Update tools.py

tools.py

@@ -2,7 +2,6 @@ import tempfile
 import requests
 import os
 
-from time import sleep
 from urllib.parse import urlparse
 from typing import Optional, List
 import yt_dlp
@@ -10,17 +9,17 @@ import imageio
 
 from PIL import Image
 from dotenv import load_dotenv
-from gaia_benchmark.tools import tool
 import whisper
 
+# ✅ Define local @tool decorator
+def tool(func):
+    func.is_tool = True
+    return func
+
 load_dotenv()
 
 @tool
 def use_vision_model(question: str, images: List[Image.Image]) -> str:
-    """
-    Use a Vision Model to answer a question about a set of images. 
-    This stub exists for potential future Gemini or Mistral multimodal integration.
-    """
     return "Vision model is not available for Mistral. Please integrate a separate endpoint for image analysis."
 
 @tool
@@ -39,7 +38,7 @@ def youtube_frames_to_images(url: str, sample_interval_seconds: int = 5) -> List
             'force_ipv4': True,
         }
         with yt_dlp.YoutubeDL(ydl_opts) as ydl:
-            info = ydl.extract_info(url, download=True)
+            ydl.extract_info(url, download=True)
 
         video_path = next((os.path.join(tmpdir, f) for f in os.listdir(tmpdir) if f.endswith('.mp4')), None)
         reader = imageio.get_reader(video_path)
@@ -114,8 +113,8 @@ def youtube_transcribe(url: str) -> str:
         with yt_dlp.YoutubeDL(ydl_opts) as ydl:
             ydl.extract_info(url, download=True)
         audio_path = next((os.path.join(tmpdir, f) for f in os.listdir(tmpdir) if f.endswith('.wav')), None)
-        return whisper.load_model("small").transcribe(audio_path)['text']
+        return model.transcribe(audio_path)['text']
 
 @tool
 def transcribe_audio(audio_file_path: str) -> str:
-    return whisper.load_model("small").transcribe(audio_file_path)['text']
+    return whisper.load_model("small").transcribe(audio_file_path)['text']