Skip to main content

Translation Specialist

Machine translation across 200+ languages (NLLB-200).

Not yet implemented

The translate-public endpoint is registered in the catalog, but the served handler currently returns {"status": "not_implemented"} — the NLLB backend is not wired up yet. Treat translation as roadmap, not a working endpoint. For translation today, use a multilingual LLM (e.g. Qwen 2.5 7B) or import a translation model from HuggingFace with task_type="translation".

Overview​

  • Endpoint: translate-public
  • Intended model: facebook/nllb-200-distilled-600M
  • Capability: text.translate
  • Backend: specialist (GPU)
  • Status: not yet implemented (see caution above)

Intended contract (roadmap)​

When implemented, the endpoint is designed to translate a text between languages identified by NLLB-200 language codes (e.g. eng_Latn, spa_Latn, fra_Latn):

# Intended shape — NOT functional today
client.endpoints.infer(
    endpoint_id="translate-public",
    input_data={"text": "Hello, how are you?", "source_language": "eng_Latn", "target_language": "spa_Latn"},
)

Until then, translate with an LLM:

from colabhive import ColabHive

client = ColabHive(api_key="hive_...", account_id="YOUR_ACCOUNT_ID")
result = client.endpoints.infer(
    endpoint_id="qwen-2.5-7b-instruct-public",
    input_data={"messages": [{"role": "user", "content": "Translate to Spanish: Hello, how are you?"}]},
    max_tokens=100,
)

Next steps​

Authors: José Luis Minich, Maximiliano Lucius.