Content
# Music class ad generation agent 🎼
A company needs an agent for a project to publicize music classes. There is an image generation agent that can create illustrations according to the company's specific brand guidelines:
- a retro illustration style: Pop Art comic strip panel
- a color palette: bold, saturated, mainly primary colors
- include musical and teaching imagery
With this illustrator agent made a remote agent, it can be shared to be used by other organizations using different programming languages or agent frameworks. This illustrator agent can then be incorporated by an ad content creator agent.
Models used: `gemini-2.5-flash` and `gemini-2.5-flash-image`
<details>
<summary>Instructions</summary>
<details open>
<summary>1. Set up your environment</summary>
### 1. Set up your environment
Enable the Vertex AI API and Cloud Run API:
```
gcloud services enable \
aiplatform.googleapis.com \
run.googleapis.com
```
```
gcloud services list --enabled
NAME: run.googleapis.com
TITLE: Cloud Run Admin API
NAME: aiplatform.googleapis.com
TITLE: Vertex AI API
...
```
Download and install the Agent Development Kit (ADK), and the Agent2Agent (A2A) SDK:
```
# Install ADK and the A2A Python SDK
cd ~
export PATH=$PATH:"/home/${USER}/.local/bin"
python3 -m pip install google-adk==1.8.0 a2a-sdk==0.2.16
pip install --upgrade google-genai
# Correcting a typo in this version
sed -i 's/{a2a_option}"/{a2a_option} "/' ~/.local/lib/python3.12/site-packages/google/adk/cli/cli_deploy.py
```
Based on: [Connect to remote agents with ADK and the Agent2Agent (A2A) SDK](https://www.cloudskillsboost.google/focuses/132170?parent=catalog)
Set up the `.env` files.
Create a Cloud Storage bucket named `PROJECT_ID-bucket`.
```
gcloud storage buckets create gs://PROJECT_ID-bucket --location=BUCKET_LOCATION --project=PROJECT_ID --no-public-access-prevention
```
Make all objects in the bucket publicly readable. To make all objects in a bucket readable to everyone on the public internet, grant the principal `allUsers` the `Storage Object Viewer` (`roles/storage.objectViewer`) role:
```
gcloud storage buckets add-iam-policy-binding gs://BUCKET_NAME --member=allUsers --role=roles/storage.objectViewer
```
</details>
<details open>
<summary>2. Explore the illustrations agent</summary>
### 2. Explore the illustrations agent
From the Cloud Shell Terminal, launch the ADK dev UI:
```
adk web
```
Output:
```
INFO: Started server process [3274]
INFO: Waiting for application startup.
+-----------------------------------------------------------------------------+
| ADK Web Server started |
| For local testing, access at http://localhost:8000. |
+-----------------------------------------------------------------------------+
INFO: Application startup complete.
INFO: Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
```
Clicking the [http://127.0.0.1:8000](https://8000-cs-70724c49-cfc8-4e60-b3d7-15a89c90e20e.cs-us-east1-pkhd.cloudshell.dev/dev-ui/) link, open the ADK Dev UI.
From the 'Select an agent' dropdown on the left, select `illustration_agent`.
Query the agent with text related to music classes, e.g.: `piano and violin`
<img src="https://github.com/eloiisa-gh/adk_and_a2a/blob/main/img_docs/illustration_agent_pianoandviolin.png?raw" alt="illustration agent" width=75% height=75% />
Resulting image:
<img src="https://github.com/eloiisa-gh/adk_and_a2a/blob/main/img_docs/illustration_agent_generated-images_pianoandviolin.png?raw" alt="illustration agent result" width=25% height=25% />
</details>
<details open>
<summary>3. Deploy the illustration agent as an A2A Server</summary>
### 3. Deploy the illustration agent as an A2A Server
Deploy the agent to Cloud Run as an [A2A server](https://google.github.io/adk-docs/a2a/intro/):
```
adk deploy cloud_run \
--project GCP_PROJECