feat(models): auto-mode intent remap + real coder models in preference

Settings "Auto model routing" picks which installed model fires for chat vs
coding intent when no model is pinned (auto_chat_model / auto_code_model).
_auto_select_model honors the remap; _MODEL_PREFERENCE["code"] prefers real
coder models first.

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

synapse/main.py

@@ -104,6 +104,11 @@ async def _auto_select_model(message: str = "") -> str:
         if s.get("model"):
             return s["model"]
         intent = _detect_intent(message) if message else "chat"
+        # Auto-mode remap: a configured model for this intent fires first;
+        # otherwise fall back to the built-in preference list.
+        remap = s.get(f"auto_{intent}_model")
+        if remap:
+            return remap
         return await get_ollama_manager().select_best_model(intent)
     except Exception:
         return DEFAULT_CHAT_MODEL

synapse/memory/store.py

@@ -986,6 +986,10 @@ class PersistentMemoryStore:
     # -----------------------------
     _SETTINGS_DEFAULTS: Dict[str, Any] = {
         "model": "",
+        # Auto-mode routing overrides (blank = built-in preference). When no model
+        # is pinned, a message's detected intent picks which of these fires first.
+        "auto_chat_model": "",
+        "auto_code_model": "",
         # Qwen3-style reasoning. Off by default: the hidden <think> block is pure
         # latency for chat/memory. Turn on for hard multi-step problems.
         "think": False,

synapse/ollama_manager.py

@@ -162,7 +162,7 @@ def _detect_gpu_backend() -> tuple[str, dict]:
 # tail differs by task. Prefix-matched against installed model names.
 _MODEL_PREFERENCE = {
     "chat": ("qwen2.5:3b", "qwen2.5", "gemma3:1b", "gemma3", "phi3", "phi-3", "gemma2", "gemma"),
-    "code": ("qwen2.5:3b", "qwen2.5", "gemma3:1b", "gemma3", "phi3", "phi-3", "codellama", "deepseek-coder", "codegemma"),
+    "code": ("qwen2.5-coder", "qwen3-coder", "deepseek-coder", "codellama", "codegemma", "qwen2.5:3b", "qwen2.5", "gemma3", "phi3", "phi-3"),
 }