NezhaGateNezhaGate

Immersive Translate can add custom AI translation services. Enter the address, key and model to translate web pages and PDFs with NezhaGate models.

Written against Immersive Translate 1.33

Add the translation service

  1. Open the extension's Settings → Translation Services, scroll to the bottom and click Add Custom Translation Service.
  2. Choose Custom AI (not Custom API, which speaks another protocol).
  3. Fill in the fields as below and save.
FieldValue
Custom API interface addresshttps://nezhagate.com/v1/chat/completions
APIKEYYour NezhaGate API key
ModelClick Set up more models, enter model ids in Custom Model List (comma-separated), then select one.
Max requests per second3–5 is a good start (default 5). Translating a page sends many requests at once, and too high a value can trigger rate limits.

Which model to use

Page translation is many short requests, so a model that answers quickly without long thinking suits it best; a thinking model thinks before every paragraph, which is slower and spends more output tokens. For example:

gemini-3.6-flash-low · claude-sonnet-4-6

Claude (optional)

You can also use the extension's built-in Claude service with your own key, with the API address set to https://nezhagate.com/anthropic/v1/messages.

Test

Click Verify service; once it shows Successful, translate any page to try it.

Troubleshooting

429, or a translation that fails halfway

Lower Max requests per second, for example to 3.

401 / 404

401 means the key is wrong; 404 means the address is wrong. It should be https://nezhagate.com/v1/chat/completions.

The service test keeps failing

Make sure the type is Custom AI, not Custom API, and that the model id is in the custom model list and selected.

The userscript version cannot send requests

The script header must declare the NezhaGate domain with // @connect.

More integrations