Updates

Migrated all core bot and user data tables from Supabase to Neon PostgreSQL, complete with primary key sequence synchronization and full foreign key integrity.

Fixed client secret whitespace trimming and interval session verification queries so /anilist connect and list sync complete instantly without errors.

Defers interaction gracefully ensuring the `/kairoku setup` does not timeout for databases experiencing latency above 3 seconds. Allows configuration to save asynchronously safely before following up.

Fixed an issue in the `/kairoku setup` command where the `character_birthdays` option wasn't properly reading as a boolean and was not persisting to the database. The status (Enabled/Disabled) correctly displays on the `/kairoku config` embed.

Users can now invoke /report <details> to preview their bug report embed, view warning prompts regarding server invite link generation, and explicitly send to the developer or cancel via button components.

Updated interaction replies across all commands to use MessageFlags.Ephemeral instead of deprecated options parameters, eliminating deprecation warnings in logs.

Silenced unhandled button warnings for ephemeral command collectors (such as /report confirmation buttons) in the central interaction button router.

Episode notification deliveries will now silently spawn a dedicated Discord thread connected to the announcement if bot permissions allow, organizing discussion without risking spoilers.

All anime-specific slash commands now read GraphQL exceptions directly, surfacing accurate descriptions for timeouts, empty accounts, or exact strings over generic trace catch-alls.

The command pipeline now properly intercepts strict AniList 429 limits and gracefully notifies the Discord user directly instead of swallowing the exception.

Anime results sourced from the Kitsu fallback provider are now blocked from being written to subscription or episode notification databases, preventing AniList ID collisions.

Search queries that are purely numeric (AniList IDs) now always go to AniList directly, preventing the Kitsu fallback from misinterpreting AniList IDs.

Non-numeric, non-positive, and overflow Kitsu IDs are dropped instead of being coerced to 0 or NaN, preventing phantom entries in search results.

Responses missing the required JSON:API data field or containing non-anime records are now rejected as malformed instead of producing empty anime objects.

Season derivation now uses standard month boundaries (Jan-Mar = Winter, etc.) instead of the previous incorrect ranges.

Kitsu subtypes are now case-insensitively matched (e.g. 'movie' maps to MOVIE), preventing unmapped lowercase variants from returning null.

Average ratings are now bounded to (0, 100] and checked for finite values, preventing NaN or negative scores from appearing in embeds.

/user stats and /user recommend now defer their reply before calling remote services, preventing Discord interaction timeouts.

The Favorite and Watchlist buttons defer immediately and handle database failures privately instead of showing Discord timeout errors.

A new GET /ready endpoint probes the PostgreSQL database and Discord gateway with measured latency. Returns 200 when both are healthy, 503 otherwise. The existing /health liveness endpoint is unchanged.

The bot pairs AniList with Kitsu as a fallback data source. When AniList fails recoverably - timeout, network error, malformed response, HTTP 403 block, HTTP 429, or HTTP 5xx - anime commands automatically retry on Kitsu exactly once. If Kitsu also fails, the original AniList error is shown so nothing silently breaks. Kitsu replaced Jikan (MyAnimeList), whose schedule feed was unreliable.

A new 'Kitsu API' card reports the backup provider's health on the status page. When AniList is reporting an outage but Kitsu is healthy, the AniList row notes that the bot is falling back to Kitsu.

The website status page now surfaces AniList 403 failures ('temporarily disabled' and Cloudflare blocks) with their exact error detail instead of a generic online status.

The /anime quiz trivia feature was removed. The quiz subcommand no longer appears in the command tree and its button interactions are gone.

The /anime poll, matchup, rankings, and vote commands replied with an in-development placeholder and were never functional. They are removed so the command list only contains working features.

New /anilist commands let users connect their AniList account: connect, disconnect, sync, and watchlist-status. Linked accounts sync watchlist and favorites in both directions.

Once linking finishes on the website callback page, the bot sends a green confirmation embed to your Discord DMs.

The connect flow moved off the implicit grant. The user authorizes on AniList, the website callback posts the code back, and the bot exchanges it server-side with the client secret before storing the encrypted token.

The post-connect confirmation link no longer drops you into the support server; it opens the Discord client on your own account.

A missing client secret or missing token-encryption key no longer crashes the bot at boot. AniList syncing disables cleanly with an explanatory log line instead.

The /anime quiz trivia feature was removed. The quiz subcommand no longer appears in the command tree and its button interactions are gone, keeping the bot focused on schedules, releases, and server automation.

WhitelistService and the whitelisted_servers table are gone. Commands are registered on every server the bot is in, so no ops DB entry is needed to grant access.

