AutoStopper

Added

- Failed /autostopper start, stop, and restart commands now state the cause in chat. In game, hovering shows the remediation and the typed outcome code (matching outcome= in velocity.log), clicking opens the matching troubleshooting section, and a suggested next command (the retry or /autostopper status) follows the line. Console senders get the same cause, remediation, outcome, and link as plain text. Player connection messages are unchanged (#124). - /autostopper help for operators, and /autostopper status when any server has failed, end with Troubleshooting · Get help links. Documentation links are pinned to the running release (#124). - docs/troubleshooting.md documents every manual command failure outcome (#124). - docs/troubleshooting.md has an outcome reference covering every outcome= code in lifecycle log lines, including player connection outcomes, NO_MAPPING, PROXY_SHUTDOWN, and success codes (#123). - /autostopper status <server> shows one server in detail: state and, while waking, the startup stage, hold, waiting players, time since activity, and the last failure with its age, cause, and fix. The console prints the exact failure time; players hover the age for it. It inspects only that server's container and tab-completes mapped server names. In game, clicking a server name in /autostopper status opens its detail view (#125). - /autostopper status <server> lists the server's last 10 completed lifecycle operations, newest first: startups, manual starts, stops, and restarts, and automatic stops, each with its age, who started it, its typed outcome (matching outcome= in velocity.log), and its duration. The history is in memory only; a proxy restart clears it, and a reload clears it for servers whose mapping was removed, added, or changed (#129).

Changed

- /autostopper status shows the last failure's age (last failure 3m ago during …) instead of a raw ISO-8601 timestamp, and leaves the remediation to /autostopper status <server>. Activity and failure ages share one format, and ages of a day or more read 2d 3h ago (#125). - Velocity 4.1.0 (build 22) and 4.2.0 (build 30) are now tested stable runtimes alongside 4.0.0, with 4.2.0 the recommended production line and the stable example. The preview line now validates Velocity 4.2.1-SNAPSHOT (build 41). The packaged-runtime CI matrix, smoke profiles (stable-4.0, stable-4.1, stable, preview), release-candidate E2E, and release metadata were updated to match. Pull requests run only legacy, stable, and preview; the release gate also runs stable-4.0 and stable-4.1 (#134). - /autostopper now shows clickable Docs, GitHub, Modrinth, and Changelog links to everyone, and the version opens the release notes; console senders get the URLs as plain text. Documentation and changelog links are pinned to the running release (#131). - /autostopper help groups commands under General, Lifecycle, and Holds, uses consistent imperative descriptions (also used by usage messages), drops the /autostopper and /autostopper help entries, and shows each command's permission on hover. Sources without operator commands get the project links instead of an empty list (#131). - A cancelled manual stop now reports Stopping server <name> was cancelled. instead of the start wording (#124). - Readiness checks no longer hold an AutoStopper worker for the whole readiness window. Each attempt runs as a short, individually bounded task, and the probe interval between attempts waits on a timer instead of a sleeping worker. Concurrent startups no longer delay other servers' status checks, startups, stops, the inactivity scan, or /autostopper status. Readiness deadlines, probe intervals, outcomes, cancellation, and shutdown are unchanged. If AutoStopper is saturated before the first attempt, the startup still reports OVERLOADED; a later attempt skipped for saturation is retried at the next interval within the same deadline (#95). - The canonical build is the checksummed Maven 3.10.0 Wrapper. Maven's auto-discovered repository prefix filter is disabled in .mvn/maven.config because the PaperMC repository can publish a prefix file that omits com.velocitypowered, which blocks Velocity API resolution.

Fixed

- A startup opened by /autostopper start or /autostopper restart now logs origin=MANUAL_COMMAND in its op=STARTUP lifecycle line instead of origin=PLAYER_CONNECTION (#129). - A startup whose next step throws unexpectedly now fails with the stage's error outcome, so the server becomes FAILED instead of staying in STARTING until the proxy restarts. A manual restart in that case reports START_FAILED instead of never completing (#118). - A manual stop that loses its server to another operation now ends CANCELLED and leaves that operation's state alone, instead of stopping the container or returning the server to READY (#119).

Fixed

- Servers whose container stops outside AutoStopper (backend crash, in-game /stop, manual docker stop) no longer stay READY after a prior successful connection. The inactivity scan and refused connection attempts now reconcile the stopped container through the revision-guarded lifecycle path, so the next player connection wakes the server again (#93). - Stops now honor the container's own stop grace period (StopTimeout, Compose stop_grace_period) instead of cutting docker stop off at a fixed 10-second deadline. The deadline is the grace period plus the Docker command timeout, with Docker's 10-second default when unset and a 10-minute cap. A stop whose CLI still times out is re-inspected and reported as stopped when the container has already exited. The stop command remains docker stop <container> (#94). - The examples give managed backends stop_grace_period: 60s so large worlds can finish saving.

