feat: introduce base_url_fixed for local model

agentflow/agentflow/models/initializer.py

@@ -225,9 +225,9 @@ def _load_single_tool(self, root: str, import_path: str, agentflow_dir: str) ->
                             if engine == "Default":
                                 tool_instance = obj()
                             elif engine == "self":
-                                tool_instance = obj(model_string=self.model_string)
+                                tool_instance = obj(model_string=self.model_string, base_url=self.base_url)
                             else:
-                                tool_instance = obj(model_string=engine)
+                                tool_instance = obj(model_string=engine, base_url=self.base_url)
                         else:
                             tool_instance = obj()
 

agentflow/agentflow/models/planner.py

@@ -13,18 +13,18 @@
 class Planner:
     def __init__(self, llm_engine_name: str, llm_engine_fixed_name: str = "dashscope",
                  toolbox_metadata: dict = None, available_tools: List = None,
-                 verbose: bool = False, base_url: str = None, is_multimodal: bool = False,
+                 verbose: bool = False, base_url: str = None, base_url_fixed: str = None, is_multimodal: bool = False,
                  check_model: bool = True, temperature : float = .0):
         self.llm_engine_name = llm_engine_name
         self.llm_engine_fixed_name = llm_engine_fixed_name
         self.is_multimodal = is_multimodal
         # self.llm_engine_mm = create_llm_engine(model_string=llm_engine_name, is_multimodal=False, base_url=base_url, temperature = temperature)
-        self.llm_engine_fixed = create_llm_engine(model_string=llm_engine_fixed_name, is_multimodal=False, temperature = temperature)
-        self.llm_engine = create_llm_engine(model_string=llm_engine_name, is_multimodal=False, base_url=base_url, temperature = temperature)
+        self.llm_engine_fixed = create_llm_engine(model_string=llm_engine_fixed_name, base_url=base_url_fixed, is_multimodal=False, temperature=temperature)
+        self.llm_engine = create_llm_engine(model_string=llm_engine_name, base_url=base_url, is_multimodal=False, temperature=temperature)
         self.toolbox_metadata = toolbox_metadata if toolbox_metadata is not None else {}
         self.available_tools = available_tools if available_tools is not None else []
-
         self.verbose = verbose
+
     def get_image_info(self, image_path: str) -> Dict[str, Any]:
         image_info = {}
         if image_path and os.path.isfile(image_path):

agentflow/agentflow/models/verifier.py

@@ -13,12 +13,12 @@
 class Verifier:
     def __init__(self, llm_engine_name: str, llm_engine_fixed_name: str = "dashscope",
                  toolbox_metadata: dict = None, available_tools: list = None,
-                 verbose: bool = False, base_url: str = None, is_multimodal: bool = False,
+                 verbose: bool = False, base_url: str = None, base_url_fixed: str = None, is_multimodal: bool = False,
                  check_model: bool = True, temperature: float = .0):
         self.llm_engine_name = llm_engine_name
         self.llm_engine_fixed_name = llm_engine_fixed_name
         self.is_multimodal = is_multimodal
-        self.llm_engine_fixed = create_llm_engine(model_string=llm_engine_fixed_name, is_multimodal=False, temperature=temperature)
+        self.llm_engine_fixed = create_llm_engine(model_string=llm_engine_fixed_name, is_multimodal=False, base_url=base_url_fixed, temperature=temperature)
         self.llm_engine = create_llm_engine(model_string=llm_engine_name, is_multimodal=False, base_url=base_url, temperature=temperature)
         self.toolbox_metadata = toolbox_metadata if toolbox_metadata is not None else {}
         self.available_tools = available_tools if available_tools is not None else []

agentflow/agentflow/solver.py

@@ -207,6 +207,7 @@ def construct_solver(llm_engine_name : str = "gpt-4o",
                      verbose : bool = True,
                      vllm_config_path : str = None,
                      base_url : str = None,
+                     base_url_fixed : str = None,
                      temperature: float = 0.0
                      ):
 
@@ -235,6 +236,7 @@ def construct_solver(llm_engine_name : str = "gpt-4o",
         available_tools=initializer.available_tools,
         verbose=verbose,
         base_url=base_url,
+        base_url_fixed=base_url_fixed,
         temperature=temperature
     )
 
@@ -246,6 +248,7 @@ def construct_solver(llm_engine_name : str = "gpt-4o",
         available_tools=initializer.available_tools,
         verbose=verbose,
         base_url=base_url if verifier_engine == llm_engine_name else None,
+        base_url_fixed=base_url_fixed,
         temperature=temperature
     )
 

agentflow/agentflow/tools/base_generator/tool.py

@@ -22,7 +22,7 @@
 class Base_Generator_Tool(BaseTool):
     require_llm_engine = True
 
-    def __init__(self, model_string="gpt-4o-mini"):
+    def __init__(self, model_string="gpt-4o-mini", base_url=None):
         super().__init__(
             tool_name=TOOL_NAME,
             tool_description="A generalized tool that takes query from the user, and answers the question step by step to the best of its ability. It can also accept an image.",
@@ -59,6 +59,7 @@ def __init__(self, model_string="gpt-4o-mini"):
 
         )
         self.model_string = model_string  
+        self.base_url = base_url
         print(f"Initializing Generalist Tool with model: {self.model_string}")
         # multimodal = True if image else False
         multimodal = False
@@ -68,6 +69,7 @@ def __init__(self, model_string="gpt-4o-mini"):
         self.llm_engine = create_llm_engine(
             model_string=self.model_string, 
             is_multimodal=multimodal, 
+            base_url=base_url,
             temperature=0.0, 
             top_p=1.0, 
             frequency_penalty=0.0,