The news service, news scheduler, and /anime news command were removed. Per-guild news channels and the /kairoku setup news option are gone. The kairoku_anime_news table and news_channel_id column are dropped by migration.

A guildCreate handler registers the command tree immediately whenever the bot is added to a new server, instead of waiting for the next startup.

Startup re-registration now covers all cached guilds and clears stale global commands, removing legacy commands (such as /music) from every server.

The site dropped its closed-beta framing: ADD TO DISCORD now points to the bot's OAuth install link, and the status/commands pages reflect the live product.

Java/Spring/JDA engine fully replaced by Node.js 22 + TypeScript + discord.js. All five command roots (/ping, /releases, /kairoku, /user, /anime) behave identically, including autocomplete, selects, and buttons.

Every embed (search, schedule, news, releases, server setup, subscriptions), custom ID prefix, and user-facing message matches the Java output byte-for-byte, verified by automated parity contracts.

The first episode-availability check runs only after the bot is online, eliminating wasted pre-login work.

The AniList response cache is now capped with expiry- and oldest-first eviction to prevent unbounded memory growth over long uptime.

Interaction handlers and promise rejections are caught and logged, so transient Discord/API failures can no longer crash the bot process.

/anime subscriptions and /kairoku subscriptions now resolve anime details concurrently instead of serially, cutting multi-subscription latency.

Multi-stage Dockerfile runs the compiled dist output as an unprivileged user with a health-check endpoint; docker-compose brings up the container with automatic restarts and graceful shutdown.

All database access is parameterized, .env stays gitignored, the bot uses no privileged Discord intents, and npm audit reports zero vulnerabilities.

src/main, src/test, pom.xml, Maven wrapper, and Spring resources were removed from the working tree. The complete Java history remains in git; MIGRATION_PARITY.md records the parity ledger.

AnimeNewsService polls Anime News Network RSS every 10 minutes with failover to Crunchyroll News and MyAnimeList feeds. Articles are deduplicated by GUID and delivered to the configured per-guild news channel.

/anime watchlist and /anime favorites render the user's saved anime from the database as clean numbered embeds.

/kairoku setupserver (Manage Server) scaffolds the ANIME category, channels, role, and permission overwrites idempotently.

/music, /filter, /radio, /history, and /settings were removed from the active bot. The music implementation is preserved at git tag/branch preserve/music-legacy.

kairoku_anime_news now has row-level security enabled (V12), consistent with other bot-internal tables.

Skipped tracks were not recorded in autoplay history, allowing them to be reselected as candidates immediately after being skipped.

Autoplay search queries could collapse to identical strings across successive picks, causing repeated track suggestions.

Tracks with empty or blank titles bypassed the deduplication normaliser and were incorrectly allowed as candidates.

/history only showed manually requested tracks. Autoplay-selected songs were never written to the play history database, leaving the server play log mostly empty. Every track that plays is now recorded (autoplay entries use a system user), and Supabase write failures are now logged instead of silently ignored.

The position/duration progress display (e.g. 00:00 / 04:27) has been removed from the Now Playing embed. Title, artist, requester, and all control buttons remain unchanged.

The /playlist and /lyrics commands have been removed entirely along with all associated infrastructure.

Resolved priority P0 through P3 security and stability findings identified in the latest audit pass.

Replaced manual ID entry in /anime subscribe and /kairoku subscribe with live interactive search. Autocomplete suggests matching anime with title, season year, and community score stars.

Added anime_role options to /kairoku setup, allowing server admins to designate specific roles for anime episode alerts. Current roles are visible in /kairoku config.

Slash command autocomplete for /kairoku setup timezone with support for popular global presets, country/city keywords (e.g., IST, Tokyo, London, NY), and automatic UTC offset calculations.

Added instant autocomplete suggestions for popular radio streams and live radio-browser directory station search under /radio play.

Updated /anime subscriptions and /kairoku subscriptions to dynamically resolve subscribed anime details from AniList, rendering clickable title links, anime IDs, and cover artwork.

The root cause was that autoplay used the exact current track title as the search query, which naturally returned the same track. It now uses a context-aware strategy based on artist and style, and rejects the seed track, queued tracks, and the last 15 played tracks.

A diversity scoring system now penalises consecutive plays from the same artist (streak dampening). After 2 songs from the same artist, autoplay switches to key title token search to discover adjacent artists in the same niche.

Track-end events could previously fire multiple simultaneous autoplay resolution requests, causing unexpected tracks to start after the queue was cleared. An AtomicBoolean CAS guard and session epoch invalidation now prevent this entirely.

