Adding a new language to Argilla

If you want to add a new language to Argilla you need to go to two places:

1. Add a new translation specification in the folder: argilla-frontend/translation E.g. for Korean with Code ko add a ko.js file by coping the en.js file. The text values need to be translated:

   export default {
       multi_label_selection: "다중 라벨",
       ranking: "순위",
       label_selection: "라벨",
       span: "범위",
       text: "텍스트",
       ...
   

2. Then update the i18n Nuxt: argilla-frontend/nuxt.config.ts

  i18n: {
    locales: [
      {
        code: "en",
        file: "en.js",
      },
      ...
      {
        code: "ko",
        file: "ko.js",
      },
    ],

How to test it

1. Start a local instance of Argilla, easiest by just using the docker recipe here. It will give you a backend API for the frontend.
2. Compile a new version of the frontend. Check this guide. This is basically:
   • git clone https://github.com/argilla-io/argilla
   • cd argilla-frontend
   • Install the dependencies: npm i
   • Build the new frontend with the updates: npm run build
   • Serve the UI via npm run start. You can reach it under localhost:3000 by default.
   • Check the translations.
3. Deploy a small test dataset to test the translation on a dataset too:

   import argilla as rg
   
   client_local = rg.Argilla(api_url="http://localhost:6900/", api_key="argilla.apikey")
   
   sample_questions = [
       rg.SpanQuestion(
           name="question1",
           field="text",
           labels={
               "PERSON": "Person",
               "ORG": "Organization",
               "LOC": "Location",
               "MISC": "Miscellaneous"
           },  # or ["PERSON", "ORG", "LOC", "MISC"]
           title="Select the entities in the text",
           description="Select the entities in the text",
           required=True,
           allow_overlapping=False,
       ),
       rg.LabelQuestion(
           name="question2",
           labels={"YES": "Yes", "NO": "No"},  # or ["YES", "NO"]
           title="Is the answer relevant to the given prompt?",
           description="Choose the option that applies.",
           required=True,
       ),
       rg.MultiLabelQuestion(
           name="question3",
           labels={
               "hate": "Hate speech",
               "sexual": "Sexual content",
               "violent": "Violent content",
               "pii": "Personal information",
               "untruthful": "False information",
               "not_english": "Not English",
               "inappropriate": "Inappropriate content"
           },  # or ["hate", "sexual", "violent", "pii", "untruthful", "not_english", "inappropriate"]
           title="Does the response contain any of the following?",
           description="Select all applicable options.",
           required=True,
           visible_labels=3,
           labels_order="natural"
       ),
       rg.RankingQuestion(
           name="question4",
           values={
               "reply-1": "Answer 1",
               "reply-2": "Answer 2",
               "reply-3": "Answer 3"
           },  # or ["reply-1", "reply-2", "reply-3"]
           title="Rank the answers by your preference",
           description="1 = best, 3 = worst. Equal ratings are allowed.",
           required=True,
       ),
       rg.RatingQuestion(
           name="question5",
           values=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
           title="How satisfied are you with the answer?",
           description="1 = very dissatisfied, 10 = very satisfied",
           required=True,
       ),
       rg.TextQuestion(
           name="question6",
           title="Please provide your feedback on the answer",
           description="Please provide your feedback on the answer",
           required=True,
           use_markdown=True
       )
   ]
   
   sample_fields = [
       rg.ChatField(
           name="chat",
           title="Previous conversation with the customer",
           use_markdown=True,
           required=True,
           description="Dialog between AI & customer up to the last question",
       ),
       rg.TextField(
           name="text",
           title="Customer's question",
           use_markdown=False,
           required=True,
           description="This is a question from the customer",
       ),
       rg.ImageField(
           name="image",
           title="Image related to the question",
           required=True,
           description="Image sent by the customer",
       ),
   ]
   
   # Create a new dataset with the same settings as the original
   settings = rg.Settings(
       fields=sample_fields,
       questions=sample_questions,
   )
   new_dataset = rg.Dataset(
       name="demo_dataset",
       workspace="default",
       settings=settings,
       client=client_local,
   )
   new_dataset.create()
   
   def fix_record():
       return rg.Record(
           fields={
               "chat": [
                   {"role": "user", "content": "What is Argilla?"},
                   {"role": "assistant", "content": "Argilla is a collaboration tool for AI engineers and domain experts to build high-quality datasets"},
               ],
               "image": "https://images.unsplash.com/photo-1523567353-71ea31cb9f73?w=900&auto=format&fit=crop&q=60&ixlib=rb-4.0.3&ixid=M3wxMjA3fDB8MHxzZWFyY2h8MTJ8fGNvcmdpfGVufDB8fDB8fHww",
               "text": "Which town has a greater population as of the 2010 census, Minden, Nevada or Gardnerville, Nevada?",
           },
       )
   
   new_records = [fix_record() for _ in range(10)]
   new_dataset.records.log(new_records)
   

4. Test if your translation also works with the dataset and in the dataset settings.