Update app.py

app.py

@@ -128,35 +128,12 @@ def generate_answer(text_input):
         if not text_input or text_input.strip() == "":
             return "No input provided"
         
-        # Web search
-        search_results = search_web(text_input, max_results=2)
-        
-        # Generate answer
-        messages = [
-            {"role": "system", "content": "You are a helpful assistant. Answer briefly using provided context. Keep responses under 40 words."},
-            {"role": "user", "content": f"Context:\n{search_results}\n\nQuestion: {text_input}\n\nAnswer:"}
-        ]
+        # Get the last chunk from streaming
+        final_answer = ""
+        for chunk in generate_answer_stream(text_input):
+            final_answer = chunk
         
-        text = tokenizer.apply_chat_template(
-            messages,
-            tokenize=False,
-            add_generation_prompt=True
-        )
-        
-        inputs = tokenizer([text], return_tensors="pt").to("cpu")
-        
-        with torch.no_grad():
-            outputs = model.generate(
-                **inputs,
-                max_new_tokens=80,
-                temperature=0.2,
-                do_sample=True,
-                top_p=0.85,
-                pad_token_id=tokenizer.eos_token_id
-            )
-        
-        response = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
-        return response.strip()
+        return final_answer
         
     except Exception as e:
         return f"Error: {str(e)}"
@@ -255,26 +232,26 @@ with gr.Blocks(title="Fast Q&A - Streaming Enabled", theme=gr.themes.Soft()) as
             inputs=text_input
         )
     
-    # API endpoints for Pluely
+    # API endpoints for Pluely - Using Gradio's built-in interface functions
     with gr.Tab("🔌 Pluely Integration"):
         gr.Markdown("""
         ## Dedicated Endpoints for Pluely
         
-        ### 1. STT Endpoint (Audio Transcription) - Non-streaming
+        ### 1. STT Endpoint (Audio Transcription)
         ```
         curl -X POST https://archcoder-basic-app.hf.space/call/transcribe_stt \\
           -H "Content-Type: application/json" \\
           -d '{"data": ["BASE64_AUDIO_DATA"]}'
         ```
-        **Returns:** `{"data": [{"text": "transcribed text"}]}`
+        **Response Format:** `{"data": [{"text": "transcribed text"}]}`
         
-        ### 2. AI Endpoint (Text to Answer) - **WITH STREAMING**
+        ### 2. AI Endpoint - Streaming
         ```
         curl -X POST https://archcoder-basic-app.hf.space/call/answer_ai_stream \\
           -H "Content-Type: application/json" \\
           -d '{"data": ["Your question here"]}'
         ```
-        **Returns:** Server-Sent Events (SSE) stream of text chunks
+        **Response Format:** Streaming text chunks
         
         ---
         
@@ -283,29 +260,43 @@ with gr.Blocks(title="Fast Q&A - Streaming Enabled", theme=gr.themes.Soft()) as
         ### Custom STT Provider:
         **Curl Command:**
         ```
-        curl --location 'https://archcoder-basic-app.hf.space/call/transcribe_stt' \\
-          --header 'Content-Type: application/json' \\
-          --data '{"data": ["{{AUDIO_BASE64}}"]}'
+        curl https://archcoder-basic-app.hf.space/call/transcribe_stt -H "Content-Type: application/json" -d '{"data": ["{{AUDIO_BASE64}}"]}'
         ```
         **Response Content Path:** `data[0].text`
-        **Streaming:** OFF (STT doesn't need streaming)
+        **Streaming:** OFF
         
         ### Custom AI Provider (Streaming):
         **Curl Command:**
         ```
-        curl --location 'https://archcoder-basic-app.hf.space/call/answer_ai_stream' \\
-          --header 'Content-Type: application/json' \\
-          --data '{"data": ["{{TEXT}}"]}'
+        curl https://archcoder-basic-app.hf.space/call/answer_ai_stream -H "Content-Type: application/json" -d '{"data": ["{{TEXT}}"]}'
         ```
-        **Response Content Path:** Leave empty for streaming text
-        **Streaming:** **ON** ✅
-        
-        ### Benefits:
-        - ⚡ Instant feedback as answer generates
-        - 🎯 Better user experience - see words appear in real-time
-        - ⏱️ Perceived latency reduced by 50%+
-        - 🔄 No actual performance penalty
+        **Response Content Path:** `data`
+        **Streaming:** ON ✅
         """)
+        
+        # Hidden interface components that create API endpoints
+        with gr.Row(visible=False):
+            stt_input = gr.Textbox()
+            stt_output = gr.JSON()
+            ai_input = gr.Textbox()
+            ai_output = gr.Textbox()
+        
+        # These create the /call/transcribe_stt and /call/answer_ai_stream endpoints
+        stt_button = gr.Button("STT", visible=False)
+        stt_button.click(
+            fn=transcribe_audio_base64,
+            inputs=[stt_input],
+            outputs=[stt_output],
+            api_name="transcribe_stt"
+        )
+        
+        ai_button = gr.Button("AI", visible=False)
+        ai_button.click(
+            fn=generate_answer_stream,
+            inputs=[ai_input],
+            outputs=[ai_output],
+            api_name="answer_ai_stream"
+        )
     
     gr.Markdown("""
     ---
@@ -314,25 +305,6 @@ with gr.Blocks(title="Fast Q&A - Streaming Enabled", theme=gr.themes.Soft()) as
     **Streaming Mode:** Words appear as they're generated - much faster perceived response!
     """)
 
-# Register API endpoints
-@demo.api(api_name="transcribe_stt")
-def api_transcribe(audio_base64: str):
-    """API endpoint for audio transcription (Pluely STT) - Non-streaming"""
-    result = transcribe_audio_base64(audio_base64)
-    return result
-
-@demo.api(api_name="answer_ai")
-def api_answer(text: str):
-    """API endpoint for text-to-answer (Pluely AI) - Non-streaming fallback"""
-    answer = generate_answer(text)
-    return answer
-
-@demo.api(api_name="answer_ai_stream")
-def api_answer_stream(text: str):
-    """API endpoint for streaming text-to-answer (Pluely AI) - Streaming enabled"""
-    for chunk in generate_answer_stream(text):
-        yield chunk
-
 if __name__ == "__main__":
     demo.queue(max_size=5)
     demo.launch()