What's new.

Reports work with no setup. The Report Message and Report User context menus file them, and the person who reported is told the outcome by DM.

Safety signals, opt-in and off by default. They flag coded language to staff, never punish anyone, and store signal keys instead of message text.

Deciding a report now needs permission. Manage Server always counts, and /report setup staff_role lets a server name one role whose holders can decide. Before this, any member could close a report.

Like and Dislike on 18+ results and on autoposts. The first press asks for a one-time opt-in. Only tags and post ids are kept, never what you search for, and anyone who can see an autopost can press its buttons.

Edit your own taste. In /nsfw taste, pick a tag and set how you feel about it, pin it so it stops fading, mute it so it is never used, add a tag by hand, or delete everything.

/nsfw foryou serves a feed ranked from the tags you like. It needs a handful of likes first, mixes in a few new ideas, skips muted tags, and never repeats a post you have already rated.

A Taste memory page on the dashboard for anyone who has opted in. After you sign in you can see your tags, change a weight, pin or mute a tag, or delete everything. It shows tag names only.

A suspicious account filter in automod, off by default. It tells your staff when several accounts join within minutes with a similar name or the same profile picture, or when Discord has marked an account as spam. It never bans, kicks or mutes anyone.

Verification now sends a code with the link. A leaked link on its own can no longer change a verification attempt.

The account age gate now bans an account whose name advertises child abuse material, instead of only kicking it.

Role grants that hand out permissions the person granting them lacks are now refused. Autorole, leveling rewards, birthday roles, the verification gate, /role add, massrole and the AI's add role action all check this.

Level-up, starboard, bump reminder, birthday, audit log, autopost and mod log output now refuses a channel that belongs to another server, even if a stored setting points there.

The privacy page now covers taste editing, the Taste memory page and the account filter, including that the filter's optional AI note sends a joining account's username and display name to TypeSafe only when a server turns it on.

Honeypot. Spam and scam bots often post the same message in every channel they can see. The honeypot gives them a channel to fall into, and Akiko deals with them quickly.

Trap channels. A trap looks like a normal public channel. Akiko can create one that blends in, or adopt a channel you already have. Real features, announcement channels and webhook channels can never become traps.

Hidden from the rest of Akiko. Trap channels are left out of the dashboard pickers, role panels and the channel list the AI sees.

Role panels. Post one message with buttons or a select menu and members give themselves roles. Set them up with /autorole panel, and edit a posted panel without reposting it.

A color picker for role panels. One button opens the full color palette sorted by hue, with a search box, and a color role is created the first time someone picks it.

Premium is now per server. One subscription upgrades one server, and everyone in it gets unlimited AI messages, a custom AI prompt, and bigger role panels. It used to follow a single person.

No XP channels and roles can now be edited from the dashboard, not only with a command.

A new premium purchase could show a renewal date in 1970 and expire straight away. The date now comes from the subscription itself.

Using the autorole as the birthday role removed it from members the day after their birthday. That choice is now refused with an explanation.

Rebuilt the premium page around what a subscription actually buys. It previously showed a single command where five features should have been.

Premium now has a place in the site navigation, which is grouped into Explore, Community, and About rather than nine flat links.

The homepage served zeroes for server, member, and command counts to anything that did not run JavaScript, search engines included. It now serves the real numbers.

Cut roughly a third off how long the homepage takes to paint on a phone. A font finishing its download was forcing the browser to lay out the entire page again.

The command theatre went from 9 scenes to 59 across 17 categories, so you can watch what a command does before adding me to anything.

Added /rank to the showcase. It was the one command most people try first and the only flagship the theatre was missing.

The settings pages now play a short clip of each module on hover, rather than describing it and hoping.

Economy commands now accept the comma-grouped amounts the bot itself prints, so copying a number back in works.

Fixed /data export and wipe, which were failing on some servers.

Auto-roles only ever ran through the verification gate, so they did nothing in almost every server that had them configured. They now apply on join, and there is a backfill for members who were missed.

Eleven of the twenty-four raid heat settings had nothing reading them. All of them work now. Enforcement stays off until a server turns it on, because nobody had been running with these values in effect.

The link filter and mention spam detectors now honour the infraction limits the dashboard had been offering for them.

Mutes can run longer than Discord's own 28-day ceiling.

