refactor -- following guidelines to push tools

tool.py

@@ -153,6 +153,8 @@ class VisualRAGTool(Tool):
     }
     output_type = "string"
 
+    model_name: str = "vidore/colqwen2-v1.0"
+
     def _init_models(self, model_name: str) -> None:
         self.device = "cuda" if torch.cuda.is_available() else "cpu"
         self.model = ColQwen2.from_pretrained(
@@ -163,27 +165,18 @@ class VisualRAGTool(Tool):
             ).eval()
         self.processor = ColQwen2Processor.from_pretrained(model_name)
 
-    def __init__(self, model_name: str = "vidore/colqwen2-v1.0", api_key: str = None, files: List[str] = None, **kwargs):
-        super().__init__(**kwargs)
-        self.model_name = model_name
-        self.api_key = api_key
-
-        self.embds = []
-        self.pages = []
-
-        self.files = files
-
-        self._init_models(self.model_name)
-
-        self.is_initialized = False 
+    def __init__(self, *args, **kwargs):
+        self.is_initialized = False
     
     def setup(self):
         """
         Overwrite this method here for any operation that is expensive and needs to be executed before you start using
         your tool. Such as loading a big model.
         """
-        if self.files:
-            _ = self.index(self.files, self.api_key)
+        self._init_models(self.model_name)
+
+        self.embds = []
+        self.pages = []
 
         self.is_initialized = True
 
@@ -195,7 +188,7 @@ class VisualRAGTool(Tool):
                     'query': "Give the general context about these pages. Give the context in the same language as the documents.",
                     'pages': [Page(image=im) for im in images[max(i-window+1, 0):i+1]],
                     'api_key': api_key,
-                    'system_prompt': DEFAULT_CONTEXT_PROMPT
+                    'system_prompt': CONTEXT_SYSTEM_PROMPT,
                 } for i in range(0, len(images), window)
             ]
             window_contexts = pqdm(args, query_openai, n_jobs=8, argument_type='kwargs')