google__adk-docs/docs/get-started/streaming/quickstart-streaming.md

# Build a streaming agent with Python

With this quickstart, you'll learn to create a simple agent and use ADK Streaming to enable voice and video communication with it that is low-latency and bidirectional. We will install ADK, set up a basic "Google Search" agent, try running the agent with Streaming with `adk web` tool, and then explain how to build a simple asynchronous web app by yourself using ADK Streaming and [FastAPI](https://fastapi.tiangolo.com/).

**Note:** This guide assumes you have experience using a terminal in Windows, Mac, and Linux environments.

## Supported models for voice/video streaming {#supported-models}

In order to use voice/video streaming in ADK, you will need to use Gemini models that support the Live API. You can find the **model ID(s)** that supports the Gemini Live API in the documentation:

- [Google AI Studio: Gemini Live API](https://ai.google.dev/gemini-api/docs/models#live-api)
- [Vertex AI: Gemini Live API](https://cloud.google.com/vertex-ai/generative-ai/docs/live-api)

## 1. Setup Environment & Install ADK { #setup-environment-install-adk }

Create & Activate Virtual Environment (Recommended):

```bash
# Create
python -m venv .venv
# Activate (each new terminal)
# macOS/Linux: source .venv/bin/activate
# Windows CMD: .venv\Scripts\activate.bat
# Windows PowerShell: .venv\Scripts\Activate.ps1
```

Install ADK:

```bash
pip install google-adk
```

## 2. Project Structure { #project-structure }

Create the following folder structure with empty files:

```console
adk-streaming/  # Project folder
└── app/ # the web app folder
    ├── .env # Gemini API key
    └── google_search_agent/ # Agent folder
        ├── __init__.py # Python package
        └── agent.py # Agent definition
```

### agent.py

Copy-paste the following code block into the `agent.py` file.

For `model`, please double check the model ID as described earlier in the [Models section](#supported-models).

```py
from google.adk.agents import Agent
from google.adk.tools import google_search  # Import the tool

root_agent = Agent(
   # A unique name for the agent.
   name="basic_search_agent",
   # The Large Language Model (LLM) that agent will use.
   # Please fill in the latest model id that supports live from
   # https://google.github.io/adk-docs/get-started/streaming/quickstart-streaming/#supported-models
   model="...",  # for example: model="gemini-2.0-flash-live-001" or model="gemini-2.0-flash-live-preview-04-09"
   # A short description of the agent's purpose.
   description="Agent to answer questions using Google Search.",
   # Instructions to set the agent's behavior.
   instruction="You are an expert researcher. You always stick to the facts.",
   # Add google_search tool to perform grounding with Google search.
   tools=[google_search]
)
```

`agent.py` is where all your agent(s)' logic will be stored, and you must have a `root_agent` defined.

Notice how easily you integrated [grounding with Google Search](https://ai.google.dev/gemini-api/docs/grounding?lang=python#configure-search) capabilities.  The `Agent` class and the `google_search` tool handle the complex interactions with the LLM and grounding with the search API, allowing you to focus on the agent's *purpose* and *behavior*.

![intro_components.png](../../assets/quickstart-streaming-tool.png)

Copy-paste the following code block to `__init__.py` file.

```py title="__init__.py"
from . import agent
```

## 3\. Set up the platform { #set-up-the-platform }

To run the agent, choose a platform from either Google AI Studio or Google Cloud Vertex AI:

=== "Gemini - Google AI Studio"
    1. Get an API key from [Google AI Studio](https://aistudio.google.com/apikey).
    2. Open the **`.env`** file located inside (`app/`) and copy-paste the following code.

        ```env title=".env"
        GOOGLE_GENAI_USE_VERTEXAI=FALSE
        GOOGLE_API_KEY=PASTE_YOUR_ACTUAL_API_KEY_HERE
        ```

    3. Replace `PASTE_YOUR_ACTUAL_API_KEY_HERE` with your actual `API KEY`.

=== "Gemini - Google Cloud Vertex AI"
    1. You need an existing
       [Google Cloud](https://cloud.google.com/?e=48754805&hl=en) account and a
       project.
        * Set up a
          [Google Cloud project](https://cloud.google.com/vertex-ai/generative-ai/docs/start/quickstarts/quickstart-multimodal#setup-gcp)
        * Set up the
          [gcloud CLI](https://cloud.google.com/vertex-ai/generative-ai/docs/start/quickstarts/quickstart-multimodal#setup-local)
        * Authenticate to Google Cloud, from the terminal by running
          `gcloud auth login`.
        * [Enable the Vertex AI API](https://console.cloud.google.com/flows/enableapi?apiid=aiplatform.googleapis.com).
    2. Open the **`.env`** file located inside (`app/`). Copy-paste
       the following code and update the project ID and location.

        ```env title=".env"
        GOOGLE_GENAI_USE_VERTEXAI=TRUE
        GOOGLE_CLOUD_PROJECT=PASTE_YOUR_ACTUAL_PROJECT_ID
        GOOGLE_CLOUD_LOCATION=us-central1
        ```

## 4. Try the agent with `adk web` { #try-the-agent-with-adk-web }

Now it's ready to try the agent. Run the following command to launch the **dev UI**. First, make sure to set the current directory to `app`:

```shell
cd app
```

Also, set `SSL_CERT_FILE` variable with the following command. This is required for the voice and video tests later.

=== "OS X & Linux"
    ```bash
    export SSL_CERT_FILE=$(python -m certifi)
    ```

=== "Windows"
    ```powershell
    $env:SSL_CERT_FILE = (python -m certifi)
    ```



Then, run the dev UI:

```shell
adk web
```

!!!info "Note for Windows users"

    When hitting the `_make_subprocess_transport NotImplementedError`, consider using `adk web --no-reload` instead.


Open the URL provided (usually `http://localhost:8000` or
`http://127.0.0.1:8000`) **directly in your browser**. This connection stays
entirely on your local machine. Select `google_search_agent`.

### Try with text

Try the following prompts by typing them in the UI.

* What is the weather in New York?
* What is the time in New York?
* What is the weather in Paris?
* What is the time in Paris?

The agent will use the google_search tool to get the latest information to answer those questions.

### Try with voice and video

To try with voice, reload the web browser, click the microphone button to enable the voice input, and ask the same question in voice. You will hear the answer in voice in real-time.

To try with video, reload the web browser, click the camera button to enable the video input, and ask questions like "What do you see?". The agent will answer what they see in the video input.

(Just clicking the microphone or camera button once is enough. Your voice or video will be streamed to models and the model response will be streamed back continuously. Clicking on the microphone or camera button multiple times is not supported.)

### Stop the tool

Stop `adk web` by pressing `Ctrl-C` on the console.

### Note on ADK Streaming

The following features will be supported in the future versions of the ADK Streaming: Callback, LongRunningTool, ExampleTool, and Shell agent (e.g. SequentialAgent).

Congratulations\! You've successfully created and interacted with your first Streaming agent using ADK\!

## Next steps: build custom streaming app

In [Custom Audio Streaming app](../../streaming/custom-streaming.md) tutorial, it overviews the server and client code for a custom asynchronous web app built with ADK Streaming and [FastAPI](https://fastapi.tiangolo.com/), enabling real-time, bidirectional audio and text communication.