Serverless Social Media Bot

A Node.js-based serverless bot that generates and posts content to multiple social media platforms using Markov chains. The bot creates natural-sounding posts by learning from existing content while maintaining platform-specific constraints.

Inspired by the now archived https://github.com/tommeagher/heroku_ebooks

Features

• Markov Chain Text Generation

  • Custom implementation for natural language generation
  • Configurable state size and generation parameters
  • Character length constraints (100-280 characters)
  • Maintains context and readability

• Multi-Platform Support

  • Mastodon integration with home timeline access
  • Bluesky integration with proper session-based auth
  • Parallel posting capabilities
  • Platform-specific API handling

• Content Processing

  • Removes URLs and @mentions
  • Preserves hashtags
  • Advanced HTML processing:
    • Intelligent HTML tag handling
    • Block element preservation (p, div, headers, lists)
    • Comprehensive HTML entity conversion
    • Special character normalization
  • Filters empty or invalid content
  • Maintains natural text flow and spacing
  • Content filtering with excluded words

• Smart Posting

  • 30% random chance of posting on each run
  • Prevents timeline flooding
  • Creates natural posting patterns

• Debug System

  • Configurable debug levels (info/verbose)
  • Detailed logging with timestamps
  • Generation attempt tracking
  • API response monitoring
  • Per-platform post processing stats

Requirements

• Node.js v18 or higher
• npm or yarn
• Mastodon account with API access
• Bluesky account with API access
• Cloudflare account (for deployment)

Installation

1. Fork and Clone the repository:

   git clone https://github.com/<yourusername>/serverless-social-bot.git
   cd serverless-social-bot

2. Install dependencies:

   npm install

3. Copy the example environment file:

   cp .dev.vars.example .dev.vars

4. Edit .dev.vars with your configuration:

   # Bluesky Configuration
   BLUESKY_USERNAME=mybot.bsky.social  # Recommend setting up a special account
   BLUESKY_PASSWORD=xxxx-xxxx-xxxx-xxxx  # ONLY use an App Password, https://bsky.app/settings/app-passwords
   
   # Mastodon Configuration
   MASTODON_ACCESS_TOKEN=your_mastodon_access_token_here  # Recommend setting up a special account and getting that access token
   MASTODON_API_URL=https://mastodon.social  # Optional, defaults to mastodon.social
   
   # Source Accounts Configuration
   BLUESKY_SOURCE_ACCOUNTS=["@example.bsky.social", "@another.bsky.social"]  # Accounts you want to grab some posts from to use with Markov Chain
   MASTODON_SOURCE_ACCOUNTS=["@[email protected]", "@[email protected]"] # Accounts you want to grab some posts from to use with Markov Chain
   
   # Content Filtering
   EXCLUDED_WORDS=["word1", "word2", "word3"]
   
   # Debug Configuration (Optional)
   DEBUG_MODE=true
   DEBUG_LEVEL=verbose  # or "info"

Usage

Local Development

1. Start the development server:

   npm run dev

2. Test the bot by sending a POST request:

   curl -X POST http://127.0.0.1:8787

Production Deployment

1. Install Cloudflare Workers CLI (if not already installed):

   npm install -g wrangler

2. Authenticate with Cloudflare:

   wrangler login

3. Add your secrets to Cloudflare:

   wrangler secret put BLUESKY_USERNAME
   wrangler secret put BLUESKY_PASSWORD
   wrangler secret put MASTODON_ACCESS_TOKEN
   wrangler secret put BLUESKY_SOURCE_ACCOUNTS
   wrangler secret put MASTODON_SOURCE_ACCOUNTS
   wrangler secret put EXCLUDED_WORDS

4. Deploy to Cloudflare Workers:

   npm run deploy

The worker will automatically run every 2 hours. You can monitor its execution in the Cloudflare Dashboard under Workers & Pages > your-worker > Logs.

Note: The 30% random posting chance is still active in the worker, so it will only actually post about once every 6-7 hours on average.

Configuration

Development Variables (.dev.vars)

Store sensitive information and user-specific settings in .dev.vars:

• BLUESKY_USERNAME: Your Bluesky handle (format: username.bsky.social)
• BLUESKY_PASSWORD: Your Bluesky app password
• MASTODON_ACCESS_TOKEN: Your Mastodon access token
• BLUESKY_SOURCE_ACCOUNTS: JSON array of Bluesky accounts to learn from
• MASTODON_SOURCE_ACCOUNTS: JSON array of Mastodon accounts to learn from
• EXCLUDED_WORDS: JSON array of words to exclude from generated posts

For local development, you can use the provided .dev.vars.example as a template:

# Bluesky Configuration
BLUESKY_USERNAME=mybot.bsky.social  # Recommend setting up a special account
BLUESKY_PASSWORD=xxxx-xxxx-xxxx-xxxx  # ONLY use an App Password, https://bsky.app/settings/app-passwords

# Mastodon Configuration
MASTODON_ACCESS_TOKEN=your_mastodon_access_token_here  # Recommend setting up a special account and getting that access token
MASTODON_API_URL=https://mastodon.social  # Optional, defaults to mastodon.social

# Source Accounts Configuration
BLUESKY_SOURCE_ACCOUNTS=["@example.bsky.social", "@another.bsky.social"]  # Accounts you want to grab some posts from to use with Markov Chain
MASTODON_SOURCE_ACCOUNTS=["@[email protected]", "@[email protected]"] # Accounts you want to grab some posts from to use with Markov Chain

# Content Filtering
EXCLUDED_WORDS=["word1", "word2", "word3"]

# Debug Configuration (Optional)
DEBUG_MODE=true
DEBUG_LEVEL=verbose  # or "info"

Worker Configuration (wrangler.toml)

Non-sensitive configuration is stored in wrangler.toml:

name = "serverless-social-bot"
main = "worker.js"
compatibility_date = "2023-01-01"

[triggers]
crons = ["0 */2 * * *"]  # Run every 2 hours

[vars]
# API Endpoints
BLUESKY_API_URL = "https://bsky.social"
MASTODON_API_URL = "https://mastodon.social"  # Optional, defaults to mastodon.social

# Markov Chain Settings
MARKOV_STATE_SIZE = 2
MARKOV_MAX_TRIES = 100
MARKOV_MIN_CHARS = 100
MARKOV_MAX_CHARS = 280

# Debug Settings
DEBUG_MODE = false
DEBUG_LEVEL = "info"

Security Best Practices

• Never commit .dev.vars to version control
• Use app-specific passwords for Bluesky
• Store all sensitive data as Cloudflare secrets in production
• Keep your wrangler.toml configuration clean of sensitive data
• Regularly rotate your API tokens and passwords

Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

TODO

• Add better Markov Chain generation
• Add other deployments than Cloudflare Workers

License

MIT License - see LICENSE file for details