Cleaned out my memory. Around 1,400 of 1,600 stored memories were noise from a bad filter, and the ones that remain are things somebody actually said.

Partner submissions are verified before they post, and approvals announce with a working invite rather than a bare link.

The dashboard says what each setting does, including the ones that do nothing yet. Automod and the Safety sidebar are ordered by what you configure first.

The dashboard has a console built for phones: its own shell, a server rail, and an edge-swipe drawer, instead of a desktop layout squeezed sideways.

The privacy policy lives here now, at /policy, rather than in a gist.

The partners page and the community art gallery read from approved submissions instead of a hardcoded list.

Server icons keep their transparency instead of being flattened onto white.

Closed a hole that let user-submitted links carry javascript: URLs, and added anti-clone guards to the verification board.

A failed build can no longer take the site down with it.

We have implemented our own custom "fleet management system" for Akiko and her sharding/deployment/fleet-wide IPC communication. This will speed up deployments and bot reboots by almost 3x as well as helping maintain an even more stable uptime.

Added an "auto-heal" feature to the Akiko fleet, so that way if something goes wrong on a cluster, it will attempt to respawn itself! Or if something is wrong with the gateway, we'll know what to do.

Refreshed the entire UI/UX for the panel for better navigability and readability.

We also swapped over to Nextjs 16!

Welcomer local image rendering issue.

Error handling in image processing improved.

AI context and workflow enhancements.

AI model configuration and workflow improvements

AI web search fallbacks and response rejection checks

Polished frontend interactions and welcomer card functionality

Improved AI model configuration and workflow.

Enhanced AI debug tools and response rejection checks.

Fixed AI web search fallbacks and response quieting.

Added AI worker embed circuit breaker and low value message quieting.

Welcomer cards and quick setup dashboard flow

Leveling card image fetching moved off loop for performance

Leveling card background previews, dashboard quick setup actions, and automod dashboard fallback dispatch

Cluster safe quick setup and warning escalation coverage

Quick setup dashboard flow and welcomer card config API

Leveling card image fetches moved off loop for performance improvement

Verification risk checks improved

Heat decay fixes and normalization

Automod dashboard fallback dispatch and leveling card background previews

Images not loading in custom level cards

Claiming expired vote rewards

Userinfo partial member roles

AI crisis and repetition handling

Verification client and bot side issues

Automod heat system dashboard and API support

Leveling system updates

AI memory tweaks

Dashboard cache scoping and verification

AI memory tweaks and improvements.

Panel API authentication and security updates.

