Create app.py

app.py

@@ -0,0 +1,50 @@
+import gradio as gr
+
+from app.pipeline import run_text, run_voice
+
+def ui_text(text):
+    # expected: returns dict or tuple; adapt if needed
+    result = run_text(text)
+    return (
+        result["user_text"],
+        result["user_normalized"],
+        result["assistant_text_raw"],
+        result["assistant_text_tts"],
+        result["audio_path"],
+    )
+
+def ui_voice(audio_fp):
+    result = run_voice(audio_fp)
+    return (
+        result["user_text"],
+        result["user_normalized"],
+        result["assistant_text_raw"],
+        result["assistant_text_tts"],
+        result["audio_path"],
+    )
+
+with gr.Blocks() as demo:
+    gr.Markdown("Egyptian Arabic Voice Chatbot")
+
+    with gr.Tab("Text"):
+        inp = gr.Textbox(label="Input (Egyptian Arabic)")
+        out_user = gr.Textbox(label="User Text")
+        out_norm = gr.Textbox(label="Normalized")
+        out_raw = gr.Textbox(label="LLM Raw Response")
+        out_tts = gr.Textbox(label="TTS-ready Response")
+        out_aud = gr.Audio(label="Bot Audio", type="filepath")
+        btn = gr.Button("Send")
+        btn.click(ui_text, inputs=inp, outputs=[out_user, out_norm, out_raw, out_tts, out_aud])
+
+    with gr.Tab("Voice"):
+        v_in = gr.Audio(sources=["microphone", "upload"], type="filepath", label="Record / Upload")
+        v_user = gr.Textbox(label="User Text (ASR)")
+        v_norm = gr.Textbox(label="Normalized")
+        v_raw = gr.Textbox(label="LLM Raw Response")
+        v_tts = gr.Textbox(label="TTS-ready Response")
+        v_out = gr.Audio(label="Bot Audio", type="filepath")
+        v_btn = gr.Button("Send Voice")
+        v_btn.click(ui_voice, inputs=v_in, outputs=[v_user, v_norm, v_raw, v_tts, v_out])
+
+demo.queue()
+demo.launch()