import math
from typing import Optional, Tuple, Literal
from smolagents import tool



@tool
def extract_text_from_audio(file_path : str) -> str:
    """given a path to an audio file, it extract and returns the text contained in it as a string"""

    import speech_recognition as sr
    r = sr.Recognizer()
    with sr.AudioFile(file_path) as source:
        # listen for the data (load audio to memory)
        audio_data = r.record(source)
        # recognize (convert from speech to text)
        text = r.recognize_google(audio_data)
    return text


class TestAgent:
    def __init__(self):
        
        # import code agent and basic tool from smolagent
        from smolagents import CodeAgent, OpenAIServerModel, DuckDuckGoSearchTool, FinalAnswerTool, VisitWebpageTool, MCPClient

        # import additional tool from langchain @ https://docs.langchain.com/oss/python/integrations/tools
        #from langchain_community.agent_toolkits import load_tools
        from langchain_community.agent_toolkits.load_tools import load_tools

        from smolagents import Tool
        wikipedia_tool = Tool.from_langchain(load_tools(["wikipedia"])[0])
        wikipedia_tool.top_k_results=3

        # import tools from MCP servers @ https://github.com/mcp
        #from mcp import StdioServerParameters
        #server_parameters = StdioServerParameters(command="uvx",
        #                                          args=["--quiet", "youtubeqa@0.2.1"],
        #                                          env={"UV_PYTHON": "3.12", **os.environ},
        #                                         )
        #youtube_tools = MCPServerTool(server_params=server_parameters)

        model = OpenAIServerModel(model_id="gpt-4o")
        #model = InferenceClientModel("Qwen/Qwen2.5-Coder-32B-Instruct")
        # Instantiate the agent
        self.agent = CodeAgent(
            tools=[extract_text_from_audio,        # homemade tool
                   DuckDuckGoSearchTool(),          # basic tools from smolagent
                   VisitWebpageTool(),
                   wikipedia_tool,                  # tool from langchain with extra parmaeters
                   #youtube_tools,                   # tool from MCP server
                   FinalAnswerTool()],
            additional_authorized_imports=["pandas","markdownify","requests"],
            model=model,
            max_steps=3,
            verbosity_level=2,
            use_structured_outputs_internally=True.   # V3. Adds structure
        )
        # V3. add Guidance
        prompt_for_guidance = "\n10. Provide the answer axactly as it is asked, be concise and precise\n\nNow Begin!"
        self.agent.prompt_templates['system_prompt'] = self.agent.prompt_templates['system_prompt'] + prompt_for_guidance

    def __call__(self, question: str) -> str:

        print(f"Agent received question (first 50 chars): {question[:50]}...")
        answer = self.agent.run(question)
        print(f"Agent returning his answer: {answer}")
        return answer