tanod-dev Mlpeek API
mlpeek, the local ML inference. Text embeddings (English or multilingual, 384 dimensions), multimodal embeddings (texts and images in one 768-dimension space, EmbeddingGemma 2), reranking, semantic similarity, named entities and zero-shot classification with open models (MIT / Apache-2.0) run on Tanod's own CPU: no third-party API; texts are not logged or stored. Also speech to text (audio of up to 10 minutes to text, timed segments and SRT / WebVTT captions with faster-whisper); audio is not logged or stored. Offline text translation between English and 14 languages (Argos Translate), text to speech in 6 languages (Kokoro) and an automated, heuristic prompt-injection check of text (a DeBERTa classifier) and JSON extraction from text under your JSON Schema (a small open language model); texts are not logged or stored (generated speech is kept only when `output` is "url", as a download file for 24 hours).