AbenzaFran
commited on
Update app.py
Browse files
app.py
CHANGED
@@ -1,76 +1,67 @@
|
|
|
|
1 |
import os
|
2 |
import re
|
3 |
-
import streamlit as st
|
4 |
from dotenv import load_dotenv
|
|
|
|
|
|
|
|
|
5 |
from langchain.agents.openai_assistant import OpenAIAssistantRunnable
|
|
|
6 |
|
7 |
-
|
8 |
-
|
9 |
-
api_key = os.getenv("OPENAI_API_KEY")
|
10 |
-
extractor_agent = os.getenv("ASSISTANT_ID_SOLUTION_SPECIFIER_A")
|
11 |
|
12 |
-
# Create the assistant
|
|
|
13 |
extractor_llm = OpenAIAssistantRunnable(
|
14 |
assistant_id=extractor_agent,
|
15 |
api_key=api_key,
|
16 |
as_agent=True
|
17 |
)
|
18 |
|
19 |
-
|
|
|
|
|
|
|
20 |
pattern = r"γ\d+β \w+γ"
|
21 |
return re.sub(pattern, "π", text)
|
22 |
|
23 |
-
|
24 |
-
if "messages" not in st.session_state:
|
25 |
-
st.session_state["messages"] = []
|
26 |
-
if "thread_id" not in st.session_state:
|
27 |
-
st.session_state["thread_id"] = None
|
28 |
-
|
29 |
-
st.title("Solution Specifier A")
|
30 |
-
|
31 |
-
def predict(user_input: str) -> str:
|
32 |
"""
|
33 |
-
|
34 |
-
|
35 |
-
Otherwise
|
36 |
"""
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
response = extractor_llm.invoke(
|
42 |
-
{"content": user_input, "thread_id": st.session_state["thread_id"]}
|
43 |
-
)
|
44 |
-
output = response.return_values["output"]
|
45 |
-
return remove_citation(output)
|
46 |
-
|
47 |
-
# Display any existing messages (from a previous run or refresh)
|
48 |
-
for msg in st.session_state["messages"]:
|
49 |
-
if msg["role"] == "user":
|
50 |
-
with st.chat_message("user"):
|
51 |
-
st.write(msg["content"])
|
52 |
-
else:
|
53 |
-
with st.chat_message("assistant"):
|
54 |
-
st.write(msg["content"])
|
55 |
-
|
56 |
-
# Create the chat input widget at the bottom of the page
|
57 |
-
user_input = st.chat_input("Type your message here...")
|
58 |
-
|
59 |
-
# When the user hits ENTER on st.chat_input
|
60 |
-
if user_input:
|
61 |
-
# Add the user message to session state
|
62 |
-
st.session_state["messages"].append({"role": "user", "content": user_input})
|
63 |
|
64 |
-
#
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
# Get the assistant's response
|
69 |
-
response_text = predict(user_input)
|
70 |
|
71 |
-
#
|
72 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
73 |
|
74 |
-
#
|
75 |
-
|
76 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
import os
|
3 |
import re
|
|
|
4 |
from dotenv import load_dotenv
|
5 |
+
load_dotenv()
|
6 |
+
|
7 |
+
import gradio as gr
|
8 |
+
|
9 |
from langchain.agents.openai_assistant import OpenAIAssistantRunnable
|
10 |
+
from langchain.schema import HumanMessage, AIMessage
|
11 |
|
12 |
+
api_key = os.getenv('OPENAI_API_KEY')
|
13 |
+
extractor_agent = os.getenv('ASSISTANT_ID_SOLUTION_SPECIFIER_A')
|
|
|
|
|
14 |
|
15 |
+
# Create the assistant. By default, we don't specify a thread_id,
|
16 |
+
# so the first call that doesn't pass one will create a new thread.
|
17 |
extractor_llm = OpenAIAssistantRunnable(
|
18 |
assistant_id=extractor_agent,
|
19 |
api_key=api_key,
|
20 |
as_agent=True
|
21 |
)
|
22 |
|
23 |
+
# We will store thread_id globally or in a session variable.
|
24 |
+
THREAD_ID = None
|
25 |
+
|
26 |
+
def remove_citation(text):
|
27 |
pattern = r"γ\d+β \w+γ"
|
28 |
return re.sub(pattern, "π", text)
|
29 |
|
30 |
+
def predict(message, history):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
"""
|
32 |
+
Receives the new user message plus the entire conversation history
|
33 |
+
from Gradio. If no thread_id is set, we create a new thread.
|
34 |
+
Otherwise we pass the existing thread_id.
|
35 |
"""
|
36 |
+
global THREAD_ID
|
37 |
+
|
38 |
+
# debug print
|
39 |
+
print("current history:", history)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
|
41 |
+
# If history is empty, this means that it is probably a new conversation and therefore the thread shall be reset
|
42 |
+
if not history:
|
43 |
+
THREAD_ID = None
|
|
|
|
|
|
|
44 |
|
45 |
+
# 1) Decide if we are creating a new thread or continuing the old one
|
46 |
+
if THREAD_ID is None:
|
47 |
+
# No thread_id yet -> this is the first user message
|
48 |
+
response = extractor_llm.invoke({"content": message})
|
49 |
+
THREAD_ID = response.thread_id # store for subsequent calls
|
50 |
+
else:
|
51 |
+
# We already have a thread_id -> continue that same thread
|
52 |
+
response = extractor_llm.invoke({"content": message, "thread_id": THREAD_ID})
|
53 |
+
|
54 |
+
# 2) Extract the text output from the response
|
55 |
+
output = response.return_values["output"]
|
56 |
+
non_cited_output = remove_citation(output)
|
57 |
|
58 |
+
# 3) Return the model's text to display in Gradio
|
59 |
+
return non_cited_output
|
60 |
+
|
61 |
+
# Create a Gradio ChatInterface using our predict function
|
62 |
+
chat = gr.ChatInterface(
|
63 |
+
fn=predict,
|
64 |
+
title="Solution Specifier A",
|
65 |
+
#description="Testing threaded conversation"
|
66 |
+
)
|
67 |
+
chat.launch(share=True)
|