Spaces:
Sleeping
Sleeping
File size: 8,522 Bytes
5d16466 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 |
import streamlit as st
import requests
import subprocess
import atexit
import os
import signal
import os
import PyPDF2
from docx import Document
from fastapi import UploadFile, FastAPI, File, Form, UploadFile, HTTPException
from langchain.text_splitter import CharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_google_genai import ChatGoogleGenerativeAI
import pickle
from datetime import datetime
import io
from dotenv import load_dotenv
class User:
def __init__(self, username):
self.username = username
self.llm = "gemini-pro"
self.embedder = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
async def upload_documents(user: User, files: list[UploadFile]) -> tuple[str, int]:
text = await _extract_text_from_document(files)
chunks = await _chunk_text(text)
pkl_name, status_code = await _create_embeddings_and_save(user, chunks)
if status_code == 200:
return "Document uploaded successfully.", 200
else:
return "Failed to upload document.", 500
async def _extract_text_from_document(files: list[UploadFile]) -> str:
text = ""
for file in files:
byte_object = await file.read()
file_name = file.filename
file_extension = os.path.splitext(file_name)[1]
if file_extension == '.txt':
text += byte_object.decode('utf-8')
elif file_extension == '.pdf':
pdf_reader = PyPDF2.PdfReader(io.BytesIO(byte_object))
for page_number in range(len(pdf_reader.pages)):
page = pdf_reader.pages[page_number]
text += page.extract_text()
elif file_extension == '.docx':
doc = Document(io.BytesIO(byte_object))
for paragraph in doc.paragraphs:
text += paragraph.text + "\n"
return text
async def _chunk_text(text: str) -> list[str]:
chunks = None
text_splitter = CharacterTextSplitter(
separator="\n",
chunk_size=512,
chunk_overlap=10,
length_function=len
)
chunks = text_splitter.split_text(text)
return chunks
async def _create_embeddings_and_save(user: User, chunks: any) -> FAISS:
embeddings = HuggingFaceEmbeddings(model_name=user.embedder)
pkl_name = os.path.join(user.username + ".pkl")
vector_store = FAISS.from_texts(chunks, embeddings, metadatas=[{"source": f"{pkl_name}:{i}"} for i in range(len(chunks))])
with open(pkl_name, "wb") as f:
pickle.dump(vector_store, f)
return vector_store
async def ask_question(user: User, question: str, api_key: str) -> tuple[str, int]:
username = user.username
vector_store = await _get_vector_file(username)
if vector_store is None:
return "Document not found.", 400
if api_key is not None:
os.environ["GOOGLE_API_KEY"] = api_key
else:
is_loaded = load_dotenv()
if is_loaded == False:
return "API key not found.", 400
llm = ChatGoogleGenerativeAI(model=user.llm, temperature=0, max_output_tokens=256, top_k = 40, top_p = 0.8)
docs = vector_store.similarity_search(question)
retrieved_chunks = docs[0].page_content + docs[1].page_content + docs[2].page_content
system_message="Figure out the answer of the question by the given information pieces. ALWAYS answer with the language of the question."
prompt = system_message + "Question: " + question + " Context: " + retrieved_chunks
try:
response = llm.invoke(prompt)
except Exception:
return "Wrong API key.", 400
answer = response.content + " **<Most Related Chunk>** " + retrieved_chunks
await _log(user, question, system_message, retrieved_chunks, response.content)
return answer, 200
async def _get_vector_file(username: str)-> any:
with open(username+".pkl", "rb") as f:
vector_store = pickle.load(f)
return vector_store
async def _log(user: User, question: str, system_message: str, retrieved_chunks: str, answer: str):
timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
log_message = (
f"{timestamp}, Username: {user.username}, Question: {question}, "
f"LLM: {user.llm}, Embedder: {user.embedder}, System Message: {system_message}, "
f"Retrieved Texts: {retrieved_chunks}, Answer: {answer}\n"
)
with open("log.txt", "a", encoding="utf-8") as file:
file.write(log_message)
app = FastAPI()
@app.post("/document-uploader")
async def document_uploader(username: str = Form(...), files: list[UploadFile] = File(...)):
user = User(username=username)
response, status_code = await upload_documents(user, files)
if status_code == 200:
return {response}
else:
raise HTTPException(status_code=status_code, detail=response)
@app.post("/question-answerer")
async def question_answerer(username: str = Form(...), question: str = Form(...), api_key = File(None)):
user = User(username=username)
response, status_code = await ask_question(user, question, api_key)
if status_code == 200:
return {response}
else:
raise HTTPException(status_code=status_code, detail=response)
def main():
st.title("Free Multilingual RAG")
tabs = ["Upload Document", "Ask Question"]
active_tab = st.radio("Upload documents first, ask questions later:", tabs)
if active_tab == "Upload Document":
upload_document()
elif active_tab == "Ask Question":
ask_question()
def upload_document():
st.write("Several files can be uploaded, each upload crushes the old one. Depending on the number and size of files, the upload process may take a long time.")
username = st.text_input("Enter a username (just something that represents you):")
uploaded_files = st.file_uploader("Upload your documents (for now it only works with files that has .txt, .pdf or .docx extension):", accept_multiple_files=True)
if uploaded_files:
st.write("Number of uploaded files:", len(uploaded_files))
for uploaded_file in uploaded_files:
file_details = {"FileName": uploaded_file.name, "FileType": uploaded_file.type, "FileSize": uploaded_file.size}
st.write(file_details)
files = [("files", (uploaded_file.name, uploaded_file, uploaded_file.type)) for uploaded_file in uploaded_files]
payload = {'username': username}
with st.spinner('Loading...'):
response = requests.post("http://localhost:8000/document-uploader/", files=files, data=payload)
if response.status_code == 200:
st.success(response.text)
else:
st.error("Error:", response.text)
def ask_question():
username = st.text_input("Enter a username (just something that represents you):")
api_key = st.text_input("Add your Google API key. It is free. Key acquisition video: [https://www.youtube.com/watch?v=brCkpzAD0gc]: (If you do not trust you can download and use the app in your local too)", type="password")
question = st.text_area("Enter the question you want to ask in your document (the more detailed your question, the more accurate an answer you will get): ")
if st.button("Ask"):
if not question:
st.warning("Please enter a question.")
elif not username:
st.warning("Please enter a username.")
else:
payload = {'username': username, 'question': question, 'api_key': api_key}
with st.spinner('Question is getting answered...'):
response = requests.post("http://localhost:8000/question-answerer/", data=payload)
if response.status_code == 200:
st.success("Answer: " + response.text)
else:
print(response)
st.error("Error:", response.text)
uvicorn_process = None
def run_fastapi():
global uvicorn_process
if uvicorn_process is None:
uvicorn_process = subprocess.Popen(["uvicorn", "app:app", "--host", "127.0.0.1", "--port", "8000"])
print("FastAPI server has been started.")
def cleanup():
global uvicorn_process
if uvicorn_process:
os.kill(uvicorn_process.pid, signal.SIGTERM)
uvicorn_process.wait()
print("FastAPI server has been closed.")
if __name__ == "__main__":
run_fastapi()
atexit.register(cleanup)
main()
|