PredictBot

An AI-powered X/Twitter bot that creates prediction markets from tweets using Grok.

When someone makes a claim like "I bet China won't invade Taiwan", anyone can reply mentioning @marketmake67808 to instantly create a tradeable prediction market.

Features

• Grok AI Integration - Uses xAI's Grok model to understand context and frame predictions
• Automatic Market Creation - Extracts claims from tweets and creates YES/NO markets
• Rich Link Previews - Auto-generated OG images showing market probabilities
• Real-time Streaming - Monitors X mentions via filtered stream
• Conversation Context - Understands full thread context for accurate prediction framing

How It Works

1. Someone tweets: "I bet the Fed won't cut rates this year"
2. Another user replies: "Hey @marketmake67808 create a market for this"
3. The bot creates a market: "Will the Fed cut rates by December 2026?"
4. Bot replies with a link showing 65% YES / 35% NO with a preview card

Quick Start

# Clone the repository
git clone https://github.com/your-username/predictbot.git
cd predictbot

# Install dependencies
npm install

# Copy environment template
cp .env.example .env

# Edit .env with your credentials (see Configuration section)

# Build and run
npm run build
npm start

Configuration

Required Environment Variables

Variable	Description
X_BEARER_TOKEN	X API Bearer Token (for reading)
X_API_KEY	X API Key
X_API_SECRET	X API Secret
X_ACCESS_TOKEN	X Access Token (for posting)
X_ACCESS_TOKEN_SECRET	X Access Token Secret
GROK_API_KEY	xAI Grok API Key
BOT_USERNAME	Your bot's X username (default: marketmake67808)
WEBSITE_URL	Base URL for link previews

Optional Environment Variables

Variable	Default	Description
BOT_PLUGIN	prediction-market	Which plugin to use
PLUGIN_SANDBOX_MODE	true	Use sandbox/mock data

Getting API Keys

1. X/Twitter API: Apply at developer.twitter.com

   • Create a project with "Read and Write" permissions
   • Generate Bearer Token, API Key/Secret, Access Token/Secret

2. Grok API: Get access at x.ai

   • Create an API key in the dashboard

---

Deployment

The bot requires a long-running process (not serverless).

Docker (Recommended)

# Build the image
docker build -t predictbot .

# Run with environment variables
docker run -d \
  -e X_API_KEY=your_key \
  -e X_API_SECRET=your_secret \
  -e X_BEARER_TOKEN=your_token \
  -e X_ACCESS_TOKEN=your_access_token \
  -e X_ACCESS_TOKEN_SECRET=your_access_secret \
  -e GROK_API_KEY=your_grok_key \
  -e BOT_USERNAME=marketmake67808 \
  -e WEBSITE_URL=https://marketmaker-nine.vercel.app \
  predictbot

Or use docker-compose:

# Create .env file with your credentials, then:
docker-compose up -d

Website Deployment (Vercel)

The web/ directory contains a Next.js app for link previews:

1. Push to GitHub
2. Import in Vercel
3. Set root directory to web
4. Add environment variables:
   • KV_REST_API_URL - Upstash Redis URL
   • KV_REST_API_TOKEN - Upstash Redis token
   • NEXT_PUBLIC_BASE_URL - Your Vercel URL
   • NEXT_PUBLIC_BOT_USERNAME - marketmake67808
5. Deploy

---

Project Structure

├── src/
│   ├── bot.ts                # Bot orchestration
│   ├── config.ts             # Configuration
│   ├── types.ts              # Core types
│   │
│   ├── framework/            # XBot Framework core
│   │   ├── grok-client.ts    # Grok AI integration
│   │   └── types.ts          # Framework types
│   │
│   ├── plugins/              # Bot plugins
│   │   ├── index.ts          # Plugin registry
│   │   └── prediction-market/# Prediction market bot
│   │
│   └── services/             # Core services
│       ├── x-client.ts       # X API client
│       ├── filtered-stream.ts# Real-time stream
│       └── conversation.ts   # Thread handling
│
├── web/                      # Next.js website
│   ├── app/
│   │   ├── r/[id]/          # Market preview pages
│   │   └── api/og/[id]/     # OG image generation
│   └── lib/
│       ├── types.ts         # Data types
│       └── config.ts        # Branding config
│
├── Dockerfile               # Container build
├── docker-compose.yml       # Easy deployment
└── package.json

---

Prediction Market Plugin

The prediction market plugin (src/plugins/prediction-market/) handles:

Tools Available to Grok

1. create_market - Creates a new prediction market

   • Extracts the claim from conversation context
   • Frames it as a YES/NO question
   • Assigns a category and resolution date
   • Returns market with initial probability

2. search_markets - Searches existing markets

   • Checks if a similar market already exists
   • Returns matching markets by topic

Market Categories

• Politics
• Economics
• Sports
• Technology
• Entertainment
• Science
• World Events
• Crypto

Example Interaction

Original Tweet:

"There's no way Bitcoin hits $100k this year"

Reply:

"@marketmake67808 make a market"

Bot Response:

"Market created: 'Will Bitcoin reach $100,000 by Dec 2026?' Currently at 72% YES." [Link with rich preview showing probability bars]

---

Future Enhancements

• Real API Integration - Connect to Kalshi, Polymarket, or Manifold APIs
• Market Resolution - Automatic resolution based on verified outcomes
• Trading - Allow users to buy/sell shares directly via replies
• Notifications - Alert market creators when resolution approaches
• Leaderboards - Track prediction accuracy

---

License

MIT