eloqnt translate

Context enrichment

Your messages hold a set of strings, but they typically lack context on how they’re used:

{
  "XPruqs": "Order",
  ...
}

Without additional information, it’s hard to tell:

1. Where this string is used
2. What kind of UI component it’s used for
3. Whether "Order" is a noun or a verb
4. How "Order" is translated elsewhere in your app
5. Which voice and tone to use

If you have srcPath configured, the CLI will analyze your code and extract a simplified source outline that keeps the relevant parts around each string:

⋮
│ export default function OrderDetailPage({params}: Props) {
│   const t = useExtracted();
│   return (
│     <main>
█       <h1>{t('Order')}</h1>
⋮

This provides clear answers to the first three questions raised above:

1. It’s used on the order detail page
2. It’s the heading of that page
3. "Order" is the noun

Styleguides

Next, all applicable styleguides are attached to the request:

• Your global styleguide, defining the voice and terminology of your project
• The styleguide of your target locale, holding tone and a glossary specific to the language

# Styleguide

## Voice and tone

- Friendly and concise.
- Address the user as "you", and prefer active voice.

...

# `es` styleguide

## Tone

- Use informal "tú", never formal "usted".

## Glossary

- "order" (a purchase): "pedido"

...

With this, we can answer the remaining questions from above:

4. "Pedido" is the expected translation
5. Informal tone should be used

Translation

With all context now available, the following is passed to an LLM:

1. Your source strings
2. Source outlines
3. Styleguides
4. CLDR data for target locales
5. General translation best practices

Depending on your plan, the actual translation work either happens on our hosted backend or is delegated to your own model.

Post-checks

Once the LLM returns, every translation has to pass various quality checks, like:

1. Are there syntactical errors?
2. Do arguments like {name} match the source?
3. Are the right plural cases defined?

If a translation doesn’t meet the bar yet, then it’s fixed:

1. Minor mechanical slips are repaired directly (e.g. removal of superfluous plural cases)
2. If the issue requires more work, then the translation step is retried with the findings attached, and with increased reasoning effort

Persistence

Finally, the translations are written to the messages of your target locales:

 {
+  "XPruqs": "Pedido",
   ...
 }