Update app.py

app.py

@@ -1,7 +1,9 @@
 import os
 import json
+import tempfile
 import gradio as gr
 from huggingface_hub import InferenceClient
+from gtts import gTTS
 
 try:
     import PyPDF2
@@ -14,25 +16,11 @@ except ImportError:
 # =========================================
 
 MODEL_OPTIONS = {
-    # Serverless-safe models
-    "Zephyr 7B (Serverless)": {
-        "id": "HuggingFaceH4/zephyr-7b-beta",
-        "fallback": False
-    },
-    "Mixtral 8x7B (Serverless)": {
-        "id": "mistralai/Mixtral-8x7B-Instruct-v0.1",
-        "fallback": False
-    },
-
-    # Requires Dedicated Endpoint
-    "Mistral 7B (Endpoint Required)": {
-        "id": "mistralai/Mistral-7B-Instruct-v0.2",
-        "fallback": True
-    }
+    "Zephyr 7B (Serverless)": "HuggingFaceH4/zephyr-7b-beta",
+    "Mixtral 8x7B (Serverless)": "mistralai/Mixtral-8x7B-Instruct-v0.1",
 }
 
-FALLBACK_MODEL_KEY = "Zephyr 7B (Serverless)"
-
+FALLBACK_MODEL = "HuggingFaceH4/zephyr-7b-beta"
 
 PERSONA_PRESETS = {
     "Balanced Assistant": "You are a helpful, intelligent AI assistant.",
@@ -61,6 +49,14 @@ Conduct a behavioral analysis.
 7. Limitations
 Material:
 {user_input}
+""",
+
+    "Explain Code (Natural Voice)": """
+Explain the following code clearly and conversationally,
+as if teaching a developer step-by-step in a calm, natural voice.
+
+Code:
+{user_input}
 """
 }
 
@@ -77,7 +73,6 @@ def get_client(model_id):
     token = os.environ.get("HF_TOKEN")
     if not token:
         raise RuntimeError("HF_TOKEN not set in Space Secrets.")
-
     return InferenceClient(model=model_id, token=token)
 
 
@@ -102,6 +97,13 @@ def extract_text(file):
     return ""
 
 
+def text_to_speech(text):
+    tts = gTTS(text=text)
+    temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
+    tts.save(temp_file.name)
+    return temp_file.name
+
+
 def run_model(client, messages, temperature):
     response = client.chat_completion(
         messages=messages,
@@ -117,19 +119,17 @@ def run_model(client, messages, temperature):
 # Core Logic
 # =========================================
 
-def generate_response(message, history, model_label, persona_label, task_mode, uploaded_file):
+def generate_response(message, history, model_label, persona_label, task_mode, uploaded_file, voice_enabled):
 
     if not message:
-        return history
+        return history, None
 
     try:
         message = message[:MAX_INPUT_CHARS]
         history = history[-MAX_HISTORY_PAIRS * 2:]
 
-        model_config = MODEL_OPTIONS[model_label]
-        model_id = model_config["id"]
-
-        system_prompt = PERSONA_PRESETS[persona_label]
+        model_id = MODEL_OPTIONS[model_label]
+        system_prompt = PERSONA_PRESETS.get(persona_label, "")
         temperature = 0.4 if "Forensic" in persona_label else 0.7
 
         file_text = extract_text(uploaded_file)
@@ -143,29 +143,25 @@ def generate_response(message, history, model_label, persona_label, task_mode, u
         formatted_input = TASK_MODES[task_mode].format(user_input=message)
         messages.append({"role": "user", "content": formatted_input})
 
-        # Primary model attempt
         try:
             client = get_client(model_id)
             answer = run_model(client, messages, temperature)
-
-        # Automatic fallback
         except Exception:
-            fallback_model = MODEL_OPTIONS[FALLBACK_MODEL_KEY]["id"]
-            fallback_client = get_client(fallback_model)
-
-            answer = (
-                "⚠️ Selected model unavailable. Fallback model used.\n\n"
-                + run_model(fallback_client, messages, temperature)
-            )
+            fallback_client = get_client(FALLBACK_MODEL)
+            answer = "⚠️ Fallback model used.\n\n" + run_model(fallback_client, messages, temperature)
 
         history.append({"role": "user", "content": message})
         history.append({"role": "assistant", "content": answer})
 
-        return history
+        audio_file = None
+        if voice_enabled:
+            audio_file = text_to_speech(answer)
+
+        return history, audio_file
 
     except Exception as e:
         history.append({"role": "assistant", "content": f"⚠️ ERROR:\n\n{str(e)}"})
-        return history
+        return history, None
 
 
 # =========================================
@@ -173,7 +169,7 @@ def generate_response(message, history, model_label, persona_label, task_mode, u
 # =========================================
 
 def clear_chat():
-    return []
+    return [], None
 
 def export_chat(history):
     file_path = "conversation.json"
@@ -188,7 +184,7 @@ def export_chat(history):
 
 with gr.Blocks(theme=gr.themes.Soft(), title="Omniscient IRIS") as demo:
 
-    gr.Markdown("## Omniscient IRIS — Adaptive Analysis Assistant")
+    gr.Markdown("## Omniscient IRIS — Voice Enabled")
 
     with gr.Row():
 
@@ -200,6 +196,8 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Omniscient IRIS") as demo:
                 show_label=False
             )
 
+            audio_output = gr.Audio(autoplay=True)
+
             send_btn = gr.Button("Send", variant="primary")
             clear_btn = gr.Button("Clear")
 
@@ -207,7 +205,7 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Omniscient IRIS") as demo:
 
             model_selector = gr.Dropdown(
                 choices=list(MODEL_OPTIONS.keys()),
-                value=FALLBACK_MODEL_KEY,
+                value=list(MODEL_OPTIONS.keys())[0],
                 label="Model"
             )
 
@@ -220,9 +218,11 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Omniscient IRIS") as demo:
             task_selector = gr.Dropdown(
                 choices=list(TASK_MODES.keys()),
                 value="General Chat",
-                label="Analysis Mode"
+                label="Mode"
             )
 
+            voice_toggle = gr.Checkbox(label="Enable Voice Output", value=True)
+
             file_upload = gr.File(
                 label="Upload Document (txt, pdf, md)",
                 file_types=[".txt", ".pdf", ".md"]
@@ -234,19 +234,19 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Omniscient IRIS") as demo:
 
     send_btn.click(
         generate_response,
-        inputs=[msg, state, model_selector, persona_selector, task_selector, file_upload],
-        outputs=chatbot,
+        inputs=[msg, state, model_selector, persona_selector, task_selector, file_upload, voice_toggle],
+        outputs=[chatbot, audio_output],
     )
 
     msg.submit(
         generate_response,
-        inputs=[msg, state, model_selector, persona_selector, task_selector, file_upload],
-        outputs=chatbot,
+        inputs=[msg, state, model_selector, persona_selector, task_selector, file_upload, voice_toggle],
+        outputs=[chatbot, audio_output],
     )
 
     clear_btn.click(
         fn=clear_chat,
-        outputs=chatbot,
+        outputs=[chatbot, audio_output],
     ).then(lambda: [], outputs=state)
 
     export_btn.click(