Update streamlit_app.py

streamlit_app.py

@@ -1,68 +1,146 @@
+# import streamlit as st
+# from src.tools.accent_tool import AccentAnalyzerTool
+# from src.app.main_agent import create_agent
+# from langchain_core.messages import HumanMessage, AIMessage
+# import re
+
+# st.set_page_config(page_title="Accent Analyzer Agent", page_icon="💬", layout="centered")
+
+# st.warning("⚠️ High latency due to CPU usage. Once migrated to GPU, response time will improve significantly.")
+
+# st.title("English Accent Analyzer (Conversational)")
+# st.subheader("Ask me to analyze a video URL, e.g.: \n\n `Analyze this video: https://github.com/ash-171/Data-mp4/raw/refs/heads/main/NWRNVTFlRGlnV0FfNDgwcA_out.mp4`")
+
+# @st.cache_resource
+# def load_tool_and_agent():
+#     tool = AccentAnalyzerTool()
+#     analysis_agent, follow_up_agent = create_agent(tool)
+#     return tool, analysis_agent, follow_up_agent
+
+# accent_tool_obj, analysis_agent, follow_up_agent = load_tool_and_agent()
+
+# if "chat_history" not in st.session_state:
+#     st.session_state.chat_history = []
+
+# if hasattr(accent_tool_obj, "last_transcript") and accent_tool_obj.last_transcript:
+#     prompt_label = "Ask more about the video..."
+#     input_key = "followup"
+# else:
+#     prompt_label = "Paste your prompt here..."
+#     input_key = "initial"
+
+# user_input = st.chat_input(prompt_label, key=input_key)
+
+# # Variable to defer assistant response
+# deferred_response = None
+# deferred_spinner_msg = ""
+
+# if user_input:
+#     st.session_state.chat_history.append(HumanMessage(content=user_input))
+
+#     if re.search(r'https?://\S+', user_input):
+#         accent_tool_obj.last_transcript = ""
+#         deferred_spinner_msg = "Analyzing new video..."
+#         def run_agent():
+#             return analysis_agent.invoke(st.session_state.chat_history)[-1].content
+#     else:
+#         deferred_spinner_msg = "Responding based on transcript..."
+#         def run_agent():
+#             return follow_up_agent.invoke(st.session_state.chat_history).content
+
+#     # Run response generation inside spinner after chat is rendered
+#     def process_response():
+#         with st.spinner(deferred_spinner_msg):
+#             try:
+#                 result = run_agent()
+#             except Exception as e:
+#                 result = f"Error: {str(e)}"
+#             st.session_state.chat_history.append(AIMessage(content=result))
+#             st.rerun() 
+
+# # Display full chat history (before running spinner)
+# for msg in st.session_state.chat_history:
+#     with st.chat_message("user" if isinstance(msg, HumanMessage) else "assistant"):
+#         st.markdown(msg.content)
+
+# # Only process response at the bottom, after chat is shown
+# if user_input:
+#     process_response()
+
 import streamlit as st
 from src.tools.accent_tool import AccentAnalyzerTool
 from src.app.main_agent import create_agent
 from langchain_core.messages import HumanMessage, AIMessage
 import re
+import os
 
 st.set_page_config(page_title="Accent Analyzer Agent", page_icon="💬", layout="centered")
 
 st.warning("⚠️ High latency due to CPU usage. Once migrated to GPU, response time will improve significantly.")
-
 st.title("English Accent Analyzer (Conversational)")
 st.subheader("Ask me to analyze a video URL, e.g.: \n\n `Analyze this video: https://github.com/ash-171/Data-mp4/raw/refs/heads/main/NWRNVTFlRGlnV0FfNDgwcA_out.mp4`")
 
+if "chat_history" not in st.session_state:
+    st.session_state.chat_history = [] 
+if "last_transcript_available" not in st.session_state:
+    st.session_state.last_transcript_available = False 
+if "processing_input" not in st.session_state:
+    st.session_state.processing_input = False 
+
 @st.cache_resource
-def load_tool_and_agent():
-    tool = AccentAnalyzerTool()
-    analysis_agent, follow_up_agent = create_agent(tool)
-    return tool, analysis_agent, follow_up_agent
+def load_tool_and_agent_cached():
+    with st.spinner("Loading AI models and tools... This might take a moment."):
+        if not os.environ.get("HF_TOKEN"):
+            st.warning("HF_TOKEN environment variable not found. Some Hugging Face models might require authentication and may fail to load.")
 
