Update app.py

app.py

@@ -1,44 +1,33 @@
 import streamlit as st
-from langchain.chat_models import ChatOpenAI
-from langchain.prompts import ChatPromptTemplate
-from langchain.transformers import JsonSegmenter
+from langchain.prompts import ConversationChain, ConversationTurnPromptTemplate, ListSummarizationPromptTemplate
 
-data = "Customer: Hi, I'm having trouble with my account. Call center guy: Sure, what seems to be the problem? Customer: I can't seem to log in. Call center guy: Okay, let me check your account. What's your username? Customer: It's johndoe123. Call center guy: Alright, I see the issue. Your account has been locked due to too many failed login attempts. Customer: Oh, I see. Can you unlock it for me? Call center guy: Yes, I can do that for you. Just give me a moment. Customer: Thank you so much. Call center guy: No problem. Is there anything else I can help you with?"
+st.title("Call Center Conversation Highlights")
 
-# Set your OpenAI API key
-openai_api_key = openai_api_key
+# Text area for user input
+conversation_text = st.text_area("Paste your call center conversation: ")
 
-# Initialize the GPT-3.5 Turbo LLM
-llm = ChatOpenAI(
-    model_name="gpt-3.5-turbo",
-    openai_api_key=openai_api_key,
-    max_tokens=1024,  # Maximum tokens per inference (adjust as needed)
+# Initialize the conversation chain
+conversation_chain = ConversationChain(
+    turn_prompt_template=ConversationTurnPromptTemplate(),
+    summary_prompt_template=ListSummarizationPromptTemplate(prompt_text="Please summarize the key points in bullet points."),
 )
 
+# Run analysis and display highlights
+if conversation_text:
+    # Segment the conversation based on line breaks
+    conversation_segments = conversation_text.split("\n")
 
-# Segment the conversation into turns
-segmenter = JsonSegmenter()
-conversation_turns = segmenter.split(data)
+    # Generate highlight summaries
+    highlights = []
+    for segment in conversation_segments:
+        response = conversation_chain.run(text=segment)
+        highlights.extend(response["summaries"])
 
-# Define prompt template for summarizing important highlights
-prompt_template = """Summarize the important highlights of the following conversation:\n\n{conversation_turn}\n\nPlease highlight key events, decisions, and customer concerns, using bullet points if possible."""
+    # Display extracted highlights
+    st.header("Extracted Highlights:")
+    for highlight in highlights:
+        st.write(f"- {highlight}")
 
-# Generate summaries for each turn
-highlights = []
-for turn in conversation_turns:
-    current_prompt = prompt_template.format(conversation_turn=turn)
-    response = llm.run(text=current_prompt)
-    summary = response["text"].strip()  # Extract summarized highlights
-    highlights.append(summary)
+else:
+    st.write("Please enter the call center conversation to analyze.")
 
-# Streamlit app
-st.title("Conversation Highlights Generator")
-
-# Display conversation data
-st.subheader("Conversation Data:")
-st.code(data, language="json")
-
-# Display predicted highlights
-st.subheader("Predicted Highlights:")
-for i, highlight in enumerate(highlights, 1):
-    st.write(f"{i}. {highlight}")