CMIBot

CMIBot is a Discord support bot built around a channel-aware /lookup workflow.

Today it fully supports the CMI context, includes a working Jobs context, and now also supports lightweight plugin contexts for SVIS, Residence, MFM, TryMe, and TradeMe, with the active plugin decided by the Discord channel ID.

Current Direction

• /lookup is now the primary slash command
• Channel ID decides which plugin context is active
• #test can be switched live between plugin contexts through the debug command
• /lookup reload is global and rebuilds caches for every plugin context
• /lookup stats and /lookup langstats are context-aware and only show data for the active plugin
• /lookup debug shows channel routing, tracked plugins, file counts, uptime, runtime versions, cache health, memory usage, disk footprint, and active test overrides

Current Contexts

CMI

The CMI context is fully wired and currently supports:

• config
• language|lang
• placeholder
• material
• command|cmd
• permission|perm
• faq
• tabcomplete
• langstats
• stats
• debug
• reload
• help

Jobs

The Jobs context is now wired for Jobs locale files, Jobs translatable-word files, shared CMILib YAML search, and Jobs log-based support data.

Right now Jobs supports:

• config
• language|lang
• placeholder
• command|cmd
• permission|perm
• faq
• help
• stats
• langstats
• debug
• reload

These are currently not part of the Jobs scope:

• material
• tabcomplete

SVIS

The SVIS context currently supports:

• config
• language|lang
• command|cmd
• permission|perm
• langstats
• stats
• debug
• reload
• help

Residence

The Residence context currently supports:

• config
• language|lang
• placeholder
• command|cmd
• permission|perm
• langstats
• stats
• debug
• reload
• help

MFM

The MFM context currently supports:

• config
• language|lang
• langstats
• stats
• debug
• reload
• help

TryMe

The TryMe context currently supports:

• config
• language|lang
• langstats
• stats
• debug
• reload
• help

TradeMe

The TradeMe context currently supports:

• config
• language|lang
• langstats
• stats
• debug
• reload
• help

Example Commands

CMI examples

/lookup help
/lookup config dynmap
/lookup config chat file:Chat.yml
/lookup config "mini message" mode:whole
/lookup config bluemap related:true
/lookup language home
/lookup lang "was fireballed by"
/lookup placeholder balance
/lookup material shulker
/lookup cmd balance
/lookup perm cmi.command.balance
/lookup faq refund
/lookup tabcomplete [playername] mode:whole
/lookup stats
/lookup langstats
/lookup debug
/lookup reload

Jobs examples

/lookup help
/lookup language points
/lookup placeholder jobsr_user_points
/lookup cmd join
/lookup perm jobs.use
/lookup faq vault
/lookup stats
/lookup langstats

SVIS examples

/lookup help
/lookup config particle
/lookup language selection
/lookup cmd gui
/lookup perm sv.worldedit.use
/lookup langstats
/lookup stats

Residence examples

/lookup help
/lookup config build
/lookup language invalid
/lookup placeholder owner
/lookup cmd set
/lookup perm residence.select
/lookup langstats
/lookup stats

MFM examples

/lookup help
/lookup config farm
/lookup language mob
/lookup langstats
/lookup stats

TryMe examples

/lookup help
/lookup config tryme
/lookup language message
/lookup langstats
/lookup stats

TradeMe examples

/lookup help
/lookup config trade
/lookup language seller
/lookup langstats
/lookup stats

Test-channel context switching

When used in a configured test channel by an admin:

/lookup debug context:cmi
/lookup debug context:jobs
/lookup debug context:auto

auto clears the manual override and returns the test channel to its default configured context.

Channel Routing

The bot uses explicit channel IDs from .env:

• DISCORD_ALLOWED_CHANNEL_IDS: every channel where the bot is allowed to run
• DISCORD_CMI_CHANNEL_IDS: channels that should route to the CMI context
• DISCORD_JOBS_CHANNEL_IDS: channels that should route to the Jobs context
• DISCORD_SVIS_CHANNEL_IDS: channels that should route to the SVIS context
• DISCORD_MFM_CHANNEL_IDS: channels that should route to the MFM context
• DISCORD_TRYME_CHANNEL_IDS: channels that should route to the TryMe context
• DISCORD_TRADEME_CHANNEL_IDS: channels that should route to the TradeMe context
• DISCORD_RESIDENCE_CHANNEL_IDS: channels that should route to the Residence context
• DISCORD_TEST_CHANNEL_IDS: channels that are allowed to override context live
• DISCORD_TEST_DEFAULT_CONTEXT: default context for test channels, currently cmi

Current defaults:

• #cmi: 526402563847880725
• #jobs-reborn: 526402919826849804
• #selectionvisualizer: 714110524731686962
• #mobfarmmanager: 713838315572559892
• #tryme: 714111148059787285
• #trademe: 713838991744434277
• #residence: 526403195476639744
• #test: 1493976695152054353

If a channel is not in DISCORD_ALLOWED_CHANNEL_IDS, the bot refuses to run there.

