Build Model
Build System
Builds define exactly what members need to prove for a role, activity, or event. Depending on the game, that proof might be screenshots, ESO CharacterMarkdown, TONL, or another adapter.
What Is a Build?
A build is a gear specification. Think of it as a checklist that says to be verified for this content, you need these items. Every build has a name, and it belongs to an activity.
Builds contain categories, and categories contain items. When the bot scores a submission, it works through each enabled section and checks how much of the required spec it can find in the submitted evidence.
Game and Evidence Source Editors
The build editor changes based on the pack's game and the selected evidence source. This keeps New World OCR builds from feeling like ESO export builds, and keeps each supported game from inheriting the wrong assumptions.
OCR Spec
Use for screenshot-based evidence. Configure categories, items, aliases, exclusions, targets, category weights, and OCR health checks.
ESO Markdown
Use the default CharacterMarkdown export. It supports character basics, equipment, skill bars, champion points, stats, and compact Discord summaries.
ESO TONL
Use advanced CharacterMarkdown TONL when staff want deeper typed data such as passives. TONL does not currently include champion point allocation, so CP checks stay in Markdown.
Disabled editors keep their saved state where possible. Switching evidence source changes which spec the bot uses for live checks.
Activities
An activity is a grouping label for your builds. Common examples: Raids, Dungeons, PvP, Wars. When you set up a gearcheck channel with /gearcheck setup activity:raids, the channel will only show builds that belong to that activity.
This keeps things organized. Your raiders see raid builds. Your dungeon runners see dungeon builds. Nobody has to scroll through a list of builds they do not care about.
Categories
Categories are the gear types within a build. For example, a build might have categories for Weapons, Armor Perks, Amulet, Ring, and Trophies. Each category gets scored separately, and the category scores combine into the overall build score.
/pack add-category build:tank-build name:Armor PerksCategories can have different weights, so you can make some categories count more than others toward the final score.
Items
Items are the individual things being checked within a category. If your Armor Perks category needs Refreshing, Freedom, and Resilient, those are three items. In OCR builds, each item gets matched against extracted text. In structured builds, fields are compared against parsed export data.
/pack add-item build:tank-build category:armor-perks name:RefreshingWhen adding items, think about what text or structured field the bot will see in the submitted evidence. The name should match what actually appears in-game or in the export, and you can add aliases to catch variations.
Aliases
Aliases are alternative names for an item. They exist because OCR and human-entered specs are not perfect, and because games sometimes display things differently than you would expect. The bot uses fuzzy matching where appropriate, so it can handle small typos on its own. But aliases help in cases where the difference is bigger.
For example:
- An item named Sword and Shield might also appear as SnS or Sword & Shield in screenshots.
- OCR might read Refreshing as Refrshing or Refreshlng.
- A perk might be shortened in the UI compared to its full name.
/pack edit-item build:tank-build category:weapons item:Sword and Shield aliases:SnS,Sword & Shield,S&SGood aliases make the difference between a 60% score and a 95% score. If your members are getting lower scores than expected, check aliases first.
Excludes
Excludes prevent false positives. Sometimes an item name is a substring of another item, and the bot might match the wrong one. For example, Refreshing is part of Refreshing Move. If your build needs Refreshing but not Refreshing Move, you can add Refreshing Move as an exclusion on the Refreshing item.
/pack edit-item build:tank-build category:armor-perks item:Refreshing excludes:Refreshing Move,Refreshing WardWhen the bot sees Refreshing Move in the OCR text, it will not count it as a match for the Refreshing item. This keeps your scores accurate.
Build Health and Screenshot Reading
The dashboard includes build-health tools for catching risky specs before members run into them. For OCR builds, the scanner can surface alias collisions, exclusion suggestions, source-pattern issues, and recurring misses from previous checks. For structured evidence, the same idea applies to missing fields, unsupported sections, and sections that are enabled in the editor but absent from member exports.
Use these tools after creating a build and again after real submissions start coming in. They are meant to turn why did this fail into a short list of concrete edits.
Weights
Not all categories are equally important. Weights let you control how much each category counts toward the final score. By default, every category has a weight of 1.0. You can increase a weight to make a category count more, or decrease it to make it count less.
For example, if trophies are harder to get and more important for your content, you might give the Trophies category a weight of 1.5. That means it contributes 50% more to the final score than a category with weight 1.0.
Example
A build with three categories: Weapons (1.0), Armor (1.0), Trophies (1.5). Someone scores 80% on weapons, 100% on armor, 60% on trophies. The weighted score: (0.8*1.0 + 1.0*1.0 + 0.6*1.5) / (1.0 + 1.0 + 1.5) = 2.7 / 3.5 = 77%
Thresholds
Thresholds determine what status a member gets based on their score. There are four status levels:
Verified
They passed. Their gear meets the build requirements. If you have a verified role configured, the bot assigns it automatically.
Tentative
Close, but not quite there. They might be missing one or two items, or the evidence parser might have had trouble reading something. Usually worth a manual review.
Borderline
Significantly below the threshold. They might have the right gear but incomplete evidence, or they might genuinely be missing several required items.
Unverified
The score is too low to indicate they have the required gear. They need to update their gear or try again with better evidence.
