Configuration

AbsoluteSolver is configured through plugins/Solver/config.yml. New options added by an update merge in automatically on startup — anything you've already customized is left untouched.

Diagnostics Reference

# Language (en_US, es_ES, pt_BR, ru_RU, de_DE, fr_FR, ja_JP, ko_KR, eo_EO)
localization: en_US

# Automatically analyze whether there was a crash when the server starts up
check-crash-on-startup: true

# Monitor the entire console for errors (red text)
monitor-console: true
# Avoid analyzing the same error multiple times (minutes to wait)
error-cooldown-minutes: 10

# Show the giant ASCII art banner in the console on startup?
show-banner: true

# AI Provider Configuration
ai-provider:
  # Provider: 'anthropic', 'google' or 'other'
  provider: "anthropic"

  # Model to use
  # Anthropic: Sonnet, Haiku, Opus
  # Google: Pro, Flash
  # Other: Model name (e.g: gpt-4o, mistral-large)
  model: "Sonnet"

  # API Keys
  anthropic-key: ""
  google-key: ""

  # Configuration for the 'other' provider (OpenAI-compatible)
  other-key: ""
  other-url: "https://api.openai.com/v1/chat/completions"

# Analysis configuration
analysis:
  auto-analyze: true
  timeout: 60
  save-to-file: true
  reports-directory: "crash-reports"
  # How many days to keep analyses saved before deleting them. 0 = keep forever.
  reports-retention-days: 0

# Notifications
notifications:
  notify-admins: true
  send-to-chat: false

Default provider changed to Anthropic

As of this release, the shipped default is Anthropic Claude (provider: "anthropic", model Sonnet), not Google Gemini. Gemini and any OpenAI-compatible endpoint remain fully supported — see Fyrx — Your AI Assistant to switch.

Option Reference — Diagnostics

Key	Type	Default	Description
localization	String	en_US	Language for Fyrx's AI responses and in-game messages.
check-crash-on-startup	Boolean	true	Scan crash-reports/ on boot.
monitor-console	Boolean	true	Intercept console errors via Log4j.
error-cooldown-minutes	Integer	10	Minutes between consecutive analyses.
show-banner	Boolean	true	Show ASCII banner on startup.
ai-provider.provider	String	anthropic	Which AI backend to use: anthropic, google, or other.
ai-provider.model	String	Sonnet	The specific model name for the selected provider.
analysis.reports-retention-days	Integer	0	Days to keep files in reports-directory before auto-deleting. 0 = forever.

Chat Moderation

Fyrx can read chat and evaluate the context of a conversation instead of matching a word blacklist. Enabling this sends chat content to the AI provider configured above. Disabled by default for privacy.

chat-moderation:
  enabled: false
  exempt-ops: true
  buffer-size: 50
  context-window-minutes: 5
  analysis-interval-seconds: 45
  scan-mode: "bulk"          # "bulk" or "individual"
  llm-analysis-level: 2      # 0-3, see below

  pre-filter:
    trigger-on-repeated-target: true
    trigger-on-caps-ratio: 0.6
    min-messages-before-trigger: 3
    urgent-score-threshold: 0.65

  action:
    mode: "alert-only"       # "alert-only", "warn-player", or "auto-action"
    severity-threshold-mute: 4
    severity-threshold-kick: 5
    mute-duration-minutes: 10
    min-confidence-to-act-alone: 85
    min-severity-to-act-alone: 4

  log-incidents-to-file: true
  log-test-results-to-console: true
  reports-retention-days: 0

llm-analysis-level — how often the AI is actually called

scan-mode decides how the AI is called (one call for the whole conversation, or one per active player); llm-analysis-level decides when:

Level	Behavior
0	Never calls the AI — the local pre-filter decides entirely on its own. Works with no AI provider configured at all.
1	Only calls the AI once the pre-filter already flagged something suspicious.
2 (default)	The AI reviews all new chat every cycle, regardless of the pre-filter.
3	Real-time: every message is sent to the AI immediately, instead of waiting for analysis-interval-seconds.

Multi-signal corroboration

A moderation verdict from the AI alone can only trigger an automatic sanction (action.mode: "auto-action") when it's both at least min-confidence-to-act-alone confident and at least min-severity-to-act-alone severe. Below that threshold, an automatic sanction also requires at least one corroborating local pre-filter signal (shouting, an insult keyword, hammering the same target, a burst of messages) — otherwise the verdict stays a staff-only alert for manual review via /solver moderation status or /solver check.

