poemsforaphrodite's picture
Upload 2 files
5974fda verified
raw
history blame
11.1 kB
import os
import io
import requests
import streamlit as st
from openai import OpenAI
from PyPDF2 import PdfReader
import urllib.parse
from dotenv import load_dotenv
from openai import OpenAI
from io import BytesIO
from streamlit_extras.colored_header import colored_header
from streamlit_extras.add_vertical_space import add_vertical_space
from streamlit_extras.switch_page_button import switch_page
import json
# ---------------------- Configuration ----------------------
st.set_page_config(page_title="Building Regulations Chatbot", layout="wide", page_icon="πŸ—οΈ", initial_sidebar_state="expanded")
# Load environment variables from .env file
load_dotenv()
# Set OpenAI API key
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
# ---------------------- Session State Initialization ----------------------
if 'pdf_contents' not in st.session_state:
st.session_state.pdf_contents = []
if 'chat_history' not in st.session_state:
st.session_state.chat_history = []
if 'processed_pdfs' not in st.session_state:
st.session_state.processed_pdfs = False
if 'id_counter' not in st.session_state:
st.session_state.id_counter = 0
if 'assistant_id' not in st.session_state:
st.session_state.assistant_id = None
if 'thread_id' not in st.session_state:
st.session_state.thread_id = None
# ---------------------- Helper Functions ----------------------
def get_vector_stores():
try:
client = OpenAI()
vector_stores = client.beta.vector_stores.list()
return vector_stores
except Exception as e:
return f"Error retrieving vector stores: {str(e)}"
def fetch_pdfs(city_code):
url = f"http://91.203.213.50:5000/oereblex/{city_code}"
response = requests.get(url)
if response.status_code == 200:
data = response.json()
print("First data:", data.get('data', [])[0] if data.get('data') else None)
return data.get('data', [])
else:
st.error(f"Failed to fetch PDFs for city code {city_code}")
return None
def download_pdf(url):
# Add 'https://' scheme if it's missing
if not url.startswith(('http://', 'https://')):
url = 'https://' + url
try:
response = requests.get(url)
response.raise_for_status() # Raise an exception for bad status codes
# Generate a unique filename
filename = f"downloaded_pdf_{st.session_state.id_counter}.pdf"
st.session_state.id_counter += 1
# Save the PDF content to a file
with open(filename, 'wb') as f:
f.write(response.content)
return filename
except requests.RequestException as e:
st.error(f"Failed to download PDF from {url}. Error: {str(e)}")
return None
# Helper function to upload file to OpenAI
def upload_file_to_openai(file_path):
try:
file = client.files.create(
file=open(file_path, 'rb'),
purpose='assistants'
)
return file.id
except Exception as e:
st.error(f"Failed to upload file {file_path}. Error: {str(e)}")
return None
def create_assistant():
assistant = client.beta.assistants.create(
name="Building Regulations Assistant",
instructions="You are an expert on building regulations. Use the provided documents to answer questions accurately.",
model="gpt-4o-mini",
tools=[{"type": "file_search"}]
)
st.session_state.assistant_id = assistant.id
return assistant.id
def chat_with_assistant(file_ids, user_message):
for file_id in file_ids:
print("File ID:", file_id)
# Create a thread and attach the file to the message
print("final file id:", file_id)
attachments = [{"file_id": file_id, "tools": [{"type": "file_search"}]} for file_id in file_ids]
print("attachments:", attachments)
if st.session_state.thread_id is None:
thread = client.beta.threads.create(
messages=[
{
"role": "user",
"content": user_message,
"attachments": attachments,
}
]
)
st.session_state.thread_id = thread.id
else:
thread = client.beta.threads.messages.create(
thread_id=st.session_state.thread_id,
role="user",
content=user_message,
attachments=attachments
)
# The thread now has a vector store with that file in its tool resources.
print(thread.tool_resources.file_search)
print("assistant_id:", st.session_state.assistant_id)
print("thread_id:", thread.id)
run = client.beta.threads.runs.create_and_poll(
thread_id=thread.id, assistant_id=st.session_state.assistant_id
)
print("run:", run)
messages = list(client.beta.threads.messages.list(thread_id=thread.id, run_id=run.id))
message_content = messages[0].content[0].text
annotations = message_content.annotations
citations = []
for index, annotation in enumerate(annotations):
message_content.value = message_content.value.replace(annotation.text, f"[{index}]")
if file_citation := getattr(annotation, "file_citation", None):
cited_file = client.files.retrieve(file_citation.file_id)
citations.append(f"[{index}] {cited_file.filename}")
print(message_content.value)
print("\n".join(citations))
return message_content.value, citations
# ---------------------- Streamlit App ----------------------
page = st.sidebar.selectbox("Choose a page", ["Home", "Documents", "Admin"])
if page == "Home":
st.title("πŸ—οΈ Building Regulations Chatbot", anchor=False)
# Sidebar improvements
with st.sidebar:
colored_header("πŸ“„ Selected Documents", description="Documents for chat")
if 'selected_pdfs' in st.session_state and st.session_state.selected_pdfs:
for pdf in st.session_state.selected_pdfs:
st.write(f"- {pdf['doc_title']}")
else:
st.write("No documents selected. Please go to the Documents page.")
