nostrdvm/nostr_dvm/tasks/textextraction_pdf.py

119 lines
4.3 KiB
Python

import json
import os
import re
from nostr_sdk import Kind
from nostr_dvm.interfaces.dvmtaskinterface import DVMTaskInterface, process_venv
from nostr_dvm.utils.admin_utils import AdminConfig
from nostr_dvm.utils.definitions import EventDefinitions
from nostr_dvm.utils.dvmconfig import DVMConfig, build_default_config
from nostr_dvm.utils.nip88_utils import NIP88Config
from nostr_dvm.utils.nip89_utils import NIP89Config, check_and_set_d_tag
from nostr_dvm.utils.nostr_utils import get_event_by_id
"""
This File contains a Module to extract Text from a PDF file locally on the DVM Machine
Accepted Inputs: Url to pdf file, Event containing an URL to a PDF file
Outputs: Text containing the extracted contents of the PDF file
Params: None
"""
class TextExtractionPDF(DVMTaskInterface):
KIND: Kind = EventDefinitions.KIND_NIP90_EXTRACT_TEXT
TASK: str = "pdf-to-text"
FIX_COST: float = 0
dependencies = [("nostr-dvm", "nostr-dvm"),
("pypdf", "pypdf==3.17.1")]
async def init_dvm(self, name, dvm_config: DVMConfig, nip89config: NIP89Config, nip88config: NIP88Config = None,
admin_config: AdminConfig = None, options=None):
dvm_config.SCRIPT = os.path.abspath(__file__)
async def is_input_supported(self, tags, client=None, dvm_config=None):
for tag in tags:
if tag.as_vec()[0] == 'i':
input_value = tag.as_vec()[1]
input_type = tag.as_vec()[2]
if input_type != "url" and input_type != "event":
return False
return True
async def create_request_from_nostr_event(self, event, client=None, dvm_config=None):
request_form = {"jobID": event.id().to_hex()}
# default values
input_type = "url"
input_content = ""
url = ""
for tag in event.tags().to_vec():
if tag.as_vec()[0] == 'i':
input_type = tag.as_vec()[2]
input_content = tag.as_vec()[1]
if input_type == "url":
url = input_content
# if event contains url to pdf, we checked for a pdf link before
elif input_type == "event":
evt = await get_event_by_id(input_content, client=client, config=dvm_config)
url = re.search("(?P<url>https?://[^\\s]+)", evt.content()).group("url")
options = {
"url": url,
}
request_form['options'] = json.dumps(options)
return request_form
async def process(self, request_form):
from pypdf import PdfReader
from pathlib import Path
import requests
options = self.set_options(request_form)
try:
file_path = Path('temp.pdf')
response = requests.get(options["url"])
file_path.write_bytes(response.content)
reader = PdfReader(file_path)
number_of_pages = len(reader.pages)
text = ""
for page_num in range(number_of_pages):
page = reader.pages[page_num]
text = text + page.extract_text()
os.remove('temp.pdf')
return text
except Exception as e:
raise Exception(e)
# We build an example here that we can call by either calling this file directly from the main directory,
# or by adding it to our playground. You can call the example and adjust it to your needs or redefine it in the
# playground or elsewhere
def build_example(name, identifier, admin_config):
dvm_config = build_default_config(identifier)
admin_config.LUD16 = dvm_config.LN_ADDRESS
# Add NIP89
nip89info = {
"name": name,
"picture": "https://image.nostr.build/c33ca6fc4cc038ca4adb46fdfdfda34951656f87ee364ef59095bae1495ce669.jpg",
"about": "I extract text from pdf documents. I only support Latin letters",
"supportsEncryption": True,
"acceptsNutZaps": False,
"nip90Params": {}
}
nip89config = NIP89Config()
nip89config.DTAG = check_and_set_d_tag(identifier, name, dvm_config.PRIVATE_KEY, nip89info["picture"])
nip89config.CONTENT = json.dumps(nip89info)
return TextExtractionPDF(name=name, dvm_config=dvm_config, nip89config=nip89config,
admin_config=admin_config)
if __name__ == '__main__':
process_venv(TextExtractionPDF)