x

app/config.py

@@ -18,6 +18,7 @@ class Settings(BaseModel):
     openai_host: str = "0.0.0.0"
     openai_port: int = 8001
     public_model_name: str = "Qwen_local_model"
+    reasoning_enabled: bool = False
     model_root: str = "/opt/model"
     offline_mode: bool = True
     max_model_len: int = 8192
@@ -46,6 +47,7 @@ def get_settings() -> Settings:
         openai_host=runtime["openai_host"],
         openai_port=runtime["openai_port"],
         public_model_name=runtime["public_model_name"],
+        reasoning_enabled=runtime["reasoning_enabled"],
         model_root=runtime["model_root"],
         offline_mode=runtime["offline_mode"],
         api_key=runtime["api_key"],

app/model_catalog.py

@@ -70,6 +70,7 @@ def resolve_runtime_settings(content: dict[str, Any]) -> dict[str, Any]:
         "openai_port": _to_int(openai_service.get("port"), 8001),
         "public_model_name": _to_str(content.get("public_model_name"), "Qwen_local_model"),
         "api_key": str(content.get("api_key", "")).strip() or None,
+        "reasoning_enabled": _to_bool(content.get("reasoning_enabled"), False),
         "tensor_parallel_size": _to_int(content.get("tensor_parallel_size"), 2),
         "dtype": str(content.get("dtype", "bfloat16")),
         "revision": str(content.get("revision", "")).strip() or None,
@@ -103,6 +104,7 @@ def resolve_model_profile(
         "served_model_name": profile.get("served_model_name", model_key),
         "dtype": _to_str(profile.get("dtype")),
         "quantization": _to_str(profile.get("quantization")),
+        "reasoning_parser": _to_str(profile.get("reasoning_parser")),
         "max_model_len": _to_int(profile.get("ctx"), 8192),
         "max_num_seqs": _to_int(profile.get("max_num_seqs"), 64),
         "max_tokens": _to_int(profile.get("max_tokens"), 4096),

app/start_openai.py

@@ -19,9 +19,11 @@ def build_command() -> list[str]:
     host = str(runtime["openai_host"])
     port = str(runtime["openai_port"])
     public_model_name = str(runtime["public_model_name"]).strip()
+    reasoning_enabled = bool(runtime["reasoning_enabled"])
     api_key = runtime["api_key"] or ""
     dtype = str(updates["dtype"] or runtime["dtype"])
     quantization = str(updates["quantization"] or "").strip()
+    reasoning_parser = str(updates["reasoning_parser"] or "").strip()
     revision = runtime["revision"] or ""
     cmd = [
         sys.executable,
@@ -54,6 +56,8 @@ def build_command() -> list[str]:
         cmd.append("--enable-auto-tool-choice")
     if updates["tool_call_parser"]:
         cmd.extend(["--tool-call-parser", str(updates["tool_call_parser"])])
+    if reasoning_enabled and reasoning_parser:
+        cmd.extend(["--reasoning-parser", reasoning_parser])
     if quantization:
         cmd.extend(["--quantization", quantization])
     if revision: