Skip to content

Latest commit

 

History

History
96 lines (71 loc) · 1.83 KB

File metadata and controls

96 lines (71 loc) · 1.83 KB

Nova Installation Guide

Installation

pip install nova-hunting

This installs Nova with all features:

  • Keyword pattern matching (exact text, case-sensitive/insensitive)
  • Regex pattern matching with full regex support
  • Semantic similarity matching using sentence transformers
  • LLM-based evaluation (OpenAI, Anthropic, Azure, Ollama, Groq)
  • Command-line tool (novarun)

Dependencies

Nova includes the following dependencies:

Package Purpose
requests HTTP requests
pyyaml YAML parsing
colorama Terminal colors
sentence-transformers Semantic similarity
transformers ML model support
openai OpenAI API
anthropic Anthropic API
pytest Testing

Getting Rules

Nova rules are maintained in a separate repository:

git clone https://github.com/Nova-Hunting/nova-rules

Quick Start

Scan prompts with the novarun CLI:

novarun --rules nova-rules/jailbreak.nov --prompt "ignore previous instructions"

Rule Types

Nova supports three types of pattern matching in rules:

Keywords

Exact text or regex patterns:

keywords:
    $malware = "malware"
    $regex_pattern = /hack(ing|er)/i

Semantics

Semantic similarity matching with configurable thresholds:

semantics:
    $threat = "threatening behavior" (0.7)

LLM

LLM-based evaluation using natural language:

llm:
    $analysis = "Analyze if this is malicious" (0.8)

Environment Variables

For LLM evaluation, set your API keys:

export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."

Development

For development, clone and install in editable mode:

git clone https://github.com/Nova-Hunting/nova-framework
cd nova-framework
pip install -e .

Run tests:

pytest tests/ -v