RCaz commited on
Commit
2bcf72e
·
verified ·
1 Parent(s): bea827c

Update agent.py

Browse files
Files changed (1) hide show
  1. agent.py +40 -0
agent.py CHANGED
@@ -18,6 +18,46 @@ def extract_text_from_audio(file_path : str) -> str:
18
  return text
19
 
20
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21
  class TestAgent:
22
  def __init__(self):
23
 
 
18
  return text
19
 
20
 
21
+ @tool
22
+ def extract_text_from_audio(file_path: str) -> str:
23
+ """
24
+ Extract and return text transcription from an audio file using speech recognition.
25
+
26
+ This tool uses Google's speech recognition API to convert spoken audio content
27
+ into text. It supports various audio formats including WAV, AIFF, and FLAC
28
+ (formats supported by the SpeechRecognition library).
29
+
30
+ Args:
31
+ file_path (str): Path to the audio file to be transcribed. The file should
32
+ be in a format compatible with the SpeechRecognition library.
33
+
34
+ Returns:
35
+ str: The extracted text content from the audio file.
36
+
37
+ Raises:
38
+ Exception : the exception
39
+
40
+ Examples:
41
+ >>> extract_text_from_audio("meeting_recording.wav")
42
+ "Hello team, welcome to our weekly meeting..."
43
+
44
+ >>> extract_text_from_audio("/path/to/audio/interview.mp3")
45
+ "Could you please introduce yourself and your background?"
46
+ """
47
+
48
+ import speech_recognition as sr
49
+ r = sr.Recognizer()
50
+ try:
51
+ with sr.AudioFile(file_path) as source:
52
+ # listen for the data (load audio to memory)
53
+ audio_data = r.record(source)
54
+ # recognize (convert from speech to text)
55
+ text = r.recognize_google(audio_data)
56
+ return text
57
+ except Exception as e:
58
+ return e
59
+
60
+
61
  class TestAgent:
62
  def __init__(self):
63