Context Sensitive is wxrks's automatic post-editing (APE) step: it takes the raw material wxrks already has for a segment and produces the final translation. Context Sensitive Translate is the action that runs it on a single segment — it blends your Translation Memories, Machine Translation output, Glossaries, and your Sous Chef Instructions — the tone, audience, and style guidance attached to your AI agent — into one suggested translation, tailored to the segment you're actually working on rather than a generic machine translation. It runs automatically the moment you open an empty segment, and you can also trigger it yourself at any time.

💡 Who is this for? This guide is for any user with Editor access who needs to use Context Sensitive Translate to auto-translate empty segments and refresh AI suggestions while translating or reviewing within the wxrks platform. The Context Sensitive Priority section is for Account Admins and Project Managers who configure which sources feed the AI and in what order.

Context Sensitive Translate doesn't just pick one source and use it as-is — it consults every source that actually has something to say about that segment, and produces a single suggestion informed by all of them:

The suggestion is not produced by a single AI request. wxrks runs the translation as a sequence of passes, ordered by the priority you configure in Translation Settings:

This is why the panel describes higher items as being "applied later": the source at the top of the list is the last one applied, and therefore the one that wins. With the default order, Glossary sits at the top, so approved terminology is enforced as the closing step of every translation.

ℹ️ Note: A source that has nothing to contribute to a specific segment is skipped for that segment — a glossary with no matching term, or a segment with no TM hits, simply doesn't produce a pass. If no source applies at all, wxrks falls back to a plain AI translation of the source text.

You don't have to guess which of these shaped a given suggestion — turning on Translation Reasoning on the card shows exactly which factors influenced that specific translation.

Context Sensitive Priority lets you decide which of the four sources feed Context Sensitive Translate, and which one gets the final word. It is the setting to reach for when a glossary term, a TM match, or a style instruction isn't being respected as consistently as you expect.

Who can change it: Account Admins and Project Managers. The Translation Settings tab on an Organizational Unit additionally requires the Organizational Unit Update permission, granted in People > Users > [user] > Permissions under ORGANIZATIONAL UNIT.

You can set the priority at two levels:

The panel lives in the Auto Translate From card on both screens, and works identically at either level.

The panel lists the four sources — it calls them providers — in priority order, numbered 1° through 4°:

That is also the default order for every Organizational Unit and project — Glossary first, MT last.

Use the up and down arrow buttons on each row. Moving a source up increases its priority, which means it is applied later in the sequence. The arrow is disabled on the row that's already at the top or bottom.

Each move saves immediately — there is no separate Save button on this panel.

⚠️ Warning: Moving any source below Machine Translation (MT) triggers the MT prioritization warning modal: "Machine Translation is recommended as the first execution step because using MT context first usually improves quality. In this priority list, that means MT should stay at the bottom. Continue anyway?" Select OK to apply the new order, or Cancel to leave the list as it was. MT is most useful as raw material for the later passes to correct, not as the final word — keep it at the bottom unless you have a specific reason not to.

Each row has its own switch. Turning a source off removes it from Context Sensitive Translate entirely for that scope — it stops being consulted, not just deprioritized. Turning off Translation Memory (TM) means TM matches no longer inform the suggestion; turning off Sous Chef Instructions means your agent's tone and style guidance is no longer applied.

The row dims to show it's inactive but keeps its position, so you can re-enable it later without rebuilding the order.

You add a term to the glossary, run Context Sensitive Translate again, and the term is applied on some segments but not others.

💡 Tip: If terminology is what your client audits, keep Glossary at 1°. If matching previously approved wording matters more than terminology — a large continuous project with a mature TM, for example — move Translation Memory (TM) to 1° instead, and accept that glossary terms will be enforced in the combined pass rather than as the closing step.

With Enable Context Sensitive Translate on in your Editor Settings panel, opening a segment that has no target text yet triggers an AI translation automatically — you don't need to click anything. This also applies to a segment that only looks empty because it was created by merging two segments together; wxrks still treats it as empty and auto-translates it the same way.

