Roadmap Inquiry: General Availability for RAG/File Search in Multimodal Live API?

Gemini API
1

Hi Team,

I’ve been experimenting with the Multimodal Live API and it’s impressive for low-latency voice interactions. However, a major blocker for our production use case is the lack of native RAG (File Search) support within the Live session.

While File Search (Managed RAG) is available for the standard Gemini API, information is needed on when this will be officially supported for the Live API, specifically for models like Gemini 2.5 Flash or the newer Gemini 3 previews.

Currently, custom workarounds (like manual tool calls to a separate RAG pipeline) must be used. This introduces significant latency that defeats the purpose of a “Live” experience.

Questions:

  1. Is there a target date for the General Availability (GA) of RAG support within the Multimodal Live API?

  2. Will future support include the Vertex AI RAG Engine as a native context store for Live sessions?

  3. Are there recommended best practices for grounding Live API responses in private datasets in the interim?

Any roadmap updates are appreciated!