</code> (Default — schnell, gut),
+          <code>claude-opus-4</code> (langsam, smarter), <code>claude-haiku-4-5</code> (sehr schnell, kleiner Kontext).
+        </div>
         <div style="display:flex;align-items:center;gap:8px;margin-bottom:8px;">
           <span style="font-size:12px;color:#8888AA;white-space:nowrap;">Aktives Model:</span>
-          <input type="text" id="setting-model" placeholder="z.B. proxy/claude-sonnet-4" style="flex:1;background:#1E1E2E;border:1px solid #333;border-radius:4px;padding:6px 8px;color:#E0E0F0;font-family:inherit;font-size:12px;">
+          <input type="text" id="setting-model" placeholder="claude-sonnet-4" style="flex:1;background:#1E1E2E;border:1px solid #333;border-radius:4px;padding:6px 8px;color:#E0E0F0;font-family:inherit;font-size:12px;">
           <button class="btn secondary" onclick="loadModel()" style="padding:4px 8px;font-size:10px;">Laden</button>
           <button class="btn" onclick="saveModel()" style="padding:4px 8px;font-size:10px;">Setzen</button>
         </div>
diff --git a/diagnostic/server.js b/diagnostic/server.js
index cca1022..6ede9aa 100644
--- a/diagnostic/server.js
+++ b/diagnostic/server.js
@@ -65,6 +65,7 @@ const RUNTIME_CONFIG_FILE = "/shared/config/runtime.json";
 const RUNTIME_CONFIG_FIELDS = [
   "RVS_HOST", "RVS_PORT", "RVS_TLS", "RVS_TOKEN",
   "ARIA_AUTH_TOKEN", "WHISPER_MODEL", "WHISPER_LANGUAGE",
+  "brainModel",
 ];
 function readRuntimeConfig() {
   const envDefaults = {
@@ -2164,42 +2165,42 @@ async function handleRestartSession(clientWs) {
 // Claude Code laeuft mit --dangerously-skip-permissions (Alles oder Nichts).
 // Root-Check wird via CLAUDE_CODE_BUBBLEWRAP=1 in docker-compose.yml umgangen.
 
-// ── Einstellungen: Model ────────────────────────────────
+// ── Einstellungen: Model (Brain) ────────────────────────
+// Liest/schreibt brainModel in /shared/config/runtime.json — Brain liest
+// das beim Container-Start. Bei Aenderung: aria-brain restarten
+// (Einstellungen → Reparatur → 🚨 aria-brain neu).
 
-async function handleGetModel(clientWs) {
+function handleGetModel(clientWs) {
   try {
-    const raw = await dockerExec("aria-core", `echo $DEFAULT_MODEL`);
-    clientWs.send(JSON.stringify({ type: "model_info", model: raw.trim(), info: "Aktuelles Model (ENV)" }));
+    const cfg = readRuntimeConfig();
+    const model = (cfg.brainModel || "").trim() || "claude-sonnet-4";
+    clientWs.send(JSON.stringify({
+      type: "model_info",
+      model,
+      info: "Brain-Model (runtime.json) — bei Aenderung: aria-brain restarten",
+    }));
   } catch (err) {
     clientWs.send(JSON.stringify({ type: "model_info", error: err.message }));
   }
 }
 
-async function handleSetModel(clientWs, model) {
+function handleSetModel(clientWs, model) {
   if (!model || typeof model !== "string") {
     clientWs.send(JSON.stringify({ type: "model_info", error: "Kein Model angegeben" }));
     return;
   }
   try {
-    // Model in Settings speichern (OpenClaw liest das)
-    const settingsPath = await findSettingsFile();
-    const script = [
-      'const fs=require("fs");',
-      `const f="${settingsPath}";`,
-      'let s={};try{s=JSON.parse(fs.readFileSync(f,"utf8"));}catch(e){}',
-      `s.model=${JSON.stringify(model)};`,
-      `const dir=f.substring(0,f.lastIndexOf("/"));`,
-      'try{fs.mkdirSync(dir,{recursive:true});}catch(e){}',
-      'fs.writeFileSync(f,JSON.stringify(s,null,2));',
-    ].join("");
-    const b64 = Buffer.from(script).toString("base64");
-    await dockerExec("aria-core", `echo ${b64} | base64 -d | node`);
-
-    clientWs.send(JSON.stringify({ type: "model_info", model, info: `Model auf "${model}" gesetzt (Neustart noetig)` }));
-    log("info", "server", `Model gesetzt: ${model}`);
+    writeRuntimeConfig({ brainModel: model.trim() });
+    log("info", "server", `Brain-Model gesetzt: ${model.trim()}`);
+    clientWs.send(JSON.stringify({
+      type: "model_info",
+      model: model.trim(),
+      info: `Model auf "${model.trim()}" gesetzt — aria-brain neu starten damit's greift`,
+    }));
   } catch (err) {
     clientWs.send(JSON.stringify({ type: "model_info", error: err.message }));
   }
+  return;
 }
 
 // OpenClaw-Config-Handler entfernt — aria-core ist raus.