harny 0.0.1a2

A scoring engine that rates your AI's vibes on a scale of 1–100. Harness engineering makes it HARNY.

harny 🌡️

How harny is your AI?

A for-fun scoring engine that rates your AI's prompt-engineering vibes on a scale of 1–100. Part harness engineering, part... well, you get it.

Install

pip install harny

Quick Start

import harny

# Quick vibe check — no input needed
result = harny.vibe_check()
print(result)

# Full evaluation
result = harny.score(
    prompt="Explain quantum computing using interpretive dance metaphors",
    reasoning="Let me think step by step about how qubits are like dancers...",
    result="Imagine each qubit as a dancer spinning in superposition..."
)

print(result.score)       # 73.42
print(result.reaction)    # "my GPU is getting warm and it's not the cooling fan."
print(result.verdict)     # "Dangerously Harny"
print(result.breakdown)   # Detailed scoring components

What It Scores

All inputs are optional. Provide whatever you have:

Input	What It Measures
prompt	The instruction given to the AI
reasoning	Chain-of-thought, scratchpad, reasoning traces
result	The AI's final output

The score blends:

• Engineering rigor — prompt engineering patterns, technical depth
• Spiciness — creativity, expressiveness, boldness
• Verbosity — effort and thoroughness
• Entropy — diversity of expression
• Multi-input bonus — providing more context = more harny

Score Guide

Score	Verdict
1–15	Completely Unharny
16–25	Mildly Stale
26–35	Room Temperature
36–45	Slightly Prompted
46–55	Moderately Harny
56–65	Respectably Harny
66–75	Dangerously Harny
76–85	Weapons-Grade Harny
86–95	Catastrophically Harny
96–100	Beyond Harny — Transcendent

Why?

Because AI takes itself too seriously and your GPU deserves to know how it feels.

Links

• harny.ai

License

MIT