What Loads On Startup

On npm start, the bot now warms a global cache and prints totals grouped by plugin context, for example:

Loaded 19563 entries from 74 files into the search cache.
CMI:
- config: 4499 entries from 21 YAML configuration files
- language: 4145 entries from 2 YAML locale files
- placeholder: 224 entries from 1 placeholder data file
- material: 1697 entries from 1 material data file
- command: 306 entries from 1 command data file
- permission: 778 entries from 2 permission data files
- faq: 54 entries from 1 FAQ data file
- tabcomplete: 77 entries from 1 tab-complete data file
Jobs:
- config: 347 entries from 1 YAML configuration file
- language: 2655 entries from 2 YAML locale files
- placeholder: 72 entries from 1 placeholder data file
- command: 57 entries from 1 command data file
- permission: 65 entries from 1 permission data file
- faq: 35 entries from 32 FAQ data files
Shared CMILib data:
- config: 36 entries from 1 YAML configuration file
- language: 2240 entries from 2 YAML locale files

/lookup reload rebuilds this cache globally for every configured plugin context.

For startup and reload summaries, shared CMILib config and language data is shown in its own Shared CMILib data: section at the bottom so it does not visually look like plugin-owned data inside the CMI or Jobs blocks.

Debug Output

/lookup debug is a compact health dashboard for the current channel context. It now reports:

• detected plugin context
• channel type and channel ID
• tracked plugin list
• known channel-route counts
• current context file counts
• uptime
• Node and discord.js versions
• project disk size
• RAM RSS and heap usage
• global and current-context cache totals
• last cache reload timestamp
• largest cache bucket
• active test-channel overrides
• tracked plugin disk footprint
• which commands are available or unsupported in the current context

Search Behavior

Shared options

• mode: exact|whole|broad
• limit: 1-15 for most commands
• summary: true|false

CMI-specific extras

• config supports file: <name>
• config, language, and lang support related: true|false
• material uses limit: 1-25 and defaults to 25

AI summary

AI support is behind two gates:

• OPENAI_ENABLED=true
• the user must have one of the configured AI_ROLE_IDS

If AI is disabled, summary:true is rejected cleanly.

Security and Abuse Controls

The bot already includes:

• per-user cooldowns
• query length checks
• blocklisted filler-word rejection
• allowlisted short-token exceptions
• disallowed @ and backtick input rejection
• no-mention Discord replies
• audit logging to logs/cmibot-usage.jsonl
• safe file: filtering against indexed files only

Because file filtering uses the active plugin context, a Jobs channel cannot search CMI config files and vice versa.

At the moment, Jobs config includes:

• JobsPlugin/generalConfig.yml
• shared CMILibPlugin/CMILib/config.yml

Jobs language now includes:

• JobsPlugin/locale/messages_en.yml
• JobsPlugin/TranslatableWords/Words_en.yml
• shared CMILibPlugin/CMILib/Translations/**/*_EN.yml

Jobs faq can now load from both:

• JobsPlugin/data/faq.log
• JobsPlugin/data/faq/*.md

That means new markdown FAQ files dropped into the Jobs FAQ folder are picked up on /lookup reload without needing a separate conversion step.

Environment

Copy .env.example to .env and fill in the values.

Discord

• DISCORD_TOKEN
• DISCORD_APPLICATION_ID
• DISCORD_GUILD_ID
• DISCORD_ALLOWED_CHANNEL_IDS
• DISCORD_CMI_CHANNEL_IDS
• DISCORD_JOBS_CHANNEL_IDS
• DISCORD_SVIS_CHANNEL_IDS
• DISCORD_MFM_CHANNEL_IDS
• DISCORD_TRYME_CHANNEL_IDS
• DISCORD_RESIDENCE_CHANNEL_IDS
• DISCORD_TEST_CHANNEL_IDS
• DISCORD_TEST_DEFAULT_CONTEXT
• ALLOWED_ROLE_IDS
• ADMIN_ROLE_IDS
• AI_ROLE_IDS

OpenAI

• OPENAI_ENABLED
• OPENAI_API_KEY
• OPENAI_MODEL

Shared bot behavior

• DISPLAY_PATH_PREFIX
• DEFAULT_RESULT_LIMIT
• LOOKUP_COOLDOWN_SECONDS
• SUMMARY_COOLDOWN_SECONDS
• QUERY_MIN_LENGTH
• QUERY_MAX_LENGTH
• QUERY_BLOCKLIST
• QUERY_ALLOWLIST
• QUERY_DEBUG_ERRORS
• AUDIT_LOG_PATH

