# ARCHITECTURE > **Placeholder.** Full architecture doc lands from a parallel research > agent in a subsequent commit. Until then, the closest existing > reference is the v1 design note: > > > > That document was written *during v1's lifecycle* as the design > sketch for what would become v2. Treat it as the starting point — > not the final architecture. ## Planned scope of this document (when written) - **State machine.** Every login state, every transition, every timeout. Diagram + prose. - **Threading model.** Which work runs on the main thread, which on async pools, where the synchronisation boundaries are. - **Persistence.** What state is stored, where, in what format. Restart-safety semantics. - **AuthMe integration.** How v2 reads `authme.db`, how it observes AuthMe's lifecycle events, how it decouples from AuthMe quirks. - **Failure modes.** Backend-down, chunk-not-loaded, player-disconnect- mid-flow, server-restart-mid-flow. One paragraph per mode. - **Observability.** Logging schema, metrics surface (if any — opt-in only), how an operator debugs a stuck player. - **Security model.** Threat model, trust boundaries, what a hostile client can and cannot do. Until the agent lands the real doc, design discussion happens in PRs and the v1 reference above.