Overview of the local assistant stack—workers, orchestration, and how they fit together.

Codmir Local AssistantCopy Anchor Link

This page summarizes the local assistant stack:

• Wakeword Listener: detects “codmir” using offline ASR
• Orchestrator: coordinates a session with screen capture, audio record, transcription, and optional desktop control
• Screen Monitor: low-FPS screenshots and blue-corner overlay
• Audio Recorder: records mic to WAV (16 kHz)
• Transcriber: Vosk-based offline speech-to-text
• Desktop Automation: controls the desktop via X11 tools
• TTS Worker: optional local speech synthesis (espeak-ng)

Run All (Docker Compose)Copy Anchor Link

File: docker/compose.codmir.yml

xhost +local:root   # X11 only
docker compose -f docker/compose.codmir.yml up --build -d

Default Intent: “open our app”Copy Anchor Link

Desktop automation opens http://localhost:3000 when asked to “open our app”. Configure via DEFAULT_APP_URL.

Security & PrivacyCopy Anchor Link

• Screen overlay indicates when capture is active.
• All processing can run offline; no cloud TTS is required (documented alternative only).