Machine Translation Services

neokapi integrates with external machine translation (MT) services as an alternative to LLM-based translation.

Supported Services

Service	Description
DeepL	High-quality neural machine translation
Google Translate	Google Cloud Translation API
Microsoft Translator	Azure Cognitive Services
MyMemory	Free translation memory API
ModernMT	Adaptive machine translation

Configuration

Each MT service is exposed as its own <provider>-translate tool, so the provider is fixed by the tool you run — deepl-translate uses DeepL, google-translate uses Google, and so on. Non-secret options (region, project, formality) are set as step config in a flow definition, which is safe to commit:

steps:
  - tool: microsoft-translate
    config:
      region: westus2

The API key is never read from the .kapi/kapi.yaml recipe. Supply it the same way as for AI translation (if more than one is present, an inline --api-key wins, then a saved --credential, then the environment variable):

1. Saved credential — store the key once and reference it by name:

   kapi credentials add my-deepl --provider deepl --api-key …
   kapi deepl-translate -i input.html --target-lang fr --credential my-deepl

2. Inline flag — --api-key … for a one-off run.

3. Per-provider environment variable — used when neither --credential nor --api-key is given (see Environment variables).

Environment variables

When no --credential and no --api-key are supplied, the key falls back to the standard environment variable for the tool's provider. Where two are listed, the first non-empty one wins:

Variable	Provider
DEEPL_API_KEY	DeepL
GOOGLE_TRANSLATE_API_KEY (then GOOGLE_API_KEY)	Google Translate
MICROSOFT_TRANSLATOR_KEY (then AZURE_TRANSLATOR_KEY)	Microsoft Translator
MODERNMT_API_KEY	ModernMT
MYMEMORY_API_KEY	MyMemory (optional)

MyMemory works without a key. For the full set of per-provider parameters, see the generated Tool Reference.

Usage

Each MT service is exposed as its own tool, named <provider>-translate — deepl-translate, google-translate, microsoft-translate, modernmt-translate, mymemory-translate. They compose into flows like any other stage. A typical production flow chains TM leverage, an MT pass, and AI refinement:

steps:
  - tool: tm-leverage
  - tool: deepl-translate
  - tool: ai-review
  - tool: qa-check

Switching providers is a configuration change — replace deepl-translate with google-translate (or ai-translate) and the rest of the flow is unchanged.

Comparison with AI Translation

Feature	MT Services	AI Translation
Speed	Faster for bulk	Slower per segment
Cost	Per-character pricing	Per-token pricing
Quality	Consistent	Context-aware
Customization	Limited	Full prompt control
Offline	No	Yes (with Ollama)

Both approaches can be combined in a flow: use MT services for bulk translation and AI for quality review.