Option Reference — Chat Moderation

Key	Type	Default	Description
chat-moderation.enabled	Boolean	false	Master switch. Sends chat to your configured AI provider when on.
chat-moderation.exempt-ops	Boolean	true	OPs are automatically exempt. solver.moderation.bypass exempts a specific player regardless of this setting.
chat-moderation.buffer-size	Integer	50	Maximum recent messages remembered per player/globally.
chat-moderation.context-window-minutes	Integer	5	Time window of context sent to the AI.
chat-moderation.analysis-interval-seconds	Integer	45	How often the unconditional sweep checks for new chat to analyze.
chat-moderation.scan-mode	String	bulk	bulk = one AI call for the whole conversation (cheaper, can confuse who said what). individual = one call per active player (no confusion, costs scale with active players).
chat-moderation.action.mode	String	alert-only	alert-only notifies staff only. warn-player also privately notifies the flagged player. auto-action additionally mutes/kicks automatically based on severity.
chat-moderation.action.severity-threshold-mute	Integer	4	Severity (1-5) at which auto-action automatically mutes.
chat-moderation.action.severity-threshold-kick	Integer	5	Severity at which auto-action automatically kicks instead of muting.
chat-moderation.action.min-confidence-to-act-alone	Integer	85	Confidence (0-100) required for the AI verdict alone to sanction, without pre-filter corroboration.
chat-moderation.action.min-severity-to-act-alone	Integer	4	Severity (1-5) required for the AI verdict alone to sanction, without pre-filter corroboration.

Tag Overrides — mapping a "situation" to a specific sanction

severity-threshold-mute/severity-threshold-kick only reach as far as a kick — auto-action never bans on its own. chat-moderation.action.tag-overrides lets you force a specific sanction type for a specific AI tag instead, regardless of severity. This is the only way to reach an automatic ban/tempban from chat moderation:

chat-moderation:
  action:
    tag-overrides:
      THREAT:
        type: tempban
        duration: 7d
      HATE_SPEECH:
        type: tempban
        duration: 3d
      SCAM:
        type: ban

• Valid types: warn, mute, tempmute, kick, ban, tempban (duration required for the temp- variants).
• Only used when action.mode is auto-action.
• If a verdict matches more than one configured tag, the harshest configured type wins.

tags.yml — define your own moderation categories

The categories the AI can use (TOXIC, HARASSMENT, THREAT, HATE_SPEECH, SCAM, SPAM by default) live in plugins/Solver/tags.yml, not hardcoded in the plugin. Add, edit, or remove a tag there and the prompt sent to the AI updates automatically — no recompiling, no code changes:

tags:
  TOXIC: "Rude, insulting, or demeaning language with real malicious intent (not friendly banter between people who are fine with it)."
  HARASSMENT: "Insistent, repeated targeting of the same player, especially after they show discomfort or ask to stop."
  THREAT: "Threats of violence, real-world harm, or encouraging self-harm directed at someone."
  HATE_SPEECH: "Attacks based on race, religion, gender, sexual orientation, nationality, or similar."
  SCAM: "Attempts to defraud or phish another player (fake giveaways, asking for passwords/account info/real money)."
  SPAM: "Repetitive, advertising, or flooding messages with no other issue."

Whatever tag name you define here is also what you reference in chat-moderation.action.tag-overrides above. Re-applied automatically on /solver reload — no restart needed to add a new category.

Standalone Commands

Lets each command group (see Commands) also register without the /solver prefix — turn a category off if another installed plugin already owns one of those command names:

standalone-commands:
  staff-mode: true  # /vanish, /freeze, /staffchat, /commandspy, /enderchest (unless EssentialsX is installed)
  sanctions: true   # /warn, /mute, /tempmute, /kick, /ban, /tempban, /unban, /unmute, /unwarn, /history, /check, /note, /checkuser, /appeal, /sanctions
  moderation: true  # /moderation
  reports: true     # /report

CommandSpy

Live relay of every command run on the server to whichever staff toggled it on with /solver commandspy — unlike retrospective command logging, this shows up as it happens.

commandspy:
  enabled: true
  watch-patterns: ["*"]
  exempt-players: []
  exempt-permissions: []

Key	Type	Default	Description
commandspy.enabled	Boolean	true	Master switch.
commandspy.watch-patterns	List	["*"]	"*" watches every command. Otherwise list specific command names (without the leading /) to only relay those, e.g. ["op", "gamemode", "give"].
commandspy.exempt-players	List	[]	Player names (case-insensitive) never relayed, no matter who's watching.
commandspy.exempt-permissions	List	[]	Anyone holding one of these permissions is never relayed either.

/solver inspect and /solver enderchest

inspect:
  live-refresh: false
  live-refresh-interval-ticks: 10

Key	Type	Default	Description
inspect.live-refresh	Boolean	false	Keep re-capturing the opened inventory every few ticks instead of a one-time snapshot. This is polling, not a real push update — Bukkit has no "inventory changed" event to react to. Shared by both /solver inspect and /solver enderchest.
inspect.live-refresh-interval-ticks	Integer	10	How often to refresh, in ticks, when the option above is on.

