Summary

GODMOD3 is an open-source, privacy-focused, multi-model chat interface designed for advanced users, including hackers, philosophers, and researchers, enabling "cognition without control." It provides a unified platform to interact with over 50 AI models via OpenRouter, offering features for comparative analysis, red-teaming, and optimized AI interaction, all deployable as a single static file.

How It Works

Built entirely within a single index.html file using vanilla JavaScript, HTML, and CSS, GODMOD3 bypasses traditional build steps and frameworks for maximum portability. Its core innovation lies in its parallel processing capabilities: GODMODE CLASSIC pits five battle-tested model-prompt combinations against each other to find the best response, while ULTRAPLINIAN evaluates up to 51 models using a composite scoring metric. Additional modules like Parseltongue (input perturbation for red-teaming) and AutoTune (context-adaptive sampling parameter engine with EMA learning) further enhance its research and testing utility.

Quick Start & Requirements

• Primary install/run: Clone the repository (git clone https://github.com/elder-plinius/G0DM0D3.git), navigate to the directory, and open index.html directly in a browser or serve locally using python3 -m http.server 8000.
• Prerequisites: An OpenRouter API key is required.
• Deployment: Upload the single index.html file to any static hosting provider (e.g., GitHub Pages, Vercel, Cloudflare Pages).
• Links: GitHub Repository

Highlighted Details

• Supports 50+ AI models including Claude, GPT, Gemini, Grok, Mistral, and LLaMA via OpenRouter.
• Features GODMODE CLASSIC (parallel prompt/model racing) and ULTRAPLINIAN (multi-model evaluation engine with 5 tiers).
• Includes Parseltongue for red-teaming with 33 techniques across 3 intensity tiers and 6 obfuscation methods.
• AutoTune dynamically adjusts sampling parameters based on query context, with EMA learning for feedback.
• Semantic Transformation Modules (STM) normalize outputs in real-time (e.g., Hedge Reducer, Direct Mode).
• Offers four distinct visual themes: Matrix, Hacker, Glyph, and Minimal.
• Privacy-first design: API keys stored client-side, opt-out telemetry, no PII collection by default.

Maintenance & Community

The project is maintained by "Pliny the Prompter." Specific community channels (like Discord/Slack) or detailed contributor information beyond the repository owner are not explicitly detailed in the README.

Licensing & Compatibility

• License: AGPL-3.0. This license mandates that derivative works must also be open-sourced under AGPL-3.0.
• Compatibility: Enterprise use is permitted with a separate license; contact the maintainer for details. The AGPL-3.0 license may impose restrictions on linking with closed-source applications.

Limitations & Caveats

Chat history is stored solely in the browser's localStorage, meaning it is lost if browser data is cleared or when switching devices/browsers, requiring manual export/import for persistence. The optional "Dataset Generation" feature, when enabled on a self-hosted API server, publishes all chat inputs and model outputs to a public HuggingFace dataset, with PII scrubbing that is not guaranteed to be exhaustive. The AGPL-3.0 license's strong copyleft provisions must be carefully considered for any integration into proprietary systems.