Improved error handling in API fetch and caching mechanism for dashboard data

Updated Akiko branding with full-body artwork

Better vote call to action

Feedback ratelimiting and forum channel support

Improved vote cta for level rewards

Updated welcomer hints

Refreshed Akiko branding with full-body artwork

Fixed images not loading in custom level cards

Fixed claiming expired vote rewards

Fixed userinfo partial member roles

Fixed ai crisis and repetition handling

Fixed verification client and bot side issues

Fixed various frontend and backend issues

Improved error handling in API fetch

Harden critical security paths

Improved vote call-to-action

Panel API authentication

Mock API for testing and web redesign

Save All button to dashboard

Promotion detection option to scam detection settings

Bot invite overlay and validation for bot presence in guild

Improved error handling in AI and AuditLog Cogs

Enhanced modal dragging functionality with position clamping

Missing GET route for /api/guild/:guildId/general endpoint

Auditlog event handling

Automod persistence issues

Channel and role input handling in special fields

VerificationClient API key loading

Guaranteed wins item and toast notification

Fixed dealers marked cards use-item to set guaranteed wins

Prisma DB pulling and generating issue

Economy gathering items variety increased

Fixes for slots SQL syntax error

Fix for item use endpoint race condition

Fix for dealers marked cards use-item functionality

Fix for coin gamble balance update

Reply context to AI feature

Improved economy gathering items variety

Updated vote button to show thanks for voting if voted recently

Fixed item removal query SQL syntax error

Fixed dealers marked cards use-item to set guaranteed_wins

Fixed coin_gamble balance update by converting to bigint

Dealers marked cards use-item to set guaranteed wins

Reply context to AI

Economy gathering items have more variety

Dealers marked cards use-item to set guaranteed_wins

Improved economy effects and login redirect

Fixed guaranteed wins item removal and toast notification

Coin gamble balance update to use bigint

Vote button to show thanks for voting if voted recently

Implemented economy effects and login redirect improvements

Updated coin_gamble balance update to use bigint

SQL syntax error in slots

Economy gathering items to have more variety

Ratelimits and cross-cluster interference

Added more variety to economy gathering items

Added negative channel caching

Reply context to AI functionality

Coin gamble balance update now uses bigint

Ratelimit lockdown issue

Fixes for dealers marked cards

Fix for slots SQL syntax error

Fix for item removal query SQL syntax error

Fix for race condition in item use endpoint

Fix for guaranteed wins item removal and toast notification

Fixed SQL syntax errors in item removal query and slots

Item use endpoint race condition

Coin gamble balance update

Ratelimits and interference issues

Guaranteed wins item removal and toast notification issue

Ratelimits and lockdown

Improved economy gathering items with more variety

Updated economy gathering items with more variety

Coin gamble balance update issue

Dealers marked cards use-item issue

Vote button glow animation

Guaranteed wins item removal and toast notification

Rule34.xxx API fallback to rule34 command

Vote button shows thanks for voting if voted recently

Dealers marked cards stack on use

Economy effects and login redirect improvements implemented

Race condition in item use endpoint fixed

SQL syntax error in item removal query fixed

Slots SQL syntax error fixed

Vote button now shows thanks for voting if voted recently

Economy gathering items now have more variety

Race condition in item use endpoint

SQL syntax error in item removal query

Slots SQL syntax error

Account bio model

Dealers marked cards now stack on use

Fixed race condition in item use endpoint

Fixed SQL syntax error in item removal query

Fixed slots SQL syntax error

Fixed lunar API 401 retry logic

Economy effects and login redirect improvements

Guaranteed wins item with removal and toast notification

SQL syntax error in slots query fixed

Negative channel caching added

Ratelimits and lockdown implemented

Dashboard usage and configuration updated

Guaranteed wins item removal and toast notification implemented

Rule34.xxx API fallback to rule34 command added

Dealers marked cards stack on use changed

Economy effects and login redirect improved

Music functionality improved

Guaranteed wins item with toast notification

Dealers marked cards use-item now sets guaranteed wins

Guaranteed wins item functionality

Slots SQL syntax error resolved

Lunar API 401 retry logic issue fixed

Account bio model issues resolved

Guaranteed wins item removal implemented

Allowed all clusters to post stats to database for panel aggregation

Guaranteed wins column to user record table

Cluster checks added to prevent cross-cluster interference

SQL syntax error in slots fixed

Negative channel caching issue fixed

More variety to economy gathering items

Negative channel caching and rule34.xxx API fallback

Economy effects and login redirect improvements were implemented

Guaranteed wins column was added to user records table

More variety was added to economy gathering items

Rule34.xxx API fallback was added to rule34 command

Allow all clusters to post stats to database for panel aggregation

Fixed account bio model

Ban tracking on the panel, so a ban is visible somewhere other than the audit log.

Ratelimit lockdown, which cuts off a source that is hammering the API rather than letting it degrade everyone.

Dashboard usage is tracked, so it is possible to tell which settings pages anyone actually opens.

Transient Discord errors are retried instead of surfacing as a failed command.

Fixed the retry path for Lunar API 401s, which had been giving up too early.

Fixed the ban handoff between clusters, which could drop a ban partway through.

Fixed the account bio model, which was rejecting valid bios.

Effects stack now, rather than the newest one replacing whatever was running.

Clusters check for each other before acting, so two of them no longer answer the same event.

Cut logging verbosity down, and added request tracing to the parts that needed it.

Colour and visual updates across the panel.

Removed the Lunar API token refresh logic, which the new retry path made redundant.