The old normaliser stripped all non-ASCII characters to empty strings, making Japanese, Korean, Arabic, and Cyrillic tracks bypass duplicate checks. The normaliser now preserves Unicode letters and numbers across all languages.

Each server now maintains an independent rolling history of the last 15 autoplay tracks. Candidates matching any entry in that history are rejected before scoring. History is never shared between servers.

Removed /setup, /setup support, and /setup content slash commands along with all associated setup and content seeding infrastructure (SupportServerRebuildService, ServerContentSeeder, and 14 related classes).

YouTube is completely removed from the audio resolution chain. No YouTube searches, fallbacks, or direct URL loading exist anywhere in the codebase.

New AudioSourceResolver implements a strict sequential failover: Primary Lavalink (SSL, port 443) → Secondary Lavalink (non-SSL, port 80) → SoundCloud via LavaPlayer. Each tier has a configurable timeout. All providers configured via environment variables in .env — passwords never logged.

Each guild now has a fully independent GuildMusicManager with its own AudioPlayer, TrackScheduler, queue, history, skip votes, and autoplay state. Playing in one server no longer affects or errors in another.

Stream exceptions now trigger a silent internal source fallback. Discord users only see a message if every fallback tier fails. Repeated "Stream failed" embeds are fully suppressed.

A /music play command now creates exactly one loading embed, which is edited in-place to become the Now Playing embed with controls. Subsequent tracks edit the same message — no new messages are created mid-session.

Autoplay now builds contextual search queries from the finished track's artist and title metadata, selecting the highest-relevance candidate from 5 options. Prevents unrelated or disproportionately off-topic track selections.

Lavalink node host, port, and password are loaded from environment variables (LAVALINK_PRIMARY_*, LAVALINK_SECONDARY_*) defined in .env. Passwords are never printed in logs, stack traces, or HTTP headers.

High-performance real-time digital signal processing pipeline supporting speed, pitch, bassboost, nightcore, vaporwave, and karaoke with strict per-guild audio state isolation.

Voice-channel track skipping using ceil(listeners * 0.5) threshold, automatically filtering deafened members and bots with dynamic threshold updates when listeners leave.

Automated server provisioning engine that builds the official 8-tier role hierarchy, 7 lowercase categories, 27 channels, and native Discord AutoMod rules with explicit confirmation.

Automated introspection of registered JDA CommandData runtime definitions to generate categorized embeds for #commands and export synchronized web metadata.

Interactive player UI buttons (pause, skip, loop, shuffle, stop) now route through centralized permission, cooldown, and voice validation pipelines with DJ and moderator overrides.

Enforces granular per-user cooldowns (10s filter, 3s play, 1.5s volume) to prevent Discord API spam and resource exhaustion while allowing trusted DJ overrides.

Eliminates race conditions during concurrent voice voting using thread-safe ConcurrentHashMap key sets and volatile state visibility across JDA event dispatchers.

Implemented comprehensive /anime search, trending, popular, and seasonal discovery powered by AniList GraphQL API v2.

Added /anime schedule, /anime today, and /anime next commands with localized Discord relative timestamp formats.

Added community-driven /anime matchup features with ephemeral button interactions and scoring.

Integrated Redis caching for frequent AniList GraphQL queries and seasonal data, drastically reducing latency and external API consumption.

Fixed HTML entity decoding and markdown escaping bugs in long anime descriptions and character biographies.

Added /radio command supporting real-time search and streaming across international radio station directories.

Integrated Supabase PostgreSQL database persistence for per-user custom playlists with /playlist save, load, and list.

Implemented automatic voice channel reconnection and playback resumption after Discord gateway disconnections.

Resolved state desynchronization issues in PlayerManager when members force-disconnect the bot during active playback.

Added /lyrics command fetching synchronized LRC lyrics with real-time highlighted playback position.

Introduced fine-tuned equalizer profiles for bass boost, vocal clarity, electronic, and rock acoustic balancing.

Added /settings dj-role and /settings 247 configuration subcommands stored in local guild configuration repositories.

Fixed millisecond offset calculations for variable bitrate audio tracks in LRC lyrics parsers.

Rebuilt backend architecture leveraging Java 21 virtual threads, Spring Boot 4, and JDA 6 for high-concurrency event processing.

Integrated Lavalink v4 audio server supporting Discord Dave end-to-end voice encryption protocols.

Launched official neo-brutalist website platform with real-time status telemetry and web changelog.

Deprecated legacy prefix-based command handling in favor of native Discord Slash Commands with rich auto-completion.