|
|
import streamlit as st |
|
|
import time |
|
|
from langchain.document_loaders import PyPDFLoader |
|
|
from langchain.text_splitter import RecursiveCharacterTextSplitter |
|
|
from langchain_together import TogetherEmbeddings |
|
|
from langchain.vectorstores import FAISS |
|
|
from langchain.chat_models import ChatOpenAI |
|
|
from langchain.chains import RetrievalQA |
|
|
from langchain.indexes import VectorstoreIndexCreator |
|
|
|
|
|
|
|
|
@st.cache_resource |
|
|
def load_chunks_and_embeddings(): |
|
|
with st.spinner("در حال بارگذاری فایل و آماده‌سازی... لطفاً صبور باشید 🙏"): |
|
|
progress_bar = st.progress(0, text="در حال بارگذاری فایل PDF...") |
|
|
|
|
|
pdf_loader = PyPDFLoader('test1.pdf') |
|
|
pages = pdf_loader.load() |
|
|
progress_bar.progress(30, text="صفحات PDF بارگذاری شد. در حال ایجاد مدل برداری...") |
|
|
|
|
|
embeddings = TogetherEmbeddings( |
|
|
api_key="0291f33aee03412a47fa5d8e562e515182dcc5d9aac5a7fb5eefdd1759005979" |
|
|
) |
|
|
progress_bar.progress(60, text="مدل Embedding ساخته شد. در حال ایجاد ایندکس...") |
|
|
|
|
|
index = VectorstoreIndexCreator( |
|
|
embedding=embeddings, |
|
|
text_splitter=RecursiveCharacterTextSplitter(chunk_size=300, chunk_overlap=0) |
|
|
).from_loaders([pdf_loader]) |
|
|
|
|
|
progress_bar.progress(100, text="بارگذاری کامل شد! ✅") |
|
|
return index |
|
|
llm = ChatOpenAI( |
|
|
base_url="https://api.together.xyz/v1", |
|
|
api_key='0291f33aee03412a47fa5d8e562e515182dcc5d9aac5a7fb5eefdd1759005979', |
|
|
model="meta-llama/Llama-3-70B-Instruct-Turbo-Free" |
|
|
) |
|
|
|
|
|
index = load_chunks_and_embeddings() |
|
|
|
|
|
chain = RetrievalQA.from_chain_type( |
|
|
llm=llm, |
|
|
chain_type='stuff', |
|
|
retriever=index.vectorstore.as_retriever(), input_key='question') |
|
|
|
|
|
|
|
|
|
|
|
if 'messages' not in st.session_state: |
|
|
st.session_state.messages = [] |
|
|
|
|
|
if 'pending_prompt' not in st.session_state: |
|
|
st.session_state.pending_prompt = None |
|
|
|
|
|
st.title("📄🤖 دستیار PDF شما") |
|
|
|
|
|
|
|
|
for msg in st.session_state.messages: |
|
|
with st.chat_message(msg['role']): |
|
|
st.markdown(f"🗨️ {msg['content']}", unsafe_allow_html=True) |
|
|
|
|
|
prompt = st.chat_input("سوالی از PDF داری؟") |
|
|
|
|
|
if prompt: |
|
|
st.session_state.messages.append({'role': 'user', 'content': prompt}) |
|
|
st.session_state.pending_prompt = prompt |
|
|
st.rerun() |
|
|
|
|
|
if st.session_state.pending_prompt: |
|
|
with st.chat_message('ai'): |
|
|
thinking = st.empty() |
|
|
thinking.markdown("🤖 در حال فکر کردن...") |
|
|
|
|
|
|
|
|
response = chain.run(f'فقط به زبان فارسی جواب بده. سوال: {st.session_state.pending_prompt}') |
|
|
answer = response.split("Helpful Answer:")[-1].strip() |
|
|
if not answer: |
|
|
answer = "متأسفم، اطلاعات دقیقی در این مورد ندارم." |
|
|
|
|
|
thinking.empty() |
|
|
full_response = "" |
|
|
placeholder = st.empty() |
|
|
|
|
|
for word in answer.split(): |
|
|
full_response += word + " " |
|
|
placeholder.markdown(full_response + "▌") |
|
|
time.sleep(0.03) |
|
|
|
|
|
placeholder.markdown(full_response) |
|
|
st.session_state.messages.append({'role': 'ai', 'content': full_response}) |
|
|
st.session_state.pending_prompt = None |
|
|
|
