Purpose

This document is a machine-ready ingestion prompt and quick-reference catalog for the GitHub organization at github.com/xtends. It is designed to help a person or AI system rapidly understand the repository landscape, identify architectural themes, and extract reusable patterns from the broadest collection of LLM tooling, desktop compositing, and intelligent application development resources assembled under one organization.

The xtends organization is the extension layer of the augmented machine learning ecosystem — a deliberate, wide-net collection of 100 repositories spanning every surface where LLMs, intelligent agents, and autonomous systems touch the real world.

Where augml provides the core local runtime, xtends provides everything that plugs into it: prompt engineering, code generation, desktop applications, 3D rendering, speech pipelines, blockchain analytics, browser extensions, tokenizers, agent frameworks, workflow automation, window management, and privacy tools.

The collection logic:

If it extends what a local LLM can do → it belongs in xtends
If it extends what a developer can build with AI → it belongs in xtends
If it extends the user's sovereignty over their computing environment → it belongs in xtends

This is the largest reference organization in the ecosystem — 100 repositories across 10 functional domains — intentionally broad because the extension surface of machine learning is broad.

The primary interaction surface. Every way a human talks to a model.

Logic: Cover every interaction modality — browser, desktop, terminal, voice, extension, mobile — so that a local or remote model can be accessed from any context.

Systems that go beyond chat — they plan, execute, iterate, and self-correct.

Logic: Autonomous agents are the destination for augmented ML. These repos represent the full spectrum — from single-agent loops (Auto-GPT) to collaborative software teams (ChatDev).

The craft of making models do what you mean.

Logic: Prompts are the programming language of LLMs. These repos collect the community's best knowledge about writing them effectively.

Model serving, routing, tokenization, and production deployment.

Logic: Running one model is augml's domain. Running many models, routing between them, managing their context windows, and tokenizing efficiently — that's xtends territory.

Using AI to write code, generate applications, and automate development.

Logic: Code generation is the highest-leverage application of LLMs for developers. These repos cover the full spectrum from CLI tools to visual-to-code pipelines.

Voice input, transcription, text-to-speech, and audio visualization.

Logic: Voice is the natural interface. These repos build the pipeline from speech → text → reasoning → response → speech, entirely locally.

Where AI meets spatial computing and visual generation.

Logic: The extension of ML into spatial and visual domains — from 3D rendering to generative humans to on-chain art.

Complete desktop window manager ecosystem — the visual sovereignty layer.

Logic: Desktop compositing is where the user's visual environment lives. Compiz provides the window management layer for bankonOS and sovereign desktop systems. The complete stack — core, plugins, config, decorator, themes — is preserved here as a buildable unit.

Chain analytics, cross-chain monitoring, IPFS, Farcaster, and ordinals.

Logic: ML + blockchain = intelligent on-chain agents. These repos provide the chain data pipelines, decentralized content access, and social protocol integration that connect AI systems to on-chain state.

Tools that protect user sovereignty and extend daily productivity.

Non-forked repositories — native to xtends.

FAICE — Framework for Autonomous and Intelligent Computer Expressions — is the philosophical anchor of the xtends collection. It represents the thesis that computer expressions should be autonomous, intelligent, and framework-driven rather than ad-hoc.

This organization is not a bookmark folder. It is a complete extension surface for local-first AI systems covering:

• Every chat interface pattern (web, desktop, terminal, extension, voice, bot)
• Every agent architecture (Auto-GPT, ChatDev, function-calling, Prolog)
• Every prompt engineering resource worth collecting
• Every model serving pattern (local, proxy, ensemble, memory-managed)
• The complete Compiz desktop stack (compositor + plugins + config + decorator + themes)
• Blockchain analytics and Web3 agent tooling
• Speech-to-text and text-to-speech pipelines
• Code generation from natural language and from screenshots
• Privacy and sovereignty tools (ad blocking, private video, sandboxed execution)

The unifying principle: if it extends what a person can do with a local machine and local intelligence, it belongs here.