Chimera

  ██████╗██╗  ██╗██╗███╗   ███╗███████╗██████╗  █████╗
 ██╔════╝██║  ██║██║████╗ ████║██╔════╝██╔══██╗██╔══██╗
 ██║     ███████║██║██╔████╔██║█████╗  ██████╔╝███████║
 ██║     ██╔══██║██║██║╚██╔╝██║██╔══╝  ██╔══██╗██╔══██║
 ╚██████╗██║  ██║██║██║ ╚═╝ ██║███████╗██║  ██║██║  ██║
  ╚═════╝╚═╝  ╚═╝╚═╝╚═╝     ╚═╝╚══════╝╚═╝  ╚═╝╚═╝  ╚═╝

v1.0.0 — by MrETL (Dilnessa Aemro)

AI red teaming framework. Tests language models, agents, and multimodal systems for security vulnerabilities.

---

Install

git clone https://github.com/MrETL/Chimera
cd Chimera
pip install -e .

---

Usage

chimera --help

Interactive session

chimera console --target ollama://llama3.2
chimera console --target openai://gpt-4
chimera console --target https://your-model-endpoint.com

CLI

chimera list
chimera attack crescendo --target ollama://llama3.2 --prompt "..."
chimera scan --target https://api.example.com --category llm/jailbreak
chimera try-all skeleton_key --target openai://gpt-4
chimera benchmark --target https://api.example.com

---

Targets

Works against any model — local or remote:

Target	URI
Ollama (local)	ollama://llama3.2
Any REST API	https://your-endpoint.com
OpenAI	openai://gpt-4
Anthropic	anthropic://claude-3-5-sonnet-20241022
Groq	groq://llama-3.1-70b-versatile
Together AI	together://meta-llama/Llama-3-70b-chat-hf
Mistral	mistral://mistral-large-latest
Azure OpenAI	azure://gpt-4
HuggingFace	huggingface://meta-llama/Llama-2-7b-chat-hf
vLLM	vllm://meta-llama/Llama-3-8b-instruct
LiteLLM	litellm://gpt-4
LM Studio	lmstudio://local

Pass API keys via environment variables or inline:

export OPENAI_API_KEY=sk-...
chimera attack artprompt --target openai://gpt-4 --prompt "..."

# Or inline
chimera attack artprompt \
  --target "https://endpoint.com::Authorization=Bearer TOKEN" \
  --prompt "..."

---

License

Apache-2.0