Pi Agent, OpenWhispr and Orca: HomeLab workflow tutorial

Christian Lempa demonstrates a mostly Mac-focused HomeLab workflow built around Markdown documentation, terminal tools and AI-assisted deployment. Obsidian stores service records and project context in plain files that his terminal agents can read. Ghostty, Zsh plugins and Starship provide his terminal setup after he moved away from Warp's bundled features.

Pi Agent uses his documentation and custom skills to review an Arcane update from 2.8.0 to 2.9.0, including pipeline results and release notes. Lempa dictates the request through OpenWhispr, which he describes as open source. He selects on-device speech recognition and explains how an optional LLM polishing step cleans up transcription before insertion. He also notes that it still makes errors and that cloud providers are available.

The next example creates an ntfy Docker Compose proof of concept bound locally on port 8081, with persistent storage and a health check. Colima runs the containers, while Zed handles manual file inspection. Orca then coordinates workers to deploy a self-hosted test environment and prepare documentation and a production migration plan. The reported deployment uses a self-signed TLS certificate and remains unsuitable for production. This is a workflow demonstration, not a local LLM installation guide; it does not establish that every agent runs inference locally.