Overview
wxrks supports multiple AI providers to power Augmented Translation Actions, including wxrks AI (bundled by default), OpenAI, Azure OpenAI, and Google Gemini.
Gemini is offered as two separate providers you can configure independently: Google Gemini (Vertex AI), which authenticates with a Google Cloud service account, and Google Gemini (Studio), which authenticates with a single API key.
You might choose Gemini instead of the bundled wxrks AI to keep usage on your own Google Cloud billing, meet a data-residency requirement, or standardize on models your organization already uses elsewhere.
💡 Who is this for? This guide is for Account Admins who need to configure Google Gemini as a custom AI provider for wxrks Augmented Actions within the wxrks platform.
1. Before You Start
Model Management is available only to Account Admins, and only on accounts with a plan that includes advanced Model Management (currently Fusion, Quantum, and Quasar). On other plans, wxrks AI is used automatically and the Model Management page shows an upgrade prompt instead of the configuration screen described below.
To connect Gemini, you'll need one of the following, depending on which option you choose:
For Gemini Vertex AI: a Google Cloud Platform (GCP) project, a Service Account with the correct permissions, and the contents of its JSON key file.
For Gemini Studio: an API Key generated in Google AI Studio.
For setup details, refer to Google's official documentation:
2. Step 1: Enable Augmented Translation Actions
Before any AI provider can be used, Augmented Translation Actions must be turned on for your account.
Go to Settings > Translation Settings > Agents.
Turn on the Enable Augmented Translation Actions toggle and confirm in the dialog that appears.
Once enabled, the Context Sensitive Translate, Alternative Suggestion, Proofread, Review Assessment, Smells, Fix Tags, Part of Speech Classifier, Learning Terms, Sous-Chef, and Active SpellCheck actions become available account-wide. This toggle only turns AI actions on or off — it does not choose which AI provider is used. That's configured separately in Model Management (Step 2).
3. Step 2: Add a Gemini Model Setting
Go to Settings > Account Settings > Model Management.
Click + Add Model Settings.
In the Add Model Settings dialog, fill in:
Scope — Account to apply to the whole account, or Org Unit to apply to one organizational unit only.
Name — a label to help you identify this configuration later (e.g. "Gemini for QA").
Operations — which Augmented Actions should use this configuration. Leave empty to make it the default for any operation not explicitly covered by another configuration.
Provider — select Google Gemini (Vertex AI) or Google Gemini (Studio) (see Section 4 to decide which one fits your setup).
Model Name — the Gemini model to use, entered as free text (e.g.
gemini-flash-latestorgemini-2.5-pro). If left blank, wxrks falls back to a current Gemini Flash model.Max Completion Tokens — optional; leave blank to use the default.
The remaining fields depend on which Gemini provider you selected — see the next section.
When you're done, turn on Enabled and click Save. wxrks runs a live connection test against Gemini using exactly what you entered before the row is saved — if that test request fails, Save is blocked and you'll see an error naming the model or configuration that couldn't be reached, so you can fix it right away instead of discovering the problem later in the Editor.
4. Choosing Between Gemini Vertex AI and Gemini Studio
Google Gemini (Vertex AI)
Best if you already manage a GCP project and want billing/quotas tied to your own Google Cloud account. Selecting this provider reveals three additional fields:
Service Account JSON — paste the full contents of your service account's JSON key file into this field. This is a text field, not a file upload — open the downloaded
.jsonkey in a text editor and paste its contents here.GCP Project ID — your Google Cloud project ID (e.g.
my-gcp-project).GCP Region — the GCP region your Vertex AI resources run in (e.g.
us-central1).
Google Gemini (Studio)
The simpler option if you just want to get started quickly with a single key from Google AI Studio. Selecting this provider reveals one field:
API Key — the API key you generated in Google AI Studio.
5. Troubleshooting
Augmented Actions don't run at all — confirm Enable Augmented Translation Actions is turned on (Step 1). A saved Gemini configuration has no effect while this toggle is off.
Save fails with an error naming a model or configuration — wxrks tested the connection using the credentials you entered (Step 2) and the test request failed. Double-check for a malformed Service Account JSON, wrong GCP Project ID/Region, or an invalid API Key, and confirm the service account has the Vertex AI permissions described in Google's setup guide above — then click Save again.
Save fails with a rate-limit or timeout message even though your credentials look correct — the connection test makes a real request to Gemini, so a transient provider-side issue can surface here too. Wait a moment and try Save again.
Model Management doesn't appear under Account Settings — your plan may not include it yet (Fusion, Quantum, and Quasar currently do). Contact your account owner or wxrks support to check your plan.
6. Useful Resources
Helpful links from Google's documentation to guide your setup:
For more on the individual Augmented Actions this configuration powers (Context Sensitive Translate, Proofread, Smells, and others), see wxrks Augmented Actions, or the wxrks Community forum at community.wxrks.com, or contact [email protected].
Evaluating a different provider instead? See Enabling AI Integration in wxrks through OpenAI or Microsoft Azure or Configuring wxrks Augmented Actions with Azure OpenAI Service.