Bot Judol

Demo

Watch Demo Video

Overview

Bot Judol is a full-stack toolkit for moderating YouTube comments, specifically targeting online gambling promotion. It works similarly to YouTube Studio but is enhanced with AI-powered classification, allowing creators to delete spammy comments in just two clicks, run the AI and delete the flagged content.

Features

• Full OAuth-based sign-in with Google for authenticated access to YouTube data.
• Comment ingestion pipeline that fetches videos and comments, persists them, and tracks moderation status.
• Predicts online gambling promotion using an IndoBERT-powered classification engine with batch moderation directly from the dashboard.

Tech Stack

• Client: React 19, TypeScript, Vite, Tailwind CSS.
• Server: FastAPI, SQLAlchemy Async, Redis, PostgreSQL, Transformers, Hugging Face models.
• Engine: Python tooling for scraping, EDA, model training, and benchmarking with Transformers, PyTorch, MLflow.

Repository Layout

• client/ - React SPA and reusable components.
• engine/ - Data collection, experimentation, model training, and benchmarking scripts.
• server/ - FastAPI service, async database layer, and ML inference helpers.
• docker-compose.yml - Orchestrates Postgres, Redis, FastAPI, and React containers.

Prerequisites

• Docker and Docker Compose.
• Python 3.10+ and Node 18+ (only required for running services outside Docker).
• Access to a Google Cloud project with YouTube Data API v3 enabled.

Environment Setup

1. Copy each example file to a working .env:
   • cp client/.env.example client/.env
   • cp engine/.env.example engine/.env
   • cp server/.env.example server/.env
2. Fill in the placeholders with the credentials described below. The same Google OAuth client ID, client secret, and API key are shared across services.
3. Generate a FERNET_KEY for the server (see section below).

Running with Docker Compose

1. Ensure docker-compose.yml is updated with desired port bindings.
2. Start the stack:
   • docker compose up --build
3. The services expose:
   • React client at http://localhost:5173
   • FastAPI backend at http://localhost:8000
   • PostgreSQL at localhost:5433
   • Redis at localhost:6379
4. Use docker compose down to stop the stack and docker compose down -v to drop volumes.

Credential Setup

Google OAuth Client (Client + Server)

1. Visit https://console.cloud.google.com/apis/credentials and create an OAuth 2.0 Client ID (type: Web application).
2. Add authorized JavaScript origin http://localhost:5173 and redirect URI http://localhost:5173.
3. Download the credentials. Use the client_id for VITE_GOOGLE_CLIENT_ID and GOOGLE_CLIENT_ID, and the client_secret values for VITE_GOOGLE_CLIENT_SECRET and GOOGLE_CLIENT_SECRET.
4. Store them in the corresponding .env files.

Google API Key (Server + Engine)

1. In the same project, create an API key with access to YouTube Data API v3.
2. Restrict the key to the YouTube Data API if desired.
3. Use the key for GOOGLE_API_KEY (server .env) and API_KEY (engine .env).

Authorized Redirect URI for Backend

• If you deploy the backend, add the public URL to the OAuth client as an authorized redirect URI. The FastAPI callback endpoint expects <backend-url>/auth/callback.

Generate a Fernet key

Go to python REPL/CLI where fernet already installed. Run the following command and copy the printed value into server/.env as FERNET_KEY:

from cryptography.fernet import Fernet
print(Fernet.generate_key().decode())