Added

- Added coordinator-owned operator lifecycle commands for mapped servers: /autostopper start <server>, /autostopper stop <server>, /autostopper restart <server>, /autostopper hold <server>, and /autostopper release <server>. - Added command-specific permission nodes (autostopper.command.start, autostopper.command.stop, autostopper.command.restart, autostopper.command.hold, autostopper.command.release) while retaining autostopper.admin as an explicit umbrella. - Implemented runtime shutdown holds that suppress automatic inactivity shutdown without blocking manual commands or player-driven wake-up. Holds survive unchanged reloads and clear on proxy restart, mapping removal, or mapping replacement. - Manual stop and restart enforce strict safety invariants, refusing execution if connected players or lifecycle waiters are present, with an immediate deterministic pre-Docker safety re-check on the worker thread. - Manual restart runs as one atomic coordinator-owned stop → start → readiness sequence under a single authoritative future. - Surfaced hold state in /autostopper status and updated /autostopper help with permission-filtered entries and argument tab completion. - Pinned stable Velocity 4.0.0 build 6 as the production support line on Java 25, tested by the packaged-runtime system tests and the release-candidate Docker/Minecraft gate. Velocity 3.5.1 on Java 21 remains the tested minimum/floor, and Velocity 4.1.0-SNAPSHOT on Java 25 remains clearly labelled preview validation. - Added the RUNNING_UNVERIFIED operational state so a Docker-running container is not reported as ready until AutoStopper has completed the configured readiness contract for the current mapping lifecycle. - Connecting players now receive authoritative inspection, startup, readiness, connection, and terminal lifecycle messages. Late arrivals see the current shared stage and unique waiter count. - Redesigned command and chat presentation around the brand identity ("Empty servers sleep. Players wake them."). Replaced legacy bracketed prefixes with clean brand prompts (AutoStopper ›, AutoStopper ✓, AutoStopper !), added scannable deterministic status rows with humanized states and durations, and introduced permission-filtered help with fuzzy command matching. - Added contributor-facing message style guide (docs/message-style-guide.md). - Added structured, typed lifecycle telemetry and bounded process-lifetime observability (#74). Authoritative lifecycle operations emit machine-parsable INFO log lines (op=, server=, origin=, outcome=, elapsed_ms=, waiters=) with monotonic duration measurement, and intermediate stage durations emit DEBUG logs. Simultaneous waiters receive 1 shared startup record plus individual connection wait records. Aggregates use thread-safe bounded memory with strictly zero PII and observational exception insulation.

Changed

- Operational status now reconciles Docker observations lazily with mapping-aware lifecycle revisions. Active transitions remain authoritative, stopped/degraded Docker state overrides stale quiescent state, and results captured before reload, mapping replacement, or newer lifecycle work are discarded without eager startup readiness probing.

Fixed

- Restored command and lifecycle messages on Velocity 3.5.1 by avoiding the Adventure 5 component-builder ABI incompatibility and validating every message factory against the pinned Velocity runtimes.

Added

- Tested support lines for Velocity 3.5.1 build 615 on Java 21 and Velocity 4.1.0-SNAPSHOT build 16 on Java 25, using one Java 21 bytecode plugin JAR. - Explicit, isolated server_name to container_name mappings with startup/reload preflight. - Real readiness strategies: Minecraft status protocol, Docker health, or either signal, all with bounded per-server deadlines. - Shared lifecycle coordination so simultaneous players use one startup/readiness operation and stop/start/reload races have legal, deterministic outcomes. - Bounded off-thread Docker execution, bounded plugin shutdown, cancellation, and saturated-worker handling without blocking Velocity event or command workers. - Capped exponential retry for failed automatic stops while preserving activity until a confirmed stop. - Atomic configuration reload: an invalid candidate is rejected completely and the previous immutable snapshot remains active. - Permission nodes autostopper.command.status and autostopper.command.reload, plus the explicit autostopper.admin umbrella. - Typed operational state and sanitized /autostopper status diagnostics with operator actions; raw Docker stderr remains in operator logs. - Pinned legacy/current Compose examples, packaged Velocity runtime tests, a real Docker/Compose/Minecraft release-candidate gate, reproducible Maven Wrapper builds, strict dependency/package checks, SpotBugs, JaCoCo, and deterministic CI gates. - Installation, security, configuration, troubleshooting, Modrinth, and 1.1.2 migration documentation.

Changed

- Generated inactivity timeout and public documentation now agree on 300 seconds. - New installations generate no monitored mappings instead of controlling example containers by default. - Velocity's actual lowercase data path is documented as plugins/autostopper/config.yml. - A running Docker container must pass configured readiness before waiting players are connected. - Inactivity scans retain activity on Docker failures and stop only explicitly mapped, running, player-free servers. - Managed containers are expected to exist in advance and use restart: "no"; unmonitored hubs remain outside AutoStopper's lifecycle. - Docker socket/group access is explicitly documented as a privileged, host-root-equivalent trust boundary rather than ordinary non-root isolation. - The canonical build is the checksummed Maven 3.9.16 Wrapper on JDK 21 through 25.

Removed

- The prototype's custom /server command/interceptor. Velocity owns /server and its permissions; AutoStopper now observes normal pre-connect events. - Unbounded fixed-delay startup behavior, blocking Docker work on platform workers, and partial or silent configuration fallback.

Migration

Read the complete [1.1.2 to 2.0.0 migration guide](https://github.com/Criseda/AutoStopper/blob/2.0.0/docs/mi… before replacing the artifact. The runtime matrix, lowercase data path, permissions, readiness policy, explicit mappings, pinned examples, and Docker security boundary all require operator review.

1.1.2: - Fixed permission denied issues from docker - Fixed autocomplete in chat - Fixed error dump when trying to enter a portal that has a server that is booting up, you get a message instead

Add functionality to work with external plugins that call velocity api directly to join different servers.