Transient Discord errors have been fixed

Account bio model has been fixed

Ban handoff issues have been resolved

Possible XSS vulnerabilities have been fixed

Security issues have been fixed

Economy commands have been improved to resolve conflicts

Effects now stack

Color updates have been made

Ban tracking has been added for panel

Welcomer now gets populated

Account bio model issues have been resolved

Possible XSS vulnerabilities have been patched

Economy command conflicts have been resolved

Tool item types have been cleaned up

Effects now stack properly

Security issues have been addressed

Detailed logging has been added to ban tracking

Ban tracking has been implemented for panel

Color updates have been applied

Datetime timezone errors in handoff monitor have been fixed

Resolved command conflicts in economy system

Fixed tool item types in economy

Removed duplicate cogs in economy

Fixed economy hot reload issue

Fixed economy gather command to use weighted loot table selection

Added command to cleanup duplicate ban records

Fixed ban crawl handoff recovery and takeover from stale clusters

Fixed possible XSS vulnerability

Improved ban tracking and handoff monitoring

Fixed economy hot reload by checking if extension already loaded

Fixed ban crawl handoff recovery by increasing heartbeat timeout to 5 minutes

Fixed datetime timezone error in handoff monitor

Improved automod event listener registration

Converted automod group to text-only to reduce slash command size

Fixed security issues

Added ban tracking for panel

Updated color scheme and made sure welcomer gets populated

Effects stacking feature

Ban tracking for panel

Economy system improved with weighted loot table selection

Automod command group converted to text-only for size reduction

Security issues fixed

DateTime timezone error in handoff monitor fixed

Economy hot reload issue fixed

Economy system was improved for better maintainability

Automod event listener registration was improved

Command tree creation error was fixed

Tag and report_setup command groups were converted to text-only

Datetime timezone error in handoff monitor was fixed

Detailed logging was added to ban tracking register function

Webp avatars were converted to png for welcomer preview api compatibility

Security issues were fixed

Possible XSS vulnerability was fixed

Ban crawl handoff recovery was fixed

Economy hot reload was fixed

Datetime timezone error was fixed

Economy gather command now uses weighted loot table selection

Command to cleanup duplicate ban records was added

Automod command group was split into top-level commands

Automod event listener registration was fixed

Webp avatars are now converted to png for welcomer preview api compatibility

Command to cleanup duplicate ban records

Detailed logging to ban tracking register function

Log message to confirm handoff monitor starts on cog load

Economy gather command to use weighted loot table selection

Automod group to text-only to reduce slash command size

Tag and report_setup command groups to text-only

Eco command group to multiple top-level commands

Datetime timezone error in handoff monitor

Cleanup duplicates to group by user_id+guild_id only

Ban crawl handoff - allow takeover from stale clusters

Economy hot reload by checking if extension already loaded

Welcomer preview retry logic and CORS issues

Eco command for item descriptions

Economy system refactored into single /eco group command structure

Automod group converted to text-only to reduce slash command size

Tag and report_setup command groups converted to text-only

Fun.py entertainment commands converted to text-only

Command tree creation error fixed by calculating command sizes

Welcomer preview retry logic fixed

Lockdown schema updated to make locked_by/locked_at optional

Economy system refactored into a single /eco group command structure to reduce slash command count.

Automod group converted to text-only to reduce slash command size below 8000 limit.

Tag and report_setup command groups converted to text-only to reduce slash command size.

Fun.py entertainment commands converted to text-only to reduce slash command count.

AcknowledgmentError added to utils.errors for shared use.

Profile command renamed to user_profile to avoid conflict with Economy profile.

Item descriptions added to eco command.

Possible XSS vulnerability fixed in panel.

Ban tracking feature added to panel.

Welcomer preview API compatibility improved by converting webp avatars to png.

Retry logic added for welcomer preview with fallback URLs.

Premium upgrade UI added for free users on account page.

Lockdown members_role configuration added to API endpoint, UI selector, and save function.

