Skip to content

higherrrrrrr/plex

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Plex: Simple and Transparent AI Bot Verification

Welcome to Plex, a lightweight and powerful tool for verifying the authenticity of AI-generated messages in the cryptocurrency ecosystem. By analyzing statistical patterns in AI outputs, Plex helps distinguish genuine bot-generated messages from potentially manipulated or malicious ones.


Building Trustworthy AI - Read on Mirror


Why Use Plex?

  1. Build Trust: Ensure AI-generated messages are authentic and reliable.
  2. Enhance Transparency: Promote trust in AI communications across the crypto landscape.
  3. Strengthen Security: Protect against manipulation by validating AI outputs with robust statistical methods.

What Does Plex Do?

Plex evaluates AI-generated messages by:

  • Analyzing log probabilities to assess the model's confidence in its predictions.
  • Calculating perplexity, a metric for determining how predictable the output is.
  • Generating a trust score to quantify authenticity.
  • Comparing outputs against trusted datasets to flag anomalies.

Quick Links


How It Works (Simplified)

  1. Submit Context and Output: Provide a sequence of input messages and the AI-generated output.
  2. Statistical Analysis: Plex evaluates the log probabilities of each token and calculates the perplexity of the message.
  3. Generate Trust Score: A trust score is computed, indicating the likelihood that the message originated from a legitimate AI model.

Features

  • Model Agnostic: Compatible with GPT-based models and any system that provides token-level log probabilities.
  • Simple API Integration: Easily integrate Plex with your crypto or AI applications.
  • Robust Analysis: Incorporates advanced statistical tests for reliable verification.
  • Lightweight and Fast: Minimal overhead for real-time validation needs.

Example Workflow

  1. Send a Request: Provide the AI context (message array) and the generated output.
  2. Get Results: Receive perplexity and trust scores as JSON responses.
  3. Evaluate Authenticity: Use the trust score to decide if the message is trustworthy.

API Overview

Endpoint: POST /calculate-trust

Request Body:

{
  "messages": [
    {"role": "user", "content": "What is Bitcoin?"},
    {"role": "assistant", "content": "Bitcoin is a decentralized digital currency."}
  ],
  "output": "It operates without a central authority."
}

Response:

{
  "trust_score": 0.87,
  "perplexity": 15.32,
  "using_default_key": false,
  "trust_classification": "HIGH",
  "trust_description": "The response appears highly reliable and consistent with expected AI behavior."
}

Get Started Today

Use Plex to bring trust, transparency, and security to your AI communications. Explore the API Documentation or dive into the Litepaper to learn more.

About

Trust scores for AI agents

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published