Skip to main content
Entri’s AI Translation turns hours of manual translation work into seconds. Rather than calling a raw machine translation API, Entri sends your strings to leading large language models with rich context — key names, descriptions, your glossary, and translation memory — so the output is consistent with your existing translations.

​
Supported Models

Model IDProviderDisplay Name
claude-haiku-4-5AnthropicClaude Haiku 4.5
claude-sonnet-4-6AnthropicClaude Sonnet 4.6
claude-opus-4-6AnthropicClaude Opus 4.6
gpt-4o-miniOpenAIGPT-4o mini
gpt-4oOpenAIGPT-4o
gpt-5OpenAIGPT-5
gpt-5-miniOpenAIGPT-5 mini
gpt-5.2OpenAIGPT-5.2
Only models whose provider has an API key configured in your organization settings will be available.

​
Default Model

Each project can specify a default model in Project Settings → General → AI Settings. The selected model is used for all single and batch translations in that project. When no project-level model is set, Entri falls back to the provider specified by the DEFAULT_AI_PROVIDER environment variable, defaulting to claude-sonnet-4-6. You can override the model on a per-request basis by passing a model field in the API.

​
How Context Awareness Works

Plain machine translation treats every string in isolation. Entri’s AI considers:
  • Key name — a key named checkout.confirm_button signals a UI button in a checkout flow, not a generic confirmation message.
  • Description — optional notes you attach to a key explaining its purpose and where it appears.
  • Glossary — terms you have defined are injected so the AI respects your preferred translations and avoids translating brand names or technical identifiers.
  • Translation memory — previously saved AI translations for similar strings guide the AI toward consistency.
The result is AI output that fits the context of your product rather than a literal word-for-word rendering.

​
Confidence Scoring

Every AI-generated translation includes a confidence score from 0 to 100%. The score reflects how certain the model is about the accuracy and appropriateness of the translation.
Confidence scores are stored on the translation record but are not currently displayed in the editor UI. They are available via the API for custom workflows.

​
Alternative Suggestions

The AI generates alternative translations alongside the primary suggestion. These are returned in the API response but are not currently displayed in the editor suggestion panel. Future editor updates will surface these alternatives directly in the UI.

​
Tone and Style Instructions

Every project supports a Voice & Style configuration. Write plain-language instructions such as:
Translate in an approachable, friendly tone. Avoid technical jargon. Use informal “you” (tu) rather than formal “vous” in French. Never translate the brand name “Entri”.
The AI reads these instructions on every request and adjusts its output accordingly. Style instructions work alongside the glossary — the glossary handles specific terms while style instructions shape the overall register and voice.

​
Translating Strings

​
Translate a Single String

In the Translation Editor, open any cell and click Suggest with AI. The editor populates with the top suggestion. Accept it or edit the value before saving.

​
Batch Translation

1

Open Batch Translate

From the project overview, click Translate → Batch Translate.
2

Choose scope

Select the target language(s). The batch job translates all untranslated strings for the selected language.
3

Review settings

The model configured in Project Settings → General → AI Settings is used automatically. Verify that your glossary and style instructions are applied.
4

Run

Click Translate. Entri processes strings in parallel and shows a real-time progress bar. Results are saved with Translated status so they enter the normal review workflow.
Run batch translation at the start of a sprint to pre-populate translations, then let human translators focus on review rather than translating from scratch.

​
Streaming Batch Translation (API)

For programmatic use, Entri offers both a synchronous batch endpoint and a streaming batch workflow via Server-Sent Events (SSE):
  1. Start a job — POST /api/projects/:projectId/translate/batch/start to create a job and get a jobId.
  2. Stream progress — GET /api/projects/:projectId/translate/batch/stream/:jobId for real-time SSE updates.
  3. Cancel — POST /api/projects/:projectId/translate/batch/cancel/:jobId to stop a running job.
The synchronous endpoint POST /api/projects/:projectId/translate/batch is also available for smaller batches.

​
Glossary and Translation Memory Integration

AI translation is strongest when combined with your organization’s knowledge:
  • Glossary — terms defined in your Glossary are automatically included in every AI prompt. Terms marked “do not translate” are passed through verbatim.
  • Translation Memory — the AI receives examples of previously saved translations for strings with high similarity, anchoring its vocabulary to what your team has already validated.
These integrations happen automatically — no extra configuration is needed once your glossary and translation memory are populated.