Today we defined what Servetus actually is. One sentence: an integrated large language model orchestrator that runs as a Docker container beside Nextcloud. You install it, and you have a fully sovereign, LLM-driven assistant that understands your entire life operating system.
What the user gets: a conversational AI assistant accessed through Nextcloud Talk. An orchestrator for all AI tools, local or cloud, routed through one interface. A sysadmin running in the background, monitoring threats and maintaining infrastructure. Full sovereignty. Self-hosted on your own hardware. Nothing leaves without explicit opt-in.
What it is not: not a plugin that adds UI chrome to Nextcloud. Not a black box. Everything is human-readable Markdown, fully auditable. Not a vendor-locked service. You own the data, the config, the rules.
The architecture is four pieces. Docker container beside Nextcloud. Talk bot as the primary interface. Obsidian vault as the transparency layer. LLM orchestration that routes queries to the right model and manages context.
Hermes is the process harness layer. Session lifecycle, tool approvals, crash recovery, skills, cron, webhooks. Servetus provides the orchestration, knowledge, rules, and Nextcloud integration on top. The user never has to know Hermes is there. They talk to Servetus through Talk. Servetus delegates execution to Hermes.
People who use Hermes already understand what this is. Servetus is that concept, integrated into Nextcloud instead of running as a standalone CLI tool. You get what everybody wants without the setup burden.