[Bug] Mismatch between api_list and tool_descriptions in class rapidapi_wrapper Initialization

[yuyq18]

In the initialization code of the rapidapi_wrapper class located in toolbench/inference/Downstream_tasks/rapidapi.py (lines 77-82):

if self.retriever is not None:
    query_json = self.retrieve_rapidapi_tools(self.input_description, args.retrieved_api_nums, args.tool_root_dir)
    data_dict = self.fetch_api_json(query_json)
    tool_descriptions = self.build_tool_description(data_dict)
else:
    data_dict = self.fetch_api_json(query_json)

When self.retriever is None, the data_dict is passed through the fetch_api_json function, but tool_descriptions is not built from the filtered data_dict (unlike the code in the if branch). This results in a mismatch between api_list and tool_descriptions when constructing the function lists later on (lines 84-85).

Here is the relevant code in toolbench/inference/Downstream_tasks/rapidapi.py (lines 84-91):

for k,api_json in enumerate(data_dict["api_list"]):
    standard_tool_name = tool_descriptions[k][0]
    openai_function_json,cate_name, pure_api_name = self.api_json_to_openai_json(api_json,standard_tool_name)
    self.functions.append(openai_function_json)

    self.api_name_reflect[openai_function_json["name"]] = pure_api_name
    self.tool_names.append(standard_tool_name)
    self.cate_names.append(cate_name)

This bug causes significant issues with prompt construction and function calling to the RapidAPI server, as the mismatch between api_list and tool_descriptions can lead to wrong function lists and subsequent wrong API requests.