Warning escalation system for automod implemented

Detailed configuration display for automod settings

Automod commands now use Discord UI components instead of embeds

Possible XSS vulnerability fixed

Lockdown endpoint errors handled properly

Warning escalation system for automod

Phishing detection into link filter

Automod commands to use discord UI components

Automod settings command to show detailed configuration

Music API to new system

Possible XSS vulnerability

Ban tracking issues

Welcomer preview issues

Unused starboard channel selector initialization

Various fixes for welcomer preview API compatibility

Error handling for link filter and phishing detection

Automod settings command now shows detailed configuration

Various fixes for welcomer preview and retry logic

Lockdown schema and endpoint fixes

Error handling fix for a command

Fixed bump reminder selectors and re-initialized when tab visible

Fixed vote reminders and triple DMs

Improved welcomer preview retry logic with fallback URLs

Fixed PATCH lockdown endpoint to handle null members role properly

Premium upgrade UI for free users on account page

Improved welcomer preview API compatibility

Updated music API and migrated old system

Fixed bump reminder selectors and triple DMs

Fixed PATCH lockdown endpoint and added error logging

Fixed vote reminders and added cluster_id checks

Removed EthicalAds integration due to application denial

Removed unused starboard channel selector initialization

Improved welcomer preview with retry logic and fallback URLs

Fixed vote reminders and stripe webhooks

Vote reminders with customizable command mention

Music API updated and old system migrated

Welcomer preview API compatibility improved

Bump reminders no longer remove themselves

Triple DMs issue resolved with cluster_id checks

Error handling improved for commands and link filter

EthicalAds integration removed due to application denial

Unused starboard channel selector initialization removed

Link filter now integrates phishing detection

Welcomer preview issues resolved

Vote reminders now work correctly

Phishing detection service with blocked counter

CORS issues with welcomer preview resolved

Welcomer preview retry logic with fallback URLs

Link filter to include phishing detection

CORS-problematic HEAD fetch in welcomer preview

Bump reminder selectors

Triple DMs issue

PATCH lockdown endpoint to handle null members_role

EthicalAds integration due to application denial

Lockdown members_role configuration and API endpoint

Welcomer preview image now fetches without CORS issues

Lockdown schema updated to make locked_by and locked_at optional

PATCH lockdown endpoint now handles null members_role properly

Improved automod verification using config endpoint

Fixed welcomer preview by removing CORS-problematic HEAD fetch

Fixed join-raid routes to prevent 400 error

Removed debug logging from saveJoinRaid and apiFetch

Enhanced selector initialization for automod modules

Welcomer preview image CORS issue resolved

Bump reminder selectors re-initialized when tab is visible

Improved selector initialization for automod modules

Fixed lockdown schema to make locked_by and locked_at optional

Fixed PATCH lockdown endpoint to handle null members_role properly

PATCH lockdown endpoint now handles null members role properly

Join-raid routes moved to prevent 400 error

Error handling improved for missing GET endpoints

Phishing detection integrated into link filter

Lockdown members role configuration with API endpoint and UI selector

EthicalAds integration with non-intrusive footer for non-premium users

Automatic verification to save functions for database updates

Lockdown schema updated to make locked_by and locked_at fields optional

PATCH lockdown endpoint updated to handle null members role properly

Fixed bump reminder selectors to re-initialize when tab is visible

Fixed 404 errors in verifySave and GET endpoints

Fixed join-raid route to prevent 400 errors

Detailed error logging to PATCH lockdown endpoint

Lockdown members_role configuration

Non-intrusive EthicalAds footer for non-premium users

Polite fallback message for ad blocker users

Phishing link detection service with blocked counter

PATCH lockdown endpoint to handle null members_role properly

Bump reminders removing themselves

Triple dms issue with cluster_id checks

Improved automod selector initialization and re-initialization

Updated save functions to use safeParseInt for consistent number parsing

EthicalAds integration with non-intrusive footer

Phishing link detection service

Improved dashboard to load channels/roles even if config fails

Bump reminder selectors to re-initialize when tab visible

SaveJoinRaid to handle empty/invalid values with proper defaults

Error handling for missing GET endpoints

Lockdown members role configuration

Improved dashboard loading with channel and role data

Fixed triple DMs issue

Fixed 404 error handling for missing GET endpoints

Automod verification now uses config endpoint

Fixed triple DMs with cluster ID checks

Fixed 404 errors before JSON parsing

Fixed saveJoinRaid to handle empty or invalid values

Phishing link detection, backed by the Sinking Yachts API and wired into the link filter with its own blocked counter.

Vote reminders, with a message you can write yourself and a thank you sent back to whoever voted.

Automatic verification, so the gate no longer needs a manual step to save.

Per-event audit log channels, rather than one channel for everything.

A preview for the welcomer, so you can see the message before a new member does.

A modal channel picker for the link filter whitelist and blacklist.

Fixed triple DMs. Vote reminders and Stripe webhooks were running on every cluster instead of one.

Fixed bump reminders deleting themselves.

Closed a set of cross-site scripting holes in the panel.

Fixed a 400 error when saving join-raid settings, and gave empty values sensible defaults.

Fixed 404s on several endpoints that were never registered.

Fixed autopost settings not persisting, and the dropdowns that had stopped opening.

Fixed disabling the AI, which did not take effect.