CMI data scopes

• LOOKUP_INCLUDE_GLOBS
• LOOKUP_EXCLUDE_GLOBS
• LANGLOOKUP_INCLUDE_GLOBS
• LANGLOOKUP_EXCLUDE_GLOBS
• PLACEHOLDER_INCLUDE_GLOBS
• PLACEHOLDER_EXCLUDE_GLOBS
• MATERIAL_INCLUDE_GLOBS
• MATERIAL_EXCLUDE_GLOBS
• COMMAND_INCLUDE_GLOBS
• COMMAND_EXCLUDE_GLOBS
• PERMISSION_INCLUDE_GLOBS
• PERMISSION_EXCLUDE_GLOBS
• FAQ_INCLUDE_GLOBS
• FAQ_EXCLUDE_GLOBS
• TABCOMPLETE_INCLUDE_GLOBS
• TABCOMPLETE_EXCLUDE_GLOBS

Jobs data scopes

These are the current Jobs search scopes:

• JOBS_LOOKUP_INCLUDE_GLOBS
• JOBS_LOOKUP_EXCLUDE_GLOBS
• JOBS_LANGUAGE_INCLUDE_GLOBS
• JOBS_LANGUAGE_EXCLUDE_GLOBS
• JOBS_PLACEHOLDER_INCLUDE_GLOBS
• JOBS_PLACEHOLDER_EXCLUDE_GLOBS
• JOBS_COMMAND_INCLUDE_GLOBS
• JOBS_COMMAND_EXCLUDE_GLOBS
• JOBS_PERMISSION_INCLUDE_GLOBS
• JOBS_PERMISSION_EXCLUDE_GLOBS
• JOBS_FAQ_INCLUDE_GLOBS
• JOBS_FAQ_EXCLUDE_GLOBS

Lightweight plugin data scopes

These are the current YAML-only plugin search scopes:

• SVIS_LOOKUP_INCLUDE_GLOBS
• SVIS_LOOKUP_EXCLUDE_GLOBS
• SVIS_LANGUAGE_INCLUDE_GLOBS
• SVIS_LANGUAGE_EXCLUDE_GLOBS
• SVIS_COMMAND_INCLUDE_GLOBS
• SVIS_COMMAND_EXCLUDE_GLOBS
• SVIS_PERMISSION_INCLUDE_GLOBS
• SVIS_PERMISSION_EXCLUDE_GLOBS
• RESIDENCE_PLACEHOLDER_INCLUDE_GLOBS
• RESIDENCE_PLACEHOLDER_EXCLUDE_GLOBS
• RESIDENCE_LOOKUP_INCLUDE_GLOBS
• RESIDENCE_LOOKUP_EXCLUDE_GLOBS
• RESIDENCE_LANGUAGE_INCLUDE_GLOBS
• RESIDENCE_LANGUAGE_EXCLUDE_GLOBS
• RESIDENCE_COMMAND_INCLUDE_GLOBS
• RESIDENCE_COMMAND_EXCLUDE_GLOBS
• RESIDENCE_PERMISSION_INCLUDE_GLOBS
• RESIDENCE_PERMISSION_EXCLUDE_GLOBS
• MFM_LOOKUP_INCLUDE_GLOBS
• MFM_LOOKUP_EXCLUDE_GLOBS
• MFM_LANGUAGE_INCLUDE_GLOBS
• MFM_LANGUAGE_EXCLUDE_GLOBS
• TRYME_LOOKUP_INCLUDE_GLOBS
• TRYME_LOOKUP_EXCLUDE_GLOBS
• TRYME_LANGUAGE_INCLUDE_GLOBS
• TRYME_LANGUAGE_EXCLUDE_GLOBS
• TRADEME_LOOKUP_INCLUDE_GLOBS
• TRADEME_LOOKUP_EXCLUDE_GLOBS
• TRADEME_LANGUAGE_INCLUDE_GLOBS
• TRADEME_LANGUAGE_EXCLUDE_GLOBS

Current Data Layout

Right now the live CMI data remains where it already works:

CMIPlugin/CMI/
CMIPlugin/data/
CMILibPlugin/CMILib/
JobsPlugin/data/
SVISPlugin/
ResidencePlugin/
MFMPlugin/
TryMePlugin/
TradeMePlugin/
src/

This refactor now supports the plugin-folder layout you created, while still keeping the data-loading rules driven by env globs.

That means:

• CMI behavior stays stable
• Jobs can be added incrementally
• new plugin contexts can follow the same folder-and-glob pattern later

Local CLI

The local CLI now understands plugin context too:

node src/cli.js cmi stats
node src/cli.js jobs stats
node src/cli.js cmi config --file Chat.yml dynmap

If no plugin is provided, it defaults to cmi.

Install and Run

npm install
npm start

Git Notes

• .env is ignored
• .env.example is tracked
• logs/ is ignored
• CMIPlugin/CMI/cmi.sqlite.db is ignored as runtime data

Next Planned Steps

• add real Jobs YAML config data when you are ready to index more than the shared CMILib files
• extend the multi-plugin structure to more support channels beyond the current CMI, Jobs, SVIS, Residence, MFM, TryMe, and TradeMe setup