Deep Research Streaming (Interactions)

Stream real-time progress (thought summaries, text, generated images) from a Deep Research task instead of waiting for the final report. Set thinking_summaries="auto" to receive intermediate reasoning during streaming. Background execution is required for agents and is enabled automatically.

Code

1import asyncio
2
3from kern.agent import Agent
4from kern.models.google import GeminiInteractions
5
6agent = Agent(
7    model=GeminiInteractions(
8        agent="deep-research-preview-04-2026",
9        thinking_summaries="auto",
10    ),
11    markdown=True,
12)
13
14if __name__ == "__main__":
15    agent.print_response(
16        "Research the history and impact of Google TPUs.",
17        stream=True,
18    )
19
20    asyncio.run(
21        agent.aprint_response(
22            "Research the current state of open-source LLM inference engines.",
23            stream=True,
24        )
25    )

Usage

Set up your virtual environment

1uv venv --python 3.12
2source .venv/bin/activate

1uv venv --python 3.12
2.venv\Scripts\activate

Set your API key

1export GOOGLE_API_KEY=xxx

Install dependencies

1uv pip install -U "google-genai>=2.0" kern-ai

Run Agent

1python cookbook/90_models/google/gemini_interactions/deep_research_streaming.py