Fixed archiving and deleting memories, and AI memory not updating.

Fixed automod selectors failing to initialise by letting them re-initialise.

The dashboard loads channels and roles even when a config fetch fails, instead of showing an empty page.

Link filter whitelist and blacklist channels are multi-select dropdowns now.

Reordered the dashboard modules, moved warning escalation to the top of the automod page, and moved mod settings under Moderation.

A design pass over the dashboard, changelog, premium and about pages.

The AI system prompt hands out the dashboard URL when someone asks where a setting lives.

Modal-based channel selector for link filter whitelist/blacklist channels

Audit log per-event channels and automod hints

Welcomer preview and autopost settings persistence

Moved mod settings from FEATURES to MODERATION section

Moved warning escalation card to top of automod page

Moved memory tab to accounts page

Updated link filter whitelist/blacklist channels to multi-select dropdowns

XSS vulnerabilities and prettier formatting issues

Broken dropdowns and api paths in dashboard.js

Autopost settings persistence and cache issues

Security issue and rendering bug

Bump reminders deleting after not bumping immediately

Deleting memories and memory archive issues

User memory view

Voteban system bugs, including timezone issues and vote claiming exploit

Voting system bugs, including database persistence and voter only decorator

AI loading issues on cog load

Premium success buttons and credits page rendering

Voteban system bugs, including database persistence and timezone issues

Voting system exploit that allowed vote count manipulation

Security issue that revealed sensitive user information

System prompt updated for better user experience

Increased max token limit for better conversation handling

Voteban system bugs fixed with database persistence and timezone fixes

Voting system exploit fixed to prevent vote count manipulation

Bump reminders no longer delete after not bumping immediately

AI memory now updates correctly

Security issue fixed to prevent sensitive information disclosure

Various bugs fixed in panel and AI functionality

Voting systems bugs fixed

AI loaded on cog load

System prompt updated

Groq rate limits and tool calling issues

Vote claiming exploit

Timezone bugs in vote expiration

Disabling AI issue

Memory archive issue

AI memory not updating issue

Improved user memory view

Memory retrieval for mentioned users

Broken dropdowns in panel

Voting system bugs

Snipe command channel parameter type annotation

Groq rate limits and memory limits

Memory archive issues

Security issue with sensitive user information

Improved error logging for conversation response parsing

Mention detection to memory manager for Discord user mentions

Memory retrieval for mentioned users when asked about them

Updated model selection to prioritize GPT-OSS models for 50% cost savings on cached tokens

Improved memory comparison in fallback mechanism to use memory ID instead of object comparison

Fixed security issue in rendering bug

Added logging to debug tool execution request

Added environment variables to docker-compose

Added /premium landing page for premium features

Updated to node 22 for prisma 7.7.0 compatibility

Dedicated AI channel awareness with enhanced user identification

Type safety issues and unused imports removed

Memory delete 422 error fixed with reason mapping

Tool calling issues and models fixed

AI worker URL port updated from 3565 to 8000

Gracious error handling for tool execution and system prompt updates

Logging for tool execution requests

Improved tool call parsing for pydantic models and XML-style function calls

User ability to delete memories

Premium success buttons and credits page fixed

Security issue fixed and rendering bug resolved

Quick view premium status in accounts page

Users can delete memories

AI memory judgment and system prompt

Autoposting issue fixed

Mention detection for memory retrieval

Cluster isolation and cross-cluster memory support

Updated memory retrieval to use empty query for recent memories

Improved Chain of Thought pattern for active memory checking

Snipe commands to accept channel name as string parameter

Voting systems bugs, including voteban system and timezone issues

Disabling AI functionality

Memory archive and deletion issues

Security issue with sensitive user information disclosure

Enhanced user memory view and premium status display

Implemented complete enhanced AI memory system with native channel and guild support

User memory view feature

Prioritized GPT-OSS models for prompt caching benefits

Updated system prompts to reflect enhanced memory capabilities

Improved memory retrieval priority and filtering

Fixed issue where AI couldn't access stored user facts

Fixed memory retrieval issue with enhanced semantic search and fallback mechanism

