Discord-Scraper

Pipeline to scrape prompt + image url pairs from Discord channels. The idea started by wanting to scrape the image-prompt pairs from share-dalle-3 Discord channel from LAION server. But now you can re-use the scraper to work with any channel you want.

How to use

Clone the repo git clone https://github.com/EduardoPach/LAION-Dalle-Scraper.git

Set up a virtual environment and install the requirements with pip install -r requirements.txt
Get your DISCORD_TOKEN and HF_TOKEN and add as environment variables.
1. DISCORD_TOKEN can be obtained by looking at developer tools in your Web Browser
2. HF_TOKEN can be obtained by logging in to HuggingFace and looking at your profile
Get the channel_id from the Discord channel you want to scrape. You can do this by enabling developer mode in Discord and right clicking the channel you want to scrape.
Create a condition_fn and a parse_fn that will be used to filter and parse the messages. You can use the ones I created as an example.
Create your scraping script and optionally your config.json

NOTE PAY ATTETNTION TO THE FUNC SIGNATURE OF parse_fn and condition_fn

import os
from typing import Any, Dict, List

from scraper import ScraperBot, ScraperBotConfig, HFDatasetScheme

def parse_fn(message: Dict[str, Any]) -> List[HFDatasetScheme]:
    ...

def condition_fn(message: Dict[str, Any]) -> bool:
    ...

if __name__ == "__main__":
    config_path = os.path.join(os.path.dirname(__file__), "config.json")
    config = ScraperBotConfig.from_json(config_path)

    bot = ScraperBot(config=config, parse_fn=parse_fn, condition_fn=condition_fn)
    bot.scrape(fetch_all=False, push_to_hub=False)

Main Components

ScraperBotConfig

Dataclass with configuration attributes to be used by the ScraperBot. You can create your own config.json file and load it with ScraperBotConfig.from_json(path_to_config).

attributes:

base_url: str, The base url of the Discord API (in chase it changes)
channel_id: str, The id of the channel you want to scrape
limit: int, The number of messages to fetch (from my tests the max allowed by Discord is 100)
hf_dataset_name: str, The name of the dataset you want to push to HuggingFace

ScraperBot

Implementation of the scraper. Get's the messages from the Discord API and filters them using the condition_fn. Then parses the messages using the parse_fn and pushes the dataset to HuggingFace.

attributes:

config: ScraperBotConfig, The configuration to be used by the bot
parse_fn: Callable[[Dict[str, Any]], List[HFDatasetScheme]], The function to parse the messages
condition_fn: Callable[[Dict[str, Any]], bool], The function to filter the messages

methods:

scrape(fetch_all: bool = False, push_to_hub: bool = False) -> Dataset

Scrapes the messages and optionally pushes the dataset to HuggingFace.

args:

fetch_all: bool, If True will fetch all the messages from the channel. If False will fetch only the messages that weren't processed yet.
push_to_hub: bool, If True will push the dataset to HuggingFace. If False will only return the dataset.

NOTE: If you want to push the dataset to HuggingFace you need to set the HF_TOKEN environment variable. NOTE 2: If the dataset doesn't exist in HuggingFace it will be created. If it already exists it will be updated.