#KRAWLER: A WEB CRAWLER IN PYTHON3 Project Source code: See inside src/* Project docs: See inside docs/* Installing the system-wide dependencies: $ bash system-dependencies.sh $ sudo pip3 install -U -r requirements.txt Starting the crawler: $ pushd src; python3 main.py; popd 