Create app.py

app.py

@@ -0,0 +1,44 @@
+import gradio as gr
+from transformers import pipeline
+import os
+
+# 1. Load the classifier from your Hugging Face Repo
+# This replaces the /content/drive path
+model_repo = "MakD1227/afriberta-hsd-model"
+classifier = pipeline("text-classification", model=model_repo)
+
+# 2. Prediction function
+def predict_speech(text): 
+    results = classifier(text)
+    # Mapping: LABEL_0 -> Free, LABEL_1 -> Offensive, LABEL_2 -> Hate
+    label_map = {"LABEL_0": "Free (Neutral)", "LABEL_1": "Offensive", "LABEL_2": "Hate"}
+
+    label = results[0]['label']
+    score = results[0]['score']
+
+    return label_map.get(label, label), f"{round(score * 100, 2)}%"
+
+# 3. Gradio Interface
+interface = gr.Interface(
+    fn=predict_speech,
+    inputs=gr.Textbox(
+        lines=2,
+        label="Input Text",
+        placeholder="Enter Amharic or Afan Oromo text..."
+    ),
+    outputs=[
+        gr.Label(label="Classification"),
+        gr.Text(label="Confidence")
+    ],
+    title="Amharic & Afan Oromo Hate Speech Detector",
+    description="Classify text into Free, Offensive, or Hate Speech",
+    article="<p style='text-align: center;'>@2025 Mequanent Degu Belete </p><p style='text-align: center;'>mekuanentde@gmail.com</p><p style='text-align: center;'>SNHCC, Academia Sinica, Taiwan</p>",
+    examples=[
+        ["ኢትዮጵያ ለዘላለም ትኑር"],
+        ["haatee sali shamtuu situ nuu beekaa waa ee baalee"]
+    ]
+)
+
+# Launch (No 'share=True' needed on Hugging Face Spaces)
+if __name__ == "__main__":
+    interface.launch()