TTC Compression - Dify Plugin

Reduce LLM costs by compressing your prompts before they reach the model.

How It Works

┌──────────────┐     ┌─────────────────────┐     ┌───────────┐
│  User Input  │ ──▶ │  TTC Compression    │ ──▶ │  LLM Node │
│              │     │  (reduces tokens)   │     │           │
└──────────────┘     └─────────────────────┘     └───────────┘

This plugin sits between your input and the LLM. It compresses the text while preserving meaning, so the LLM receives fewer tokens but understands the same thing. The output from the compression node connects directly to your LLM node's input.

Quick Start

1. Get an API key from thetokencompany.com
2. Install the plugin in Dify
3. Go to Tools → Find "TTC Compression" → Click Authorize → Enter your API key

Using in Workflows

1. Add a Semantic Compression tool node to your workflow
2. Connect your input text to the text parameter
3. Connect the tool's output to your LLM node's input ← This is the key step
4. The LLM now receives compressed text, using fewer tokens

Parameters

Parameter	Description
text	The text to compress (required)
aggressiveness	Compression level: 0.2 (light), 0.5 (moderate), 0.7 (aggressive)
protect_json	Keep JSON structures intact (default: true)

Output

• Text output: The compressed text → use this as your LLM input
• JSON output: Stats including tokens saved, compression ratio, processing time

Compression Levels

Level	Token Reduction	Best For
Light (0.2)	10-20%	When quality is critical
Moderate (0.5)	25-40%	General use
Aggressive (0.7)	40-60%	Maximum savings

Support

• Website: thetokencompany.com
• Email: support@thetokencompany.com

License

MIT License