About
Tokativ Engine is a multilingual live broadcast translation backend (LiveKit + STT + Translation + TTS). Use the links above to test individual services and the live broadcast flow.
Server time
Feb. 12, 2026, 2:10 p.m.
Notes
  • Whisper STT runs locally via faster-whisper when enabled.
  • NCAIR HuggingFace ASR for yo/ha/ig is controlled by USE_NATLAS.
  • If a model is gated, you must set HUGGINGFACE_TOKEN.
  • ffmpeg must be available for many audio formats (both STT and decoding).
Checks
Status Check Detail
OK Local storage /tmp/tokativ_storage
OK Database django.db.backends.postgresql
OK Redis redis://redis:6379/0
OK LiveKit URL configured wss://lk.tokativ.wype.app
OK DeepSeek API key configured set
OK STT general provider whisper
OK NATLAS (NCAIR HF) enabled USE_NATLAS=1 uses NCAIR models for yo/ha/ig
OK HuggingFace token present set
OK ffmpeg on PATH /usr/bin/ffmpeg
OK Python import: transformers
OK Python import: torch
OK Python import: torchaudio
OK Python import: faster_whisper
This page runs lightweight checks only (no heavy model loads). If a check fails, open /api/services/ to reproduce and see full error messages.