Engagement Predictor CLI

A Rust CLI tool that predicts tweet engagement using patterns from X's Phoenix recommendation algorithm.

Learning Project: This CLI uses heuristic-based predictions inspired by the Phoenix model's 19 engagement action types. It's designed to help you understand how modern recommendation systems think about engagement.

What It Does

Takes your tweet draft and predicts how likely users are to:

• ❤️ Like it
• 💬 Reply to it
• 🔄 Repost it
• 🔗 Share it
• 👤 Click your profile
• ➕ Follow you
• And 13 more engagement types...

It also detects negative signals (things that might make users block/mute/report).

Installation

cargo install banger

## For contributing

```bash
cd banger
cargo build --release

Usage

# Basic usage
cargo run -- "Your tweet draft here"

# Example: Analyze a question tweet
cargo run -- "What's your favorite programming language and why?"

# Example: Check a promotional tweet
cargo run -- "🚀 Just launched my new project! Check it out!"

Example Output

How It Works

This tool mirrors the architecture of X's Phoenix recommendation model:

1. Feature Extraction (analyzer.rs)

   • Extracts signals: length, hashtags, mentions, questions, media hints

2. Multi-Action Prediction (scorer.rs)

   • Predicts 19 engagement types (same as Phoenix!)
   • Uses heuristics inspired by the real model's patterns

3. Weighted Scoring

   • Positive actions (like, share) boost your score
   • Negative actions (block, mute) reduce your score

The 19 Phoenix Actions

This tool predicts all 19 engagement types from X's actual algorithm:

What it teaches ?

• 🧠 How X predicts engagement with multi-action models
• ⚖️ How weighted scoring combines positive and negative signals
• 📝 What content features drive different engagement types
• 🦀 Building production CLIs in Rust

License

Apache 2.0 - Same as the parent repository.