# Main chat area improvements
colored_header("πŸ’¬ Chat", description="Ask questions about building regulations")
# Display chat messages with improved styling
for chat in st.session_state.chat_history:
with st.container():
if chat['role'] == 'user':
st.markdown(f"""
<div class="chat-message user">
<div><strong>You</strong></div>
<div class="message-content">{chat['content']}</div>
</div>
""", unsafe_allow_html=True)
else:
st.markdown(f"""
<div class="chat-message assistant">
<div><strong>Assistant</strong></div>
<div class="message-content">{chat['content']}</div>
</div>
""", unsafe_allow_html=True)
# Chat input improvements
with st.form("chat_form", clear_on_submit=True):
user_input = st.text_area("Ask a question about building regulations...", height=100)
col1, col2 = st.columns([3, 1])
with col2:
submit = st.form_submit_button("Send", use_container_width=True)
if submit and user_input.strip() != "":
# Add user message to chat history
st.session_state.chat_history.append({"role": "user", "content": user_input})
if not st.session_state.file_ids:
st.error("Please process PDFs first.")
else:
with st.spinner("Generating response..."):
try:
response, citations = chat_with_assistant(st.session_state.file_ids, user_input)
# Add assistant response to chat history
st.session_state.chat_history.append({"role": "assistant", "content": response+"\n\n"+"\n".join(citations)})
except Exception as e:
st.error(f"Error generating response: {str(e)}")
# Rerun the app to update the chat display
st.rerun()
# Footer improvements
add_vertical_space(2)
st.markdown("---")
col1, col2 = st.columns(2)
with col1:
st.caption("Powered by OpenAI GPT-4 and Pinecone")
with col2:
st.caption("Β© 2023 Your Company Name")
elif page == "Documents":
st.title("πŸ“š Document Selection")
city_code_input = st.text_input("Enter city code:", key="city_code_input")
load_documents_button = st.button("Load Documents", key="load_documents_button")
if load_documents_button and city_code_input:
with st.spinner("Fetching PDFs..."):
pdfs = fetch_pdfs(city_code_input)
if pdfs:
st.session_state.available_pdfs = pdfs
st.success(f"Found {len(pdfs)} PDFs")
else:
st.error("No PDFs found")
if 'available_pdfs' in st.session_state:
st.write(f"Total PDFs: {len(st.session_state.available_pdfs)}")
selected_pdfs = []
for pdf in st.session_state.available_pdfs:
title = f"{pdf.get('doc_title', 'Untitled')} - {pdf.get('doc_type', 'No type')}"
if pdf.get('number'):
title += f" ({pdf['number']})"
if st.checkbox(title, key=f"pdf_{pdf['file_href']}"):
selected_pdfs.append(pdf)
if st.button("Process Selected PDFs"):
if selected_pdfs:
st.session_state.selected_pdfs = selected_pdfs
st.session_state.assistant_id = create_assistant()
with st.spinner("Processing PDFs and creating/updating assistant..."):
file_ids = []
for pdf in selected_pdfs:
file_name = download_pdf(pdf['file_href'])
if file_name:
file_path = f"./{file_name}"
file_id = upload_file_to_openai(file_path)
if file_id:
file_ids.append(file_id)
else:
st.warning(f"Failed to upload {pdf['doc_title']}. Skipping this file.")
else:
st.warning(f"Failed to download {pdf['doc_title']}. Skipping this file.")
st.session_state.file_ids = file_ids
st.success("PDFs processed successfully. You can now chat on the Home page.")
else:
st.warning("Select at least one PDF.")
if st.button("Go to Home"):
switch_page("Home")
elif page == "Admin":
st.title("πŸ”§ Admin Panel")
colored_header("Vector Stores Information", description="List of all vector stores")
vector_stores = get_vector_stores()
json_vector_stores = json.dumps([vs.model_dump() for vs in vector_stores])
st.write(json_vector_stores)
# Add a button to go back to the main page
if st.button("Back to Home"):
switch_page("Home")