Nanosamur.ai is an Open-Source Speech-to-Text Stack for Local and Cloud Deployment

Summary

Nanosamur.ai is an open-source speech-to-text platform designed for realtime, semi-realtime, and batch transcription. It provides a unified stack for speech processing, agentic workflows, and webhooks, and can run through an Electron application with Docker Compose on a local computer. The same stack is intended to run in the cloud, on Kubernetes, or on premises, including air-gapped environments for organizations handling sensitive data. The project currently supports Whisper, Qwen ASR, Nemotron ASR, and Parakeet TDT, with additional models being added. Its built-in observability stack is intended to support deployments designed to scale. The author is also adding KServe and Triton integration for batch transcription so that MLOps infrastructure can be abstracted. The repository provides a Docker Compose starter setup and links to the services used by the platform. The voice-AI services are written in Python, while the user interface, backend-for-frontend, and other services use Java, Clojure, and ClojureScript.