Spaces:
Sleeping
Sleeping
File size: 2,029 Bytes
c07fa76 d300309 c07fa76 314c1d9 c07fa76 d300309 c07fa76 d300309 c07fa76 314c1d9 d300309 c07fa76 d300309 c07fa76 d300309 c07fa76 d300309 | 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 | import os
import tempfile
import requests
import re
from fastapi import HTTPException
from langchain_community.document_loaders import PyMuPDFLoader, Docx2txtLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.schema import Document
def load_and_chunk(url: str) -> list[Document]:
resp = requests.get(url)
if resp.status_code != 200:
raise HTTPException(400, "Document download failed")
content_type = resp.headers.get("Content-Type", "").lower()
url_lower = url.lower()
text = ""
if "application/pdf" in content_type or ".pdf" in url_lower:
with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp:
tmp.write(resp.content)
tmp_path = tmp.name
try:
loader = PyMuPDFLoader(tmp_path)
# Extract full text first for better chunking
pages = loader.load()
text = "\n".join([p.page_content for p in pages])
finally:
os.remove(tmp_path)
elif ("application/vnd.openxmlformats-officedocument.wordprocessingml.document" in content_type
or ".docx" in url_lower):
with tempfile.NamedTemporaryFile(delete=False, suffix=".docx") as tmp:
tmp.write(resp.content)
tmp_path = tmp.name
try:
loader = Docx2txtLoader(tmp_path)
docs = loader.load()
text = "\n".join([d.page_content for d in docs])
finally:
os.remove(tmp_path)
elif "text/plain" in content_type or ".txt" in url_lower:
text = resp.content.decode("utf-8", errors="ignore")
else:
raise HTTPException(400, f"Unsupported document type: {content_type}")
# Clean and normalize text
text = re.sub(r'\s+', ' ', text).strip()
# Semantic chunking
splitter = RecursiveCharacterTextSplitter(
chunk_size=1000,
chunk_overlap=150,
separators=["\n\n", "\n", ". ", "! ", "? ", " ", ""]
)
return splitter.split_text(text) |