aim-ops

Server lifecycle for ai-memory (engine by akitaonrails/ai-memory, packaged and extended by an ops repo that builds the images and holds the runbooks). This skill is the reproducible procedure; the ops repo holds the real compose, images and instance-specific runbooks. Never write secrets, digests or hostnames of a real instance into this skill — they belong in the operator's private runbooks/memory.

Client-side wiring (marker, MCP entry, hooks, local CLI) → aim-init.

Route intent

• "deploy / self-host / subir uma instância nova", "new environment" (PT: "instala o ai-memory num servidor", "sobe uma instância") → install.
• "upgrade / update the server", "saiu versão nova, qual o impacto?" (PT: "atualiza o engine", "faz upgrade") → upgrade (references/upgrade-playbook.md).
• "change the LLM / embeddings", "too expensive", "run a local model?" (PT: "troca o modelo", "desliga o Cloudflare") → providers (references/providers.md).
• "issue/revoke an API key for X", "hooks 401/403" → keys.
• "backup / restore / rollback" → backup-restore.
• "is it healthy?", "what's running?", "why is capture failing?" (PT: "como está o servidor?") → doctor (read-only).