Spaces:
Sleeping
Sleeping
IngoTB303
Enhance agent functionality: add new agents for audio, visual, and Python tasks; update question fetching to handle attachments; improve tool initialization and requirements.
9e58d03
| from smolagents import tool, Tool | |
| from tavily import TavilyClient | |
| from smolagents import DuckDuckGoSearchTool, GoogleSearchTool, VisitWebpageTool, WikipediaSearchTool, PythonInterpreterTool, FinalAnswerTool | |
| from dotenv import load_dotenv | |
| import os | |
| load_dotenv() # take environment variables | |
| # init tools | |
| duck_search = DuckDuckGoSearchTool() | |
| google_search = GoogleSearchTool() | |
| visit_page = VisitWebpageTool() | |
| wiki_search = WikipediaSearchTool() | |
| do_python = PythonInterpreterTool() | |
| final_answer = FinalAnswerTool() | |
| tavily_search = TavilyClient() | |
| speech_to_text_tool = Tool.from_space("hf-audio/whisper-large-v3-turbo", | |
| name="speech_to_text_tool", | |
| description="""Converts audio to text by providing a file or url. | |
| Use the with the command speed_to_text_tool(filename). | |
| Example: 'speech_to_text_tool("cache\\audio.mp3'""", | |
| api_name="/predict") | |
| visual_qa_tool = Tool.from_space("sitammeur/PicQ", | |
| name="visual_qa_tool", | |
| description="""Can answer question about a provided image. | |
| Use it with visual_qa_tool(question=<question>, image=<image filename>). | |
| Example visual_qa_tool(question='How many items are in the image?', image='cache\\image.png').""", | |
| api_name="/predict") | |
| # image_generation_tool = Tool.from_space( | |
| # "black-forest-labs/FLUX.1-schnell", | |
| # name="image_generator", | |
| # description="Generate an image from a prompt" | |
| # ) | |
| # image = image_generation_tool("A sunny beach") | |
| # print(image) | |
| ## Testing the tools | |
| # print(duck_search.forward("smolagents")) # connect error, does not take proxy | |
| # print(google_search.forward("smolagents")) | |
| # print(visit_page.forward("https://www.wikipedia.de")) | |
| # print(wiki_search.forward("Python_(programming_language)")) | |
| # print(do_python('print("hello world")')) | |
| # print(tavily_search.search("smolagents")) | |
| # print(speech_to_text_tool("cache\\1f975693-876d-457b-a649-393859e79bf3.mp3")) | |
| # print(visual_qa_tool(question="Name all black and white figures with each coordinates, e. g. black king on g8", image="cache\\cca530fc-4052-43b2-b130-b30968d8aa44.png")) |