translateit

A simple translate api that proxies to openai written by Golang.

Run

1. Setup environment variables:

• OPENAI_KEY, key of openai, required.
• OPENAI_MODEL, model of openai, optional, default to gpt-3.5-turbo.
• HTTP_PORT, port of the server, optional, default to 8080.

2. Build by make build.
3. Start server by ./translate. Try it with:

curl -X POST \
-d '{"content": "Hi, how are you?", "lang": "Chinese", "model": "gpt-3.5-turbo"}' \
http://localhost:8080/translate

4. translate-cli is a way to test translate endpoints, start translate server before using the cli tool. Try it like ./translate-cli -c "content" -l "Chinese" -m "gpt-3.5-turbo", -l and -m are optional. The endpoint is http://localhost:8080 by default, customize it by setting TRANSLATE_ENDPOINT env var.

CORS is enabled by default, set up ALLOW_ORIGINS env var to control the origin otherwise all is allowed.

Endpoints

There are 3 endpoints for now:

1. POST /translate, translate the content and return the simplified response with reply and reason only. Ask is the request and Answer is the response.

   // Ask the request of /translate
   type Ask struct {
   	Content string `json:"content"`
   	Lang    string `json:"lang,omitempty"`  //optional, default to English
   	Model   string `json:"model,omitempty"` //optional, default to gpt-3.5-turbo
   }
   
   // Answer the response of Ask
   type Answer struct {
   	Reply  string `json:"reply"`
   	Reason string `json:"reason"`
   }

2. POST /translate/openai, translate the content and return the original openai response for debugging purpose. The request is still Ask. The response is what described in the openai doc.
3. GET /translate/ping, a server healthy check api.

Docker

Use the image build by github action or build the image yourself.

• build: docker build -t translate .
• setup OPENAI_KEY, OPENAI_MODEL in an env file
• start: docker run --env-file env -p 8081:8080 -d translate

Additional

Use the companion translateit-btn with the api.