If your account, Organizational Unit, or your own Editor preferences don't already have Augmented Actions enabled, opening an empty segment won't trigger anything — see wxrks Augmented Actions for the full three-level (Account, Organizational Unit, User) enablement walkthrough.

Two ways to get a fresh (or first) Context Sensitive Translate suggestion without waiting for the automatic trigger:

If a segment already has a translation, click into it and select the icon-only Regenerate button on the Context Sensitive Translation card (or Translate, if the card doesn't show a suggestion yet) to have the AI reconsider it — worth doing whenever a TM, glossary, MT, or Context Sensitive Priority setting has changed since the segment was last translated, since Context Sensitive Translate otherwise keeps working from what it saw at the time.

Selecting any segment reveals an Augmented Actions button in its bottom-right corner. Its dropdown menu includes Context Sensitive Translate — selecting it runs the exact same underlying translation as the automatic and Regenerate cases above, just triggered manually on whichever segment you have selected, translated or not.

For the other actions in this menu — Alternative Suggestion, Proofread, Check Smells, Fix Tags — see wxrks Augmented Actions.

Once a segment has a Context Sensitive Translation, a Translation reasoning toggle appears directly on that card. Turning it on shows the rationale behind the suggestion as a short set of categorized bullets — for example Glossary enforcement for a specific glossary-driven word choice, or Fluency adaptation for how the wording was adjusted to read naturally in the target language. Other categories appear when they apply — TM Reuse, Audience/Tone, Terminology Choice, Style Adjustment, Localization, Error Correction, Formatting — only the factors that actually influenced that translation are shown.

Translation reasoning is off by default and, once you turn it on, stays on for you across segments and sessions. Turning it on for a segment that already has a translation refreshes that translation immediately so the reasoning appears. See Translation Reasoning for the full walkthrough of every category and why you'd turn it on.

💡 Tip: Translation reasoning is the fastest way to confirm a priority change did what you wanted. If you moved Glossary to 1° and the reasoning now shows Glossary enforcement on segments where it previously didn't, the new order is taking effect.

Augmented Term Lookup is a separate, independent feature — it runs on its own as soon as you open a segment containing an inflected form of a glossary term, surfacing suggested variants in the Editor's GLOSSARY VARIANTS panel. It never feeds into or changes Context Sensitive Translate's own output, and nothing it suggests is applied automatically. It is also unaffected by Context Sensitive Priority — turning Glossary off there does not turn off Glossary Variants.

In practice this means a segment can show two different forms of the same word at once: Context Sensitive Translate might produce a literal, unadapted glossary substitution, while Glossary Variants separately offers a better-inflected form as its own suggestion. If you want the inflected form instead of what Context Sensitive Translate produced, apply the variant yourself from that panel — it's never applied for you.

Opening an empty segment doesn't trigger anything. Confirm Augmented Actions is enabled at all three levels — Account, Organizational Unit, and your own Editor Settings panel's Enable Context Sensitive Translate toggle. See wxrks Augmented Actions for where each toggle lives.

The card shows a TM badge instead of a fresh AI translation. This is expected, not a bug — the project has a strong enough Translation Memory match that wxrks reused it verbatim rather than generating a new suggestion. Select Regenerate if you specifically want a fresh AI translation instead.

A glossary term is applied on some segments but not others. Open Context Sensitive Priority in the project's Translation Settings and confirm Glossary is enabled and sitting at 1°, so it runs as the final pass rather than competing inside the combined pass. Then Regenerate the affected segments — existing translations are not re-run automatically.

I can't see the Context Sensitive Priority panel. The caret beside the Context Sensitive toggle is disabled while that toggle is off. Turn Context Sensitive on first.

My Sous Chef Instructions are being ignored. Check that Sous Chef Instructions is switched on in Context Sensitive Priority — when it's off, your agent's tone and style guidance is not consulted at all, regardless of what the agent contains.

I changed the priority but nothing changed in the Editor. Priority is read at translation time, so it applies to the next run. Existing translations keep whatever they were produced with until you Regenerate them.

Still stuck? Search the wxrks Community forum or contact [email protected].