Use function-tool schema for non-reasoning OpenAI models

OpenAI's Responses API rejects tools[i].type="custom" on non-reasoning
models like gpt-4o (400 with code=unknown_parameter, param=tools).
Strix's SDK-native Filesystem capability registers CustomTool entries
by default, so a bare STRIX_LLM=gpt-4o run failed at the first tool
invocation even though warm-up (a tool-less call) succeeded.

uses_chat_completions_tool_schema now consults
litellm.model_cost[<name>].supports_reasoning for OpenAI routes and
flips to the chat-completions function-tool schema for models that
don't carry the reasoning flag. Same registry-lookup pattern as
is_known_openai_bare_model. Non-OpenAI prefixes and configs with
LLM_API_BASE are unchanged (still function tools).

strix/config/models.py

@@ -117,7 +117,17 @@ def uses_chat_completions_tool_schema(model_name: str, settings: Settings) -> bo
     model = model_name.strip().lower()
     if "/" in model and not model.startswith("openai/"):
         return True
-    return bool(settings.llm.api_base)
+    if settings.llm.api_base:
+        return True
+    return not _supports_responses_custom_tools(model_name)
+
+
+def _supports_responses_custom_tools(model_name: str) -> bool:
+    import litellm
+
+    name = model_name.strip().lower().removeprefix("openai/")
+    entry = litellm.model_cost.get(name) or {}
+    return bool(entry.get("supports_reasoning"))
 
 
 def is_known_openai_bare_model(model_name: str) -> bool: