This devlog is about getting Logos AI from “works on my machine” to “a real person can download this and try it without me sitting beside them.”

The current push has four practical goals:

• keep the bundled SQLite library and shipped content central instead of treating the desktop app like a thin shell
• make Windows, macOS, and Linux downloads feel direct instead of making people hunt through release pages
• prepare local voice and AI dependencies on first launch when they are not already installed
• tighten the release filenames so the website can link to stable installer URLs

That matters because Logos AI already ships with a real local payload: SQLite, imported Bible content, commentary, and prebaked read-aloud data. The app is not starting from empty every time someone installs it. The setup work should only handle what is too large, too platform-specific, or too changeable to sensibly embed.

The new local setup flow is aimed at exactly that line:

• keep shipped Bible and commentary data in the app experience
• create a private Python environment for Kokoro and Piper when needed
• download Kokoro voice assets when they are missing
• install and start Ollama when possible
• pull a smaller local chat model plus an embeddings model for future local search work

That does not magically solve every packaging problem, but it does move the product closer to a real beta surface instead of a developer-only setup.

The public website is also being tightened so people can click straight into:

• a Windows installer
• a macOS installer
• a Linux .deb or AppImage

That is the difference between a project that sounds promising and a product that can actually get tested across different machines.

The immediate next step is not mass launch. It is a small beta: a handful of real users on different laptops and desktops, different Windows setups, and different comfort levels with local AI. That is where the weak points will show up fastest.