Create app.py

app.py

@@ -0,0 +1,36 @@
+import gradio as gr
+from transformers import pipeline
+
+print("Loading the Islomov STT model onto Hugging Face servers...")
+
+# Load the model. (HF Free Spaces use CPUs, so we don't specify a cuda device here)
+stt_pipeline = pipeline(
+    "automatic-speech-recognition", 
+    model="islomov/rubaistt_v2_medium"
+)
+
+def transcribe_for_api(audio_filepath):
+    # Failsafe if the API receives an empty request
+    if audio_filepath is None:
+        return "Error: No audio file received by the Space."
+    
+    try:
+        # Pass the audio file directly into the model
+        result = stt_pipeline(audio_filepath)
+        return result["text"].strip()
+    except Exception as e:
+        return f"Error during transcription: {str(e)}"
+
+# Build the Gradio interface
+# We set type="filepath" so Gradio automatically saves the incoming API audio to a temporary file
+interface = gr.Interface(
+    fn=transcribe_for_api,
+    inputs=gr.Audio(type="filepath", label="Input Audio"),
+    outputs=gr.Textbox(label="Uzbek Transcription"),
+    title="b-til.uz STT API Engine",
+    description="This Space processes audio for the b-til.uz language platform."
+)
+
+# Launch the server and enable the API
+if __name__ == "__main__":
+    interface.launch()