Create openai_speech_to_text_tool.py

tools/openai_speech_to_text_tool.py

@@ -0,0 +1,34 @@
+import os
+import whisper
+from smolagents import Tool
+
+
+class OpenAISpeechToTextTool(Tool):
+    """
+    Tool to convert speech to text using OpenAI's Whisper model.
+
+    Args:
+        audio_path (str): Path to the audio file.
+
+    Returns:
+        str: Transcribed text from the audio file.
+    """
+
+    name = "transcribe_audio"
+    description = "Transcribes audio to text and returns the text"
+    inputs = {
+        "audio_path": {"type": "string", "description": "Path to the audio file"},
+    }
+    output_type = "string"
+
+    def forward(self, audio_path: str) -> str:
+        try:
+            model = whisper.load_model("small")
+
+            if not os.path.exists(audio_path):
+                return f"Error: Audio file not found at {audio_path}"
+
+            result = model.transcribe(audio_path)
+            return result["text"]
+        except Exception as e:
+            return f"Error transcribing audio: {str(e)}"