from django.core.management import BaseCommand from django.utils import timezone from vacancies.hh_parser.models import Vacancy as ExternalVacancy from vacancies.main.models import Vacancy, JobTitle from flashrank import Ranker, RerankRequest import re tags_regex = re.compile('<.*?>') reranker = Ranker("ms-marco-TinyBERT-L-2-v2") class Command(BaseCommand): help = "Collect vacancies from hh.ru parser" def _remove_tags(self, text): return re.sub(tags_regex, "", text) def handle(self, *args, **options): job_titles = list(JobTitle.objects.values("id", "title")) passages = [{"text": job_title["title"], **job_title} for job_title in job_titles] queryset = ExternalVacancy.objects.filter(title__isnull=False, description__isnull=False) total_vacancies = queryset.count() for index, vacancy in enumerate(queryset): results = reranker.rerank(RerankRequest(query=vacancy.title, passages=passages)) ordered_results = sorted(results, key=lambda i: i["score"], reverse=True) job_title_id = ordered_results[0]["id"] vacancy, created = Vacancy.objects.get_or_create( external_id=vacancy.id, defaults=dict( job_title_id=job_title_id, min_salary_rub=vacancy.min_payment, max_salary_rub=vacancy.max_payment, company_name="", requirements=self._remove_tags(vacancy.description), content=self._remove_tags(vacancy.description), timestamp=timezone.make_aware(vacancy.created_at), link=vacancy.link, ), ) print(f"{index+1}/{total_vacancies} Vacancy: {vacancy}, created: {created}")