optimizing the speed

app.py

@@ -1896,8 +1896,8 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
         # Generate audio with Bark (wait for it)
         start = time.perf_counter()
         cleaned_text = first_q.strip().replace("\n", " ")
-        audio_future = tts_async(cleaned_text)
-        audio_path = audio_future.result()
+        audio_future = executor.submit(tts_async, next_q)
+        audio_path = audio_future.result().result()
         print("⏱️ TTS (edge-tts) took", round(time.perf_counter() - start, 2), "seconds")
 
         # Log question
@@ -1942,7 +1942,12 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
         last_q = state["questions"][-1]
         q_eval = state["question_evaluations"][-1]
         ref_answer = generate_reference_answer(last_q, data["job_role"], data["seniority"])
-        answer_eval = evaluate_answer(last_q, transcript, ref_answer, data["job_role"], data["seniority"], None)
+        answer_eval_future = executor.submit(
+            evaluate_answer,
+            last_q, transcript, ref_answer,
+            data["job_role"], data["seniority"], None
+        )
+        answer_eval = answer_eval_future.result()
         state["answer_evaluations"].append(answer_eval)
         answer_score = answer_eval.get("Score", "medium") if answer_eval else "medium"
 
@@ -1988,7 +1993,13 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
             start = time.time()
             next_q = groq_llm.predict(prompt)
             print("⏱️ Groq LLM Response Time:", round(time.time() - start, 2), "seconds")
-            q_eval = eval_question_quality(next_q, data["job_role"], data["seniority"], None)
+            start = time.time()
+            q_eval_future = executor.submit(
+                eval_question_quality,
+                next_q, data["job_role"], data["seniority"], None
+            )
+            q_eval = q_eval_future.result()
+            print("⏱️ Evaluation time:", round(time.time() - start, 2), "seconds")
             state["questions"].append(next_q)
             state["question_evaluations"].append(q_eval)
             state["conversation_history"].append({'role': 'Interviewer', 'content': next_q})