Upload tool

app.py

@@ -0,0 +1,7 @@
+from smolagents import launch_gradio_demo
+from typing import Optional
+from tool import HFSpeech2TextFromFile
+
+tool = HFSpeech2TextFromFile()
+
+launch_gradio_demo(tool)

requirements.txt

@@ -0,0 +1,2 @@
+smolagents
+requests

tool.py

@@ -0,0 +1,26 @@
+from smolagents.tools import Tool
+import requests
+
+class HFSpeech2TextFromFile(Tool):
+    name = "HFSpeech2TextFromFile"
+    description = "This is a tool that returns text transcription from an audio          file path, calling huggingface-api under the hood."
+    inputs = {'filepath': {'type': 'string', 'description': 'Path to audio file'}, 'hf_token': {'type': 'string', 'description': 'your huggingface token to call huggingface api'}, 'model': {'type': 'string', 'description': "optional, the model to call. Optional. Defaults to                 'whisper-small.en'", 'nullable': True}}
+    output_type = "string"
+
+    def forward(self, filepath: str, hf_token: str, \
+                model: str = "whisper-small.en") -> str:
+        import requests
+
+        with open(filepath, "rb") as f:
+            data = f.read()
+
+        response = requests.post(
+            f"https://api-inference.huggingface.co/models/openai/{model}", 
+            headers={"Authorization": f"Bearer {hf_token}"}, 
+            data=data
+        )
+
+        return response.json()['text']
+
+    def __init__(self, *args, **kwargs):
+        self.is_initialized = False