Staff Mode Toolkit

Beta

staff-mode:
  profiles: {}
  #  moderator:
  #    vanish: true
  #    freeze-immunity: true
  #    flight: true
  #    god-mode: true
  #    clear-inventory: true
  #  builder:
  #    flight: true
  vanish:
    invulnerable-while-vanished: false
    hide-from-server-list: true
  freeze:
    allowed-commands: ["msg", "tell", "r", "helpop"]

Key	Type	Default	Description
staff-mode.profiles.<name>	Section	(empty)	Named combined profile for /solver staffmode <name> — toggles a bundle of individual staff tools at once. Every field is optional and defaults to off (vanish, freeze-immunity, flight, god-mode, clear-inventory). clear-inventory always saves your items first and restores them when you leave the profile.
staff-mode.vanish.hide-from-server-list	Boolean	true	Also subtract vanished players from the server-list player count.
staff-mode.vanish.invulnerable-while-vanished	Boolean	false	Makes a vanished player immune to damage while vanished.
staff-mode.freeze.allowed-commands	List	["msg", "tell", "r", "helpop"]	Commands (without the leading /) a frozen player can still run. The freeze-toggle command itself always works regardless of this list, so a frozen staff member can never get permanently stuck.

/solver vanish strict and solver.staffmode.vanish.see-strict (a separate permission, not inherited from regular vanish) add a second tier that's invisible even to most staff. Staff with solver.staffmode.silentjoin connect already vanished, with no join message.

Sanctions System

Beta

Storage backend

punishments:
  storage: sqlite
  mysql:
    host: localhost
    port: 3306
    database: solver
    user: solver
    password: ""

sqlite (default, zero configuration, sanctions.db) or mysql for multi-server setups sharing one sanctions database — same commands, same data either way. Sanction history and active mutes/bans are restored automatically on join or server restart regardless of which backend you use.

Reason templates & warn escalation

punishments:
  reason-templates: {}
  #  griefing:
  #    text: "Griefing / destroying other players' builds"
  #    duration: "1d"
  warn-escalation: {}
  #  "3":
  #    action: sanction
  #    type: kick
  #    reason: "Auto: 3 warnings"
  #  "5":
  #    action: sanction
  #    type: tempban
  #    duration: "1d"
  #    reason: "Auto: 5 warnings"

• reason-templates — /solver ban Player #griefing expands #griefing to the configured text (and duration, if the sanction type needs one and none was given explicitly). A #name with no matching template is left as a literal reason, so this never breaks a reason that legitimately starts with #.
• warn-escalation — after a WARN is recorded, if the player's total warn count matches a key here, the configured action fires automatically: action: sanction applies another sanction (type/reason/duration, same fields as a manual one), action: command runs a console command instead ({player} is replaced with the player's name).

Appeals

punishments:
  appeals:
    show-in-sanction-message: true

Shows a line pointing to /solver appeal <id> <reason> in warn/mute/kick/ban messages, entirely in-game — no Discord webhook needed. Staff are notified immediately when a new appeal comes in, the same way a new report notifies them.

Usage Statistics

metrics:
  enabled: true

Anonymous usage statistics via bStats — server count, which AI provider is configured, which features are enabled. Nothing player-identifiable. Set metrics.enabled: false to opt out, independent of the global bStats toggle in plugins/bStats/config.yml (which also always applies).

Proxy Relay

Requires a separate plugin on the proxy

This section only configures the Bukkit side. It does nothing unless the matching proxy plugin is also installed — see Proxy Relay for the full setup.

proxy-relay:
  enabled: false

Off by default. When enabled, relays staffchat messages and moderation/integrity alerts to staff connected to any backend server on the same BungeeCord/Waterfall/Velocity network, not just the one where the alert happened.

Supported Languages

Code	Language
en_US	English (United States)
es_ES	Spanish (Spain)
pt_BR	Portuguese (Brazil)
de_DE	German
fr_FR	French
ru_RU	Russian
ja_JP	Japanese
ko_KR	Korean
eo_EO	Esperanto

Every player/staff-facing message (help, errors, sanctions, moderation alerts) lives in plugins/Solver/lang/<locale>/messages.yml with full MiniMessage support, and is safe to edit — your customizations are preserved across updates.

Using Alternative AI Providers

Google Gemini

ai-provider:
  provider: "google"
  model: "Flash"
  google-key: "AIza..."

OpenAI or any Compatible API

ai-provider:
  provider: "other"
  model: "gpt-4o"
  other-key: "sk-..."
  other-url: "https://api.openai.com/v1/chat/completions"

This also works with local models via Ollama or LM Studio by pointing other-url to their local endpoint.