-accent_tool_obj, analysis_agent, follow_up_agent = load_tool_and_agent()
+        tool = AccentAnalyzerTool()
+        analysis_agent, follow_up_agent = create_agent(tool)
+        return tool, analysis_agent, follow_up_agent
 
-if "chat_history" not in st.session_state:
-    st.session_state.chat_history = []
+accent_tool_obj, analysis_agent, follow_up_agent = load_tool_and_agent_cached()
 
 if hasattr(accent_tool_obj, "last_transcript") and accent_tool_obj.last_transcript:
-    prompt_label = "Ask more about the video..."
-    input_key = "followup"
+    st.session_state.last_transcript_available = True
 else:
-    prompt_label = "Paste your prompt here..."
-    input_key = "initial"
-
-user_input = st.chat_input(prompt_label, key=input_key)
+    st.session_state.last_transcript_available = False
 
-# Variable to defer assistant response
-deferred_response = None
-deferred_spinner_msg = ""
+prompt_label = "Ask more about the video..." if st.session_state.last_transcript_available else "Paste your prompt here..."
+user_input = st.chat_input(prompt_label, disabled=st.session_state.processing_input)
 
-if user_input:
-    st.session_state.chat_history.append(HumanMessage(content=user_input))
+if user_input and not st.session_state.processing_input:
+    st.session_state.processing_input = True 
+    st.session_state.chat_history.append(HumanMessage(content=user_input)) 
 
     if re.search(r'https?://\S+', user_input):
-        accent_tool_obj.last_transcript = ""
-        deferred_spinner_msg = "Analyzing new video..."
-        def run_agent():
-            return analysis_agent.invoke(st.session_state.chat_history)[-1].content
+        accent_tool_obj.last_transcript = "" 
+        st.session_state.last_transcript_available = False 
+        agent_to_run = analysis_agent
+        spinner_msg = "Analyzing new video and transcribing audio..."
     else:
-        deferred_spinner_msg = "Responding based on transcript..."
-        def run_agent():
-            return follow_up_agent.invoke(st.session_state.chat_history).content
-
-    # Run response generation inside spinner after chat is rendered
-    def process_response():
-        with st.spinner(deferred_spinner_msg):
-            try:
-                result = run_agent()
-            except Exception as e:
-                result = f"Error: {str(e)}"
-            st.session_state.chat_history.append(AIMessage(content=result))
-            st.rerun() 
-
-# Display full chat history (before running spinner)
+        agent_to_run = follow_up_agent
+        spinner_msg = "Generating response based on transcript..."
+        
+    with st.spinner(spinner_msg):
+        try:
+            if agent_to_run == analysis_agent:
+                response_content = agent_to_run.invoke(st.session_state.chat_history)[-1].content
+            else:
+                response_content = follow_up_agent.invoke(st.session_state.chat_history).content
+
+            if agent_to_run == analysis_agent and hasattr(accent_tool_obj, "last_transcript") and accent_tool_obj.last_transcript:
+                st.session_state.last_transcript_available = True
+
+        except Exception as e:
+            response_content = f"Error: {str(e)}"
+            st.error(f"An error occurred during processing: {e}")
+
+    st.session_state.chat_history.append(AIMessage(content=response_content))
+    st.session_state.processing_input = False
+    st.rerun()
+
 for msg in st.session_state.chat_history:
     with st.chat_message("user" if isinstance(msg, HumanMessage) else "assistant"):
         st.markdown(msg.content)
 
-# Only process response at the bottom, after chat is shown
-if user_input:
-    process_response()
+if not st.session_state.chat_history:
+    st.info("Start by pasting a public MP4 video URL above!")