fix: revert to env var approach for camel-ai LLM config

model_config_dict passes values to create() call, not client constructor.
api_key in model_config_dict causes 'unexpected keyword argument' error.
Revert to env vars: OPENAI_API_KEY + OPENAI_BASE_URL (camel-ai reads both).

backend/scripts/run_parallel_simulation.py

@@ -1031,17 +1031,13 @@ def create_model(config: Dict[str, Any], use_boost: bool = False):
     
     print(f"{config_label} model={llm_model}, base_url={llm_base_url[:40] if llm_base_url else '默认'}...")
     
-    # Build model config dict for camel-ai
-    model_config = {}
-    if llm_api_key:
-        model_config["api_key"] = llm_api_key
-    if llm_base_url:
-        model_config["base_url"] = llm_base_url
+    # camel-ai reads OPENAI_API_KEY from env automatically
+    # For base_url, we must also set it as env var
+    # (model_config_dict passes values to create() call, not client constructor)
     
     return ModelFactory.create(
         model_platform=ModelPlatformType.OPENAI,
         model_type=llm_model,
-        model_config_dict=model_config,
     )