Setup & Installation
What This Skill Does
A unified API for calling Gemini models and agents from Python and TypeScript. Replaces the older generateContent approach with server-managed conversation state, background task execution, streaming, and access to built-in agents like Deep Research.
Server-side state management via previous_interaction_id removes the need to send full conversation history on every request, cutting payload size and simplifying chat logic.
When to use it
- Building a multi-turn chatbot without manually tracking conversation history
- Running a Deep Research agent in the background and polling for results
- Streaming a long Gemini response token-by-token to a user interface
- Calling custom functions from within a Gemini conversation using tool definitions
- Migrating existing generateContent code to use previous_interaction_id for stateful sessions