Meta unveils Llama Assist API for in-app personalization and conversational overlays

AI · 5 min read

Llama Assist offers a lightweight runtime for conversational features that can run either on-device for sensitive tasks or in the cloud for more complex reasoning. The API supports session personalization based on local preferences and recent interactions without centralized profiling.

Developers can embed conversational overlays that appear contextually within apps — for example, help tips in productivity apps or contextual NPC dialogues in games. Meta provided templates for onboarding assistants, conversion funnels, and contextual help systems.

Meta emphasized user agency and privacy controls: users can view and export the local preference store and opt out of cloud personalization. Early partners reported faster prototype cycles for contextual assistants but cautioned about moderation requirements for open conversational outputs.