Offline Client

File: main_logic/omni_offline_client.py

The OmniOfflineClient provides text-based LLM conversation as a fallback when the Realtime API is unavailable.

When it's used

• When the selected provider doesn't support Realtime API
• When using local LLM deployments (Ollama, etc.)
• When voice input is disabled and text-only mode is preferred

Capabilities

• Text-in, text-out conversation
• Compatible with any OpenAI-compatible API endpoint
• Uses LangChain for LLM integration
• Supports conversation history and system prompts

Differences from Realtime Client

Feature	Realtime Client	Offline Client
Audio I/O	Native	Requires separate STT/TTS
Streaming	WebSocket bidirectional	HTTP streaming
Multi-modal	Native (audio + images)	Text only
Latency	Lower (persistent connection)	Higher (per-request)
Provider support	Limited (Realtime API required)	Any OpenAI-compatible