Spaces:
Sleeping
Sleeping
Update tools.py
Browse files
tools.py
CHANGED
|
@@ -1,16 +1,20 @@
|
|
| 1 |
import tempfile
|
| 2 |
import requests
|
| 3 |
import os
|
| 4 |
-
|
| 5 |
from urllib.parse import urlparse
|
| 6 |
from typing import Optional, List
|
| 7 |
import yt_dlp
|
| 8 |
import imageio
|
| 9 |
-
|
| 10 |
from PIL import Image
|
| 11 |
-
from dotenv import load_dotenv
|
| 12 |
-
from gaia_benchmark.tools import tool
|
| 13 |
import whisper
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
|
| 15 |
load_dotenv()
|
| 16 |
|
|
@@ -34,11 +38,11 @@ def youtube_frames_to_images(url: str, sample_interval_seconds: int = 5) -> List
|
|
| 34 |
'force_ipv4': True,
|
| 35 |
}
|
| 36 |
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
| 37 |
-
|
| 38 |
|
| 39 |
video_path = next((os.path.join(tmpdir, f) for f in os.listdir(tmpdir) if f.endswith('.mp4')), None)
|
| 40 |
reader = imageio.get_reader(video_path)
|
| 41 |
-
fps = reader.get_meta_data().get('fps')
|
| 42 |
frame_interval = int(fps * sample_interval_seconds)
|
| 43 |
images = [Image.fromarray(frame) for idx, frame in enumerate(reader) if idx % frame_interval == 0]
|
| 44 |
reader.close()
|
|
@@ -107,28 +111,11 @@ def youtube_transcribe(url: str) -> str:
|
|
| 107 |
'force_ipv4': True,
|
| 108 |
}
|
| 109 |
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
| 110 |
-
ydl.
|
|
|
|
| 111 |
audio_path = next((os.path.join(tmpdir, f) for f in os.listdir(tmpdir) if f.endswith('.wav')), None)
|
| 112 |
return model.transcribe(audio_path)['text']
|
| 113 |
|
| 114 |
@tool
|
| 115 |
def transcribe_audio(audio_file_path: str) -> str:
|
| 116 |
-
return whisper.load_model("small").transcribe(audio_file_path)['text']
|
| 117 |
-
|
| 118 |
-
# 🔧 Stub Tools for Mistral Compatibility
|
| 119 |
-
|
| 120 |
-
@tool
|
| 121 |
-
def web_search(query: str) -> str:
|
| 122 |
-
return f"(Stub) Web search results for: {query}"
|
| 123 |
-
|
| 124 |
-
@tool
|
| 125 |
-
def wikipedia_search(topic: str) -> str:
|
| 126 |
-
return f"(Stub) Wikipedia summary for: {topic}"
|
| 127 |
-
|
| 128 |
-
@tool
|
| 129 |
-
def visit_webpage(url: str) -> str:
|
| 130 |
-
return f"(Stub) Visited webpage at: {url}"
|
| 131 |
-
|
| 132 |
-
@tool
|
| 133 |
-
def final_answer(response: str) -> str:
|
| 134 |
-
return f"(Stub) Final answer: {response}"
|
|
|
|
| 1 |
import tempfile
|
| 2 |
import requests
|
| 3 |
import os
|
|
|
|
| 4 |
from urllib.parse import urlparse
|
| 5 |
from typing import Optional, List
|
| 6 |
import yt_dlp
|
| 7 |
import imageio
|
|
|
|
| 8 |
from PIL import Image
|
|
|
|
|
|
|
| 9 |
import whisper
|
| 10 |
+
from dotenv import load_dotenv
|
| 11 |
+
|
| 12 |
+
# Fallback tool decorator if gaia_benchmark.tools is not available
|
| 13 |
+
try:
|
| 14 |
+
from gaia_benchmark.tools import tool
|
| 15 |
+
except ImportError:
|
| 16 |
+
def tool(func):
|
| 17 |
+
return func
|
| 18 |
|
| 19 |
load_dotenv()
|
| 20 |
|
|
|
|
| 38 |
'force_ipv4': True,
|
| 39 |
}
|
| 40 |
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
| 41 |
+
ydl.download([url])
|
| 42 |
|
| 43 |
video_path = next((os.path.join(tmpdir, f) for f in os.listdir(tmpdir) if f.endswith('.mp4')), None)
|
| 44 |
reader = imageio.get_reader(video_path)
|
| 45 |
+
fps = reader.get_meta_data().get('fps', 25)
|
| 46 |
frame_interval = int(fps * sample_interval_seconds)
|
| 47 |
images = [Image.fromarray(frame) for idx, frame in enumerate(reader) if idx % frame_interval == 0]
|
| 48 |
reader.close()
|
|
|
|
| 111 |
'force_ipv4': True,
|
| 112 |
}
|
| 113 |
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
|
| 114 |
+
ydl.download([url])
|
| 115 |
+
|
| 116 |
audio_path = next((os.path.join(tmpdir, f) for f in os.listdir(tmpdir) if f.endswith('.wav')), None)
|
| 117 |
return model.transcribe(audio_path)['text']
|
| 118 |
|
| 119 |
@tool
|
| 120 |
def transcribe_audio(audio_file_path: str) -> str:
|
| 121 |
+
return whisper.load_model("small").transcribe(audio_file_path)['text']
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|