Fixed security issue

Fixed tool calling issues and models being unresponsive

Fixed memory delete 422 error with reason mapping

Enhanced AI memory system with native channel and guild support

Automatic memory classification and sensitive content filtering

Improved memory retrieval with Groq Chain of Thought pattern

Memory retrieval issue where AI couldn't access stored user facts

Security issue with sensitive memories about other users

Memory delete 422 error with reason mapping

Tool calling issues and models being unresponsive

Updated model selection to prioritize GPT-OSS models for cost savings

Improved memory retrieval priority and semantic search

Fixed memory retrieval issue where AI couldn't access stored user facts

Fixed security issue and added privacy-aware memory filtering

Premium features landing page and updated premium status display

Fixed tool calling issues and improved error handling

Automatic memory classification with regex-based scoring

Intelligent memory routing and cross-channel user memory access

Sensitive content filtering for public contexts

Updated model selection to prioritize GPT-OSS models for prompt caching benefits

Improved memory retrieval priority with user, channel, and guild memories

Fixed memory comparison in fallback mechanism to use memory ID instead of object comparison

Updated premium status display in accounts page

Fixed security issue in panel

Logging for tool execution and memory retrieval

Improved error handling for tool execution and memory retrieval

Optimized memory retrieval priority to user, channel, and guild

Fixed issue where AI couldn't access stored cat/rabbit memories

Implemented mention detection to retrieve memories for mentioned users

Improved conversation context with comprehensive memory filtering

Fixed security issue where sensitive memories were revealed

Added premium landing page for premium features

Fixed rendering bug and credits page issues

Implemented enhanced AI memory system with native channel and guild support

Mention detection to retrieve memories for mentioned users

Fixed memory deletion 422 error with reason mapping

Fixed security issue in premium status rendering

Improved memory filtering for sensitive content and memory types

Updated AI worker client to support new native parameters

Fixed issue where AI memory wasn't updating

Increased max token limit

Updated system prompt

Issue with deleting memories

AI memory not updating

Improved memory retrieval priority

Preferred GPT-OSS models for prompt caching benefits

Security issue with revealing sensitive memories about other users

Improved memory retrieval priority and filtering for sensitive content

Fixed issue where AI couldn't access stored user facts and memories

Fixed memory deletion and retrieval issues

Fixed security issue and added protection for sensitive user information

Added premium landing page and updated premium status display

Issue where AI couldn't access stored user facts

Memory comparison in fallback mechanism

Paginator for listing memories

Memory import/export commands

AI system prompt updated

Task loop changed to 15 minutes

AI blocks on valid questions fixed

User conflation issue fixed

Vote claims issue fixed

Premium management and email storage from Stripe improved

Security issue fixed

Rendering bug fixed

Credits page fixed

Premium success buttons fixed

Ability to judge if something is worth remembering

AI blocks on valid questions

User conflation issue

AI issue with vote claims

Rendering bug and security issue

Premium badges and changelog automation added

Pay-as-you-go image generation with credit system

Stripe checkout for credit packages

Credit reserve buffer to prevent negative balances

Protection mechanisms for credit system

Reduced free tier limits for sustainability

Updated sustainability messaging

AI issue

Vote claims

Security issue

Rendering bug

Premium management

Premium landing page

Clarified user premium and guild premium

Credit purchase buttons to limit messages

Updated sustainability messaging explaining GPU costs

AI issue with inspect.stack() causing heartbeat blocks

Vote claims issue

Rendering bug in panel

Security issue in panel

Admin-facing bug

User premium status now visible on accounts page

Heartbeat blocks caused by inspect.stack()

Autoposting issues

Premium management and email storage from Stripe

Heartbeat blocks caused by inspect.stack() usage

Autoposting issues with legacy loop autoposts

Updated sustainability messaging to explain GPU costs

Premium management and Stripe customer ID lookup

Task loop interval changed to 15 minutes

Free tier limits reduced for sustainability

Stripe customer lookup in premium manage command

Premium features and AI customization

Legacy autoposting migration

Legacy loop autoposts insert

Unused imports and webapp uses

Unused imports and legacy autoposting migration

Unused imports and autopull feature

Legacy autoposting migration and loop autoposts insert

AI response storage without premium tag

Updated AI demo chat JavaScript to external file

Updated GitHub avatar CSP and Soheab avatar URL

Fixed GitHub avatar CSP issues

Fixed AI demo chat typewriter animation JavaScript formatting

Fixed NSFW check to use autoposts.enabled instead of hentaichannel

Removed unused imports with autoflake

Removed autopull as it is not used

Premium landing page for premium features

User premium and guild premium features

Stripe integration for premium purchases

AI demo chat JavaScript moved to external file

AI customization with premium prompts

GitHub avatar CSP issues fixed

Changelog sorting by actual date instead of db timestamp

Unused imports and webapp uses removed

Premium integration with Stripe webhook and AI custom prompts

AI visualizations to dashboard, including status cards and memory distribution bars

