<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> Lavish-AXI addresses the challenge of effective human-agent collaboration on rich HTML artifacts. It provides a local-first, zero-cloud-dependency editor that allows users to directly annotate HTML elements and text, sending precise feedback to AI agents. This enhances the feedback loop for agent-generated content, moving beyond screenshots and text descriptions to leverage HTML's inherent interactivity. It is designed for developers and power users interacting with AI agents that produce visual or structured HTML outputs.

How It Works

Lavish-AXI functions as an Agent-executable Interface (AXI). When an agent generates an HTML artifact, lavish-axi <file_path> opens it in a local browser. A lightweight SDK within the artifact enables users to annotate specific text ranges or elements and queue feedback. The CLI then long-polls for agent responses, facilitating a tight, interactive loop. This approach prioritizes agent ergonomics with efficient polling and contextual disclosure, minimizing token usage. Artifacts are portable, rendering identically outside Lavish, and local assets must use relative paths. Live reload with scroll position preservation is supported.

Quick Start & Requirements

Installation is streamlined: add the skill via npx skills add kunchenguid/lavish-axi --skill lavish, or run the CLI directly with npx -y lavish-axi for agents capable of executing npx commands without prior setup. For advanced features like session hooks, install globally (npm install -g lavish-axi) and run lavish-axi setup hooks. Building from source requires pnpm. No specific hardware like GPUs is mandated.

Highlighted Details

• Local-First Operation: Entirely local, ensuring zero cloud dependency and enhanced privacy.
• Precise Human-AI Feedback: Enables annotation of text ranges and elements for granular agent instruction.
• Portable HTML Artifacts: Generated HTML renders consistently across Lavish and standard browsers.
• Agent Ergonomics: Optimized for agent interaction via efficient long-polling and contextual disclosure.
• Live Reload: Automatic artifact reloading in the browser, preserving scroll position.

Maintenance & Community

The provided README does not detail specific contributors, sponsorships, or community channels (e.g., Discord, Slack). Its integration as an AXI suggests compatibility with agent frameworks like Hermes.

Licensing & Compatibility

The project's license is not specified in the README. This omission presents a significant adoption blocker, particularly for commercial use or integration into closed-source projects.

Limitations & Caveats

Local assets must use relative paths; root-prefixed paths are unsupported. Binding the server beyond loopback (127.0.0.1) exposes an unauthenticated interface, requiring careful network configuration. The absence of a clear license is a critical limitation for evaluating adoption.