Project
Oremi Andika
Empowering seamless transcription with cutting-edge STT (Speech-to-Text) technology, revolutionizing interaction through accurate speech recognition.
153 commits 1 contributor Updated Nov 2026
Overview
Oremi Andika serves as the foundational pillar of the Oremi project, introducing cutting-edge STT (Speech-to-Text) capabilities to enable seamless transcription of spoken language across multiple engines.
Built upon a unified speech recognition architecture, this module empowers the Oremi personal assistant to accurately transcribe spoken words into text using multiple backends (Vosk and Mega-ASR), facilitating effortless communication and interaction.
Drawing inspiration from the Swahili word for 'transcribe', Andika encapsulates the essence of transforming spoken language into written text. By embodying this linguistic concept, Oremi Andika not only facilitates efficient communication but also bridges the gap between spoken and written expression, empowering users to seamlessly interact with their personal assistant through natural speech while enjoying the convenience of textual representation.
Features
- Real-Time Streaming: Stream audio bytes directly to the server and receive recognition results incrementally with boundary overlap deduplication.
- Extensible STT Platform: Clean abstraction separating application logic from specific speech recognition backends.
- Multiple STT Backends: Native support for Vosk and Mega-ASR.
- Audio Quality Routing & LoRA Switching: Dynamic routing of noisy/degraded audio using an audio quality router and zero-latency LoRA delta switching.
- vLLM Acceleration: Ultra-fast KV-cache streaming inference with materialized LoRA checkpoints.
- Privacy-First: Operates completely offline without external cloud API dependencies.
- Docker Support: Ready-to-use Docker images for easy orchestration on CPU or GPU.
- Home Assistant Integration: Standard-compliant integration with Home Assistant's Assist voice pipeline.
Project Links
Technology
- automatic speech recognition (ASR)
- Python
- speech-to-text
- stt
Repository information
- Primary language
- Python
- Created
- July 13, 2023
- Last activity
- November 16, 2026
Activity
Daily commits on the default branch, across the whole project history.