User premium activation on Stripe checkout completion

Conversation continuity guidance to AI personality

AI opt-in check for msgtracking and credit-aware voting CTA

GitHub API used instead of local git for docker compatibility

Changelog sorting now by actual date instead of db timestamp

NSFW check now uses autoposts.enabled instead of hentaichannel

Startup hang issue fixed

Error in AI demo chat typewriter animation JavaScript formatting fixed

GitHub avatar CSP issue fixed

Unused imports removed with autoflake

Autopull feature removed as it was not used

Webapp uses removed

Legacy autoposting migration feature

AI opt-in check for message tracking and credit-aware voting

Slash commands to panel sync

NSFW check to use autoposts.enabled instead of hentaichannel

Store original AI response without premium tag to memory

Legacy loop autoposts insert issue

Startup hang issue

Error with low credits

Unused imports

Unused autopull feature

Uses of webapp

AI visualizations to dashboard, including status cards and activity chart

AI demo chat JavaScript to external file for formatting issues

GitHub avatar CSP issue

AI demo chat typewriter animation JavaScript formatting issue

Timezone offsets for checkouts

Unused webapp

Autopull feature

User_premium activation on Stripe checkout completion

Stripe integration for premium subscriptions — $5.99/month unlocks custom AI system prompts

User-level premium — one subscription applies to all guilds you manage

AI custom system prompts — premium users can define custom personalities per guild

AI configuration dashboard — manage dedicated channels, blacklists, and custom prompts

Premium badge and animated front page showcase with AI demo chat

Automatic database table creation for user_premium on startup

Updated AI backend to check user premium instead of guild premium for custom prompts

AI demo chat typewriter animation now works correctly on front page

Enhanced security headers and CSRF protection across all routes

Rate limiting on dashboard and API endpoints to prevent abuse

Secure cookie attributes for session management

Welcomer color and background selectors now show visual previews

Improved input validation on user profile and guild configuration endpoints

All dashboard dropdowns now render safely without potential injection risks

Channel and role dropdown selectors in dashboard — no more copying IDs manually

Autopost fully configurable from panel — source, tags, interval, and channel with automatic webhook creation

Per-guild autopost intervals with dedicated asyncio tasks — interval changes take effect immediately

Custom commands CRUD — create, edit and delete server slash commands directly from the panel

Account page with Discord-style profile card — banner, bio editing, economy stats, badges and vote count

User data deletion — request permanent removal of all personal data stored by Akiko

Automod duplicate message detection and warning escalation config now configurable from panel

Link filter domain and channel whitelists/blacklists now editable from panel

Dashboard server picker now only shows servers where Akiko is actually present

Welcomer card appearance fully configurable — text color, background, template and font

Bump reminder supports fallback ping type and fallback role for when the primary bumper is unavailable

Starboard max message age now configurable in days

Autopost channel changes now take effect immediately via pg_notify without requiring a bot restart

Music node health check infinite reload loop caused by enum comparison bug

Consecutive failure counter to gate auto-reloads on music nodes

Guild member count now excludes bots for more accurate statistics

Various stability improvements across autoposting and NSFW cogs

e621 support for autoposting alongside rule34

Persistent post ID tracking per guild to prevent duplicate content

Autopost now runs every 5 minutes on a fixed schedule

Webhook cache invalidation on Forbidden errors

Vote ban system — community-driven moderation voting

Sticky roles — roles re-applied automatically when members rejoin

Counting channels with configurable goals and reset behavior

Economy overhaul with job system, streaks and gambling bonuses

Anti-zalgo, anti-hoist, and duplicate message detection in automod

Account age and avatar checks on member join

Link filter with per-channel whitelist and blacklist support

Report system — members can report users directly to staff

Slash command support across all major commands

Sharded architecture across multiple clusters

Real-time web panel at akiko.motzumoto.com

Complete internal rewrite for stability and performance

Legacy prefix-only commands deprecated in favor of hybrid commands