Glossary
The organization-wide terminology dictionary the AI consults on every translation — one per organization, applied across all namespaces.
The glossary is a reference dictionary for your brand. It tells the AI three kinds of things:
- Don't translate this. Brand names, product names, identifiers.
- Translate this specific way. Industry jargon with a fixed equivalent in each language.
- Watch the casing. Terms where case carries meaning.
There is one glossary per organization, and it applies to AI translation in every namespace — batch auto-translate and single-cell suggestions consult it automatically, and manual translation has an on-by-default Use glossary toggle. You set it up once and every AI call across your organization respects it.
Structure
The glossary itself has:
- A default language — the language you write glossary entries in. Usually your product's source language. Changeable later from the glossary's settings dialog.
- A set of enabled languages — the languages entries can carry translations in. Independent of any namespace's language list.
- The items — the entries themselves.
A glossary item has
| Field | What it does |
|---|---|
| Value | The term being defined, in the glossary's default language (e.g. Norcube). |
| Type | Word class: noun, verb, adjective, adverb, pronoun, preposition, conjunction, interjection, phrase, idiom, or other. Helps the AI inflect the term correctly in grammatical languages. |
| Context | Optional disambiguation note when the term has multiple meanings. |
| Translatable | Off = the term must pass through verbatim in every language (the item list notes: "Items marked as fixed are kept verbatim across all languages."). On = the term may be translated, using your per-language values where you've provided them. |
| Case sensitive | Whether matching should respect the exact casing. Leave off unless the same word means different things in different cases. |
| Translations | For translatable items: one value per enabled language, managed on the item's detail page. |
The item list shows each entry with its type badge, its flags, and a Complete translations count so you can see which entries still lack values in your enabled languages.
How items influence the AI
On every AI translation, LangSync selects the glossary entries relevant to the source → target language pair and passes them to the model along with your text:
- Non-translatable items instruct the AI to keep the term verbatim.
- Translatable items with a target-language value instruct the AI to use exactly that value.
- Translatable items without a value in the target language don't constrain that translation — the AI translates the term normally. Missing values are never "enforced as empty".
When to mark not translatable
- Brand and product names —
Norcube,LangSync, your product line. - Technical identifiers appearing in copy —
API key,webhook, command names. - Trademarks you're not allowed to alter.
Behaviour and edge cases
- One per organization. Creating a second is rejected. If you need two genuinely separate vocabularies (two brands with conflicting terminology), use two organizations.
- Deleting an item takes effect on the next AI call — the constraint simply stops being applied. Existing stored translations are untouched; nothing is retroactively re-translated.
- Adding or fixing an item after a big batch also doesn't re-run anything — set the glossary up before running large auto-translate batches.
- Glossary changes are recorded in the audit log
(
glossary.created,glossary.language_attached,glossary_item.created/updated/deleted).
Related
- Use the glossary — setup walkthrough.
- Custom prompts — the style counterpart to the glossary's terminology.