What problem it solves

Researchers, red-teamers, and system tinkerers need a chat interface that can work with multiple hosted and local models while exposing evaluation, perturbation, and sampling controls. They also need clear visibility into data flows, storage, telemetry, and deployment boundaries so they can make informed choices about privacy and self-hosting.

Key features

• Multi-model races and evaluation — GODMODE CLASSIC races several prompt and model combinations, while ULTRAPLINIAN evaluates responses across five model tiers using a composite score. Configured local models can also join the race.

  Source: README

• Local and remote model support — The interface supports OpenRouter, Venice, and OpenAI-compatible local servers. Local setups can use Ollama, LM Studio, llama.cpp, or vLLM.

  Source: README, LOCAL_MODELS.md

• Red-teaming input perturbation — Parseltongue classifies prompts and applies obfuscation techniques across configurable intensity levels. Its documented transformations include leetspeak, braille, morse, Unicode substitutions, and layered encodings.

  Source: README

• Context-adaptive sampling — AutoTune classifies a query into one of 20 contexts and selects parameters such as temperature, top_p, top_k, frequency penalty, presence penalty, and repetition penalty.

  Source: README

• Local history and explicit privacy controls — Conversations and settings remain in browser storage, with export and import support. No-Log and Local-only modes can disable app telemetry, while Local-only mode also excludes OpenRouter and Venice model calls.

  Source: README

Quick start

Copied verbatim from the project's README.

# Clone the repository
git clone https://github.com/elder-plinius/G0DM0D3.git
cd G0DM0D3

# Open directly in your browser
open index.html
# or serve locally
python3 -m http.server 8000

ollama pull qwen3:8b
ollama serve

Who it may suit

This project may suit AI red-teamers, cognition researchers, and system tinkerers who want model comparison, prompt perturbation, and configurable sampling in a self-hostable interface. It may also suit users who want to run model servers on their own hardware and retain conversation history in browser storage.

Topics to discuss

• How should comparative model evaluation account for different response styles and refusal behavior?

• Which local-model workflows and provider adapters should the community prioritize next?

• How can telemetry and dataset contribution remain transparent while supporting reproducible research?

Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.

Read the original README https://github.com/elder-plinius/G0DM0D3#readme

ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.