Spaces:
Sleeping
Sleeping
File size: 5,586 Bytes
7a837d4 |
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 |
import os
import shutil
import tempfile
from langchain_community.document_loaders import PyPDFLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_ollama import OllamaEmbeddings
from langchain_community.vectorstores import Chroma
from langchain.chains import RetrievalQA
from langchain_community.llms import Ollama
from book_title_extractor import BookTitleExtractor
from duplicate_detector import DuplicateDetector
from langchain_core.callbacks.base import BaseCallbackHandler
from langchain_community.chat_models import ChatOllama
class StreamingHanlder(BaseCallbackHandler):
def __init__(self):
self.buffer =[]
self.token_callback = None
def on_llm_new_token(self, token:str, **kwargs):
self.buffer.append(token)
if self.token_callback:
self.token_callback(token)
class RagEngine:
def __init__(self, embed_model= "nomic-embed-text",llm_model="qwen:1.8b", temp_dir ="chroma_temp"):
self.embed_model = embed_model
self.llm_model = llm_model
self.embedding = OllamaEmbeddings(model=self.embed_model)
self.vectorstore = None
self.qa_chain = None
self.handler = StreamingHanlder()
self.llm = ChatOllama (model=self.llm_model, streaming= True, callbacks=[self.handler] )
self.temp_dir = temp_dir
os.makedirs(self.temp_dir, exist_ok=True)
self.title_extractor = BookTitleExtractor(llm=self.llm)
self.duplicate_detector = DuplicateDetector()
if os.path.exists(os.path.join(self.temp_dir, "chroma.sqlite3")):
print("🔁 Loading existing Chroma vectorstore...")
self.vectorstore = Chroma(
persist_directory=self.temp_dir,
embedding_function=self.embedding
)
self.qa_chain = RetrievalQA.from_chain_type(
llm=self.llm,
retriever=self.vectorstore.as_retriever(),
return_source_documents=True
)
print("Vectorstore and QA chain restored.")
def clear_temp(self):
shutil.rmtree(self.temp_dir,ignore_errors=True)
os.makedirs(self.temp_dir, exist_ok=True)
def index_pdf(self, pdf_path):
if self.duplicate_detector.is_duplicate(pdf_path):
raise ValueError(f"duplicate book detected, skipping index of: {pdf_path}")
return
self.duplicate_detector.store_fingerprints(pdf_path)
self.clear_temp()
filename = os.path.basename(pdf_path)
loader = PyPDFLoader(pdf_path)
documents = loader.load()
title = self.title_extractor.extract_book_title_from_documents(documents,max_docs=10)
for doc in documents:
doc.metadata["source"] = title
documents = [doc for doc in documents if doc.page_content.strip()]
if not documents:
raise ValueError("No Reasonable text in uploaded pdf")
splitter = RecursiveCharacterTextSplitter(chunk_size = 1000,chunk_overlap = 500 )
chunks = splitter.split_documents(documents)
if self.vectorstore is None:
self.vectorstore = Chroma.from_documents(
documents=chunks,
embedding=self.embedding,
persist_directory=self.temp_dir
)
self.vectorstore.persist()
else:
self.vectorstore.add_documents(chunks)
self.qa_chain = RetrievalQA.from_chain_type(
llm = self.llm,
retriever = self.vectorstore.as_retriever(),
return_source_documents = True
)
def ask_question(self, question):
print (question)
if not self.qa_chain :
return "please upload and index pdf document first"
result = self.qa_chain({"query":question})
answer = result["result"]
sources =[]
for doc in result["source_documents"]:
source = doc.metadata.get("source", "Unknown")
sources.append(source)
print (answer)
return {
"answer": answer,
"sources": list(set(sources)) # Remove duplicates
}
def ask_question_stream(self, question: str):
if not self.qa_chain:
yield "❗ Please upload and index a PDF document first."
return
from queue import Queue, Empty
import threading
q = Queue()
def token_callback(token):
q.put(token)
self.handler.buffer = []
self.handler.token_callback = token_callback
def run():
result = self.qa_chain.invoke({"query": question})
print (result)
self._latest_result = result
q.put(None)
threading.Thread(target=run).start()
print("Threading started", flush=True)
while True:
try:
token = q.get(timeout=30)
if token is None:
print("Stream finished", flush=True)
break
yield token
except Empty:
print("Timed out waiting for token", flush=True)
break
sources = []
for doc in self._latest_result.get("source_documents",[] ):
source = doc.metadata.get("source", "Unknown")
sources.append(source)
if sources:
yield "\n\n📚 **Sources:**\n"
for i, src in enumerate(set(sources)):
yield f"[{i+1}] {src}\n" |