qwen audio streamlit

nexa/gguf/streamlit/streamlit_audio_lm.py

@@ -9,6 +9,13 @@
 from nexa.general import pull_model
 from nexa.gguf.nexa_inference_audio_lm import NexaAudioLMInference
 
+# Initialize session state
+if "has_result" not in st.session_state:
+    st.session_state.has_result = False
+    st.session_state.prompt = ""
+    st.session_state.key = 0
+    st.rerun()
+
 default_model = sys.argv[1]
 is_local_path = False if sys.argv[2] == "False" else True
 hf = False if sys.argv[3] == "False" else True
@@ -49,7 +56,12 @@ def process_audio(nexa_model, audio_file, prompt=""):
         if os.path.exists(temp_audio_path):
             os.unlink(temp_audio_path)
 
-st.title("Nexa AI AudioLM Generation")
+def start_new_callback():
+    st.session_state.has_result = False
+    st.session_state.prompt = ""
+    st.session_state.key += 1
+
+st.markdown("# Nexa AI AudioLM Generation [![Nexa SDK](https://img.shields.io/badge/SDK-Nexa-blue)](https://github.com/NexaAI/nexa-sdk)")
 st.caption("Powered by Nexa AI SDK🐙")
 
 # Sidebar configuration
@@ -68,39 +80,56 @@ def process_audio(nexa_model, audio_file, prompt=""):
     if st.session_state.nexa_model is None:
         st.stop()
 
-# Text prompt input
-prompt = st.text_input("Enter optional prompt text:", "")
-
-# Option 1: Upload Audio File
-st.subheader("Option 1: Upload Audio File")
-uploaded_file = st.file_uploader("Choose an audio file", type=["wav", "mp3"])
-
-if uploaded_file is not None:
-    st.audio(uploaded_file, format="audio/wav")
-
-    if st.button("Process Audio"):
-        with st.spinner("Processing audio..."):
-            response = process_audio(st.session_state.nexa_model, uploaded_file, prompt)
-
-        if response:
-            st.subheader("Model Response:")
-            st.write(response)
-        else:
-            st.error("Processing failed. Please try again with a different audio file.")
-
-# Option 2: Real-time Recording
-st.subheader("Option 2: Record Audio")
-wav_audio_data = st_audiorec()
-
-if wav_audio_data:
-    if st.button("Process Recorded Audio"):
-        with st.spinner("Processing audio..."):
-            response = process_audio(st.session_state.nexa_model, io.BytesIO(wav_audio_data), prompt)
-
-        if response:
-            st.subheader("Model Response:")
-            st.write(response)
-        else:
-            st.error("Processing failed. Please try recording again.")
-else:
-    st.warning("No audio recorded. Please record some audio before processing.")
+# Wrap input sections
+input_container = st.container()
+
+with input_container:
+    # Text prompt input
+    prompt = st.text_input("Enter optional prompt text:", value=st.session_state.prompt, key=f"prompt_{st.session_state.key}")
+
+
+    # Option 1: Upload Audio File
+    st.subheader("Option 1: Upload Audio File")
+    uploaded_file = st.file_uploader("Choose an audio file", type=["wav", "mp3"], key=f"uploader_{st.session_state.key}")
+
+    if uploaded_file is not None:
+        st.audio(uploaded_file, format="audio/wav")
+
+        if st.button("Process Audio", key="process_upload"):
+            with st.spinner("Processing audio..."):
+                response = process_audio(st.session_state.nexa_model, uploaded_file, prompt)
+                if response:
+                    st.session_state.has_result = True
+
+            if response:
+                st.subheader("Model Response:")
+                st.write(response)
+                if st.session_state.has_result:
+                    if st.button("Start new", key="start_new_upload", on_click=start_new_callback):
+                        pass
+                    st.write("Start new conversation to try next prompt")
+            else:
+                st.error("Processing failed. Please try again with a different audio file.")
+
+    # Option 2: Real-time Recording
+    st.subheader("Option 2: Record Audio")
+    wav_audio_data = st_audiorec()
+
+    if wav_audio_data:
+        if st.button("Process Audio", key="process_record"):
+            with st.spinner("Processing audio..."):
+                response = process_audio(st.session_state.nexa_model, io.BytesIO(wav_audio_data), prompt)
+                if response:
+                    st.session_state.has_result = True
+
+            if response:
+                st.subheader("Model Response:")
+                st.write(response)
+                if st.session_state.has_result:
+                    if st.button("Start new", key="start_new_record", on_click=start_new_callback):
+                        pass
+                    st.write("Start new conversation to try next prompt")
+            else:
+                st.error("Processing failed. Please try recording again.")
+    else:
+        st.warning("No audio recorded. Please record some audio before processing.")