app.py

import os
import re
import streamlit as st
import openai
from dotenv import load_dotenv
from langchain.agents.openai_assistant import OpenAIAssistantRunnable

# Load environment variables
load_dotenv()
api_key = os.getenv("OPENAI_API_KEY")
extractor_agent = os.getenv("ASSISTANT_ID_SOLUTION_SPECIFIER_A")

# Create the assistant
extractor_llm = OpenAIAssistantRunnable(
    assistant_id=extractor_agent,
    api_key=api_key,
    as_agent=True
)

def remove_citation(text: str) -> str:
    pattern = r"【\d+†\w+】"
    return re.sub(pattern, "📚", text)

# Initialize session state
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
# A flag to indicate if a request is in progress
if "is_in_request" not in st.session_state:
    st.session_state["is_in_request"] = False

st.title("Solution Specifier A")

def predict(user_input: str) -> str:
    """
    This function calls our OpenAIAssistantRunnable to get a response.
    If st.session_state["thread_id"] is None, we start a new thread.
    Otherwise, we continue the existing thread.

    If a concurrency error occurs ("Can't add messages to thread..."), we reset
    the thread_id and try again once on a fresh thread.
    """
    try:
        if st.session_state["thread_id"] is None:
            # Start a new thread
            response = extractor_llm.invoke({"content": user_input})
            st.session_state["thread_id"] = response.thread_id
        else:
            # Continue existing thread
            response = extractor_llm.invoke(
                {"content": user_input, "thread_id": st.session_state["thread_id"]}
            )

        output = response.return_values["output"]
        return remove_citation(output)

    except openai.error.BadRequestError as e:
        # If we get the specific concurrency error, reset thread and try once more
        if "while a run" in str(e):
            st.session_state["thread_id"] = None
            # Now create a new thread for the same user input
            try:
                response = extractor_llm.invoke({"content": user_input})
                st.session_state["thread_id"] = response.thread_id
                output = response.return_values["output"]
                return remove_citation(output)
            except Exception as e2:
                st.error(f"Error after resetting thread: {e2}")
                return ""
        else:
            # Some other 400 error
            st.error(str(e))
            return ""
    except Exception as e:
        st.error(str(e))
        return ""

# Display any existing messages
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"])

# Chat input at the bottom of the page
user_input = st.chat_input("Type your message here...")

# Process the user input only if:
# 1) There is some text, and
# 2) We are not already handling a request (is_in_request == False)
if user_input and not st.session_state["is_in_request"]:
    # Lock to prevent duplicate requests
    st.session_state["is_in_request"] = True

    # 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 assistant response
    response_text = predict(user_input)

    # Add assistant response to session state
    st.session_state["messages"].append({"role": "assistant", "content": response_text})

    # Display assistant response
    with st.chat_message("assistant"):
        st.write(response_text)

    # Release the lock
    st.session_state["is_in_request"] = False