Update app.py

app.py

@@ -1,66 +1,76 @@
 import os
 import re
+import streamlit as st
 from dotenv import load_dotenv
-load_dotenv()
-
-import gradio as gr
-
 from langchain.agents.openai_assistant import OpenAIAssistantRunnable
-from langchain.schema import HumanMessage, AIMessage
 
-api_key = os.getenv('OPENAI_API_KEY')
-extractor_agent = os.getenv('ASSISTANT_ID_SOLUTION_SPECIFIER_A')
+# Load environment variables
+load_dotenv()
+api_key = os.getenv("OPENAI_API_KEY")
+extractor_agent = os.getenv("ASSISTANT_ID_SOLUTION_SPECIFIER_A")
 
-# Create the assistant. By default, we don't specify a thread_id,
-# so the first call that doesn't pass one will create a new thread.
+# Create the assistant
 extractor_llm = OpenAIAssistantRunnable(
     assistant_id=extractor_agent,
     api_key=api_key,
     as_agent=True
 )
 
-# We will store thread_id globally or in a session variable.
-THREAD_ID = None
-
-def remove_citation(text):
+def remove_citation(text: str) -> str:
     pattern = r"【\d+†\w+】"
     return re.sub(pattern, "📚", text)
 
-def predict(message, history):
+# Initialize session state for messages and thread_id
+if "messages" not in st.session_state:
+    st.session_state["messages"] = []
+if "thread_id" not in st.session_state:
+    st.session_state["thread_id"] = None
+
+st.title("Solution Specifier A")
+
+def predict(user_input: str) -> str:
     """
-    Receives the new user message plus the entire conversation history 
-    from Gradio. If no thread_id is set, we create a new thread. 
-    Otherwise we pass the existing thread_id.
+    This function calls our OpenAIAssistantRunnable to get a response.
+    If we don't have a thread_id yet, we create a new thread on the first call.
+    Otherwise, we continue the existing thread.
     """
-    global THREAD_ID
-    
-    # debug print
-    print("current history:", history)
-    
-    # If history is empty, this means that it is probably a new conversation and therefore the thread shall be reset
-    if not history:
-        THREAD_ID = None
-    
-    # 1) Decide if we are creating a new thread or continuing the old one
-    if THREAD_ID is None:
-        # No thread_id yet -> this is the first user message
-        response = extractor_llm.invoke({"content": message})
-        THREAD_ID = response.thread_id  # store for subsequent calls
+    if st.session_state["thread_id"] is None:
+        response = extractor_llm.invoke({"content": user_input})
+        st.session_state["thread_id"] = response.thread_id
     else:
-        # We already have a thread_id -> continue that same thread
-        response = extractor_llm.invoke({"content": message, "thread_id": THREAD_ID})
-    
-    # 2) Extract the text output from the response
+        response = extractor_llm.invoke(
+            {"content": user_input, "thread_id": st.session_state["thread_id"]}
+        )
     output = response.return_values["output"]
-    non_cited_output = remove_citation(output)
-    
-    # 3) Return the model's text to display in Gradio
-    return non_cited_output
+    return remove_citation(output)
 
-# Create a Gradio ChatInterface using our predict function
-chat = gr.ChatInterface(
-    fn=predict, 
-    title="Solution Specifier A", 
-    #description="Testing threaded conversation"
-)
-chat.launch(share=True)
+# Display any existing messages (from a previous run or refresh)
+for msg in st.session_state["messages"]:
+    if msg["role"] == "user":
+        with st.chat_message("user"):
+            st.write(msg["content"])
+    else:
+        with st.chat_message("assistant"):
+            st.write(msg["content"])
+
+# Create the chat input widget at the bottom of the page
+user_input = st.chat_input("Type your message here...")
+
+# When the user hits ENTER on st.chat_input
+if user_input:
+    # Add the user message to session state
+    st.session_state["messages"].append({"role": "user", "content": user_input})
+    
+    # Display the user's message
+    with st.chat_message("user"):
+        st.write(user_input)
+        
+    # Get the assistant's response
+    response_text = predict(user_input)
+    
+    # Add the assistant response to session state
+    st.session_state["messages"].append({"role": "assistant", "content": response_text})
+    
+    # Display the assistant's reply
+    with st.chat_message("assistant"):
+        st.write(response_text)