Rule Types
Each item in a build can have a rule type that controls how it affects scoring:
Required (default)
Must be present. Full credit if matched, half credit if partially matched, zero if missing.
Optional (bonus)
Bonus credit if found, no penalty if missing. Use for nice-to-have perks.
Prohibited
Must NOT be present. Penalizes if detected. Only shows in results when found.
Game Data Search (New World)
New World packs have access to the full game database in the web build editor. Type a perk or item name to search, and selecting a result auto-fills aliases, excludes, and description fragments for accurate OCR matching. You can also type custom names without selecting from the dropdown.
Trophy Detection (New World)
Enable trophy detection on builds to auto-detect combat trophies from trophy grid screenshots. Select which types are required (Angry Earth, Wildlife, Lost) and the bot handles the rest. Ultimate Combat Trophies count toward all types. Target is always 3, one per house. Requires the trophy_detector_nw plugin to be enabled.
Editing builds on the web dashboard
The web Build Editor is the full authoring surface for builds. It does everything the /pack commands do (categories, items, aliases, weights) and adds what the commands cannot: the OCR risk scanner, revision history, and, for Elder Scrolls Online, the structured rules and evidence source that /pack cannot author at all (see the note below).
Open your Dashboard , pick a server and pack, then open a build to land in the Build Editor. Its work is organized into tabs:
Info
Name, activity, difficulty, and the role, spec, and emoji that the build is granted under, including an emoji picker for custom emoji. This covers everything in the Discord edit-build command and adds the role/spec/emoji fields.
Build Spec
Visual create, edit, delete, and reorder of categories and items, with weights, aliases, excludes, and item description phrases. Advanced rules let you set manual_review_if conditions and a progression_gate on a category.
ESO Spec (ESO builds only)
Choose the evidence source for the build: character Markdown, Markdown + TONL, or OCR screenshots, and import a character export to populate gear, sets, and attributes. The old separate Markdown and TONL tabs are now merged into this one tab.
Plugins
Game-specific detection. For New World this is attribute detection (a target total or a min/max range) and trophy detection (a target per trophy type).
Build Health
An OCR risk scanner that flags items likely to misread and offers one-click safe fixes. It is informational, never blocking, and has no Discord command equivalent.
Test
The web version of /pack test: paste evidence to preview how it scores, and accept OCR alias suggestions straight into the spec.
History
Revision history with provenance, so you can see what changed between saves and who made each edit.
The ESO Spec tab only appears for Elder Scrolls Online builds. The Plugins tab shows attribute and trophy detection only for games that support them.
ESO builds are set up here, not with /pack
Elder Scrolls Online builds are scored from a character Markdown or TONL export using structured rules (sets, slots, skills, attributes). Those rules and the build's evidence source are authored only in the ESO Spec tab above. The /pack item commands and the OCR category/item model on this page apply to screenshot-based games like New World and are not used to score ESO builds. Create and edit ESO builds in the web Build Editor; if you build one with /pack item commands, members will be asked for a character export and the OCR items will be ignored. You don't have to write every rule by hand either: import a build from Hyperioxes, or paste a character Markdown or TONL export, from the build editor's import tab or the equivalent import option on the pack page. A build imported from Hyperioxes carries a short attribution note that shows up on the pack's public requirements page.
Pack view and sharing
The pack detail page that lists your builds has a couple of controls worth knowing:
- Model / Library toggle: switch between the activity-role-spec model view and a flat library list of every build in the pack.
- The Library view is the quickest way to find and open a single build when a pack has many.
- Public Requirements toggle: generate a shareable read-only page of the pack's requirements that members can view without Discord.
Walkthroughs
- Role and spec model how activities, roles, and specs map to builds
- Creating packs by evidence type choosing OCR, Markdown, or TONL for a build
- OCR optimization and Build Health using the risk scanner to tighten accuracy
- Public requirements pages sharing a build's requirements outside Discord
Templates vs Custom Builds
You can start from a template pack or build everything from scratch. Templates are pre-configured builds with categories, items, and aliases already set up. They save a lot of time if your game has one available.
Template Packs
- +Ready to go in seconds
- +Aliases already tuned for common OCR issues
- +Covers standard builds for the game
- -Requires Pro plan
- -Might not match your specific requirements
Custom Builds
- +Total control over every item and alias
- +Works on the Free plan
- +Perfect for niche or custom content
- -Takes more time to set up
- -Need to tune aliases, evidence rules, or import settings yourself
You can also install a template and then customize it. Installed templates are fully editable, so you can add, remove, or change anything after installation. There is also a third source: the web Community Marketplace, where guilds publish and discover packs built by other communities. Browse it on the Community Marketplace in the dashboard, then install any listing into your own server (Pro plan required to install) with an optional custom pack name. Publishing a pack to the marketplace is the web counterpart of /pack share, and installing from it mirrors /pack import-code. Some listings also carry a Curated badge: GearCheck staff flag a pack as Curated when they approve it, so seeing that badge on a marketplace card or its preview means staff already checked it as a solid starting point.
Related
- How Scoring Works to understand how builds get scored
- Command Reference for all the /pack commands
- Troubleshooting if scores are not looking right