Capabilities
Realtime / Live API
Bidirectional realtime audio + video streaming
Read this when
- Building voice agents
- Building low-latency multimodal interfaces
Compatibility
| Provider | API | pi-go | Notes |
|---|---|---|---|
| Anthropic | ❌ | — | |
| OpenAI | ✅ Realtime API (WebRTC / WebSocket); voice + tools | ❌ | not wired |
| Google Gemini | ✅ Live API; 24-language voice; multimodal realtime | ❌ | not wired |
| Claude CLI | ❌ | — |
Provider Documentation
pi-go Gaps
- No bidirectional streaming primitive —
EventStreamis one-way. - No audio I/O. See Multimodal Input and Multimodal Output.
- WebSocket / WebRTC transports not present anywhere in the codebase.
- This is a fundamentally different shape from
StreamText; would warrant a separateRealtimeProviderinterface.