Yes, the Gemini Live API supports structured output. This is confirmed by the provider comparison table, where Google Gemini is marked as supporting structured output, unlike some other providers. The LiveKit Agents framework includes utilities for converting Python types and dictionaries into Gemini's schema format, with robust handling of JSON schema transformation, optional fields, and nested objects. There are also practical usage examples and comprehensive tests demonstrating structured output for various complex scenarios, like controlling TTS behavior and handling detailed response content.
For details on implementation and examples using structured output, refer to the relevant sections in your documentation and practical usage guides.
Sources:
Large language model (LLM) integrations | LiveKit Docs |
examples/voice_agents/structured_output.py |
LiveKit | Recipes and examples for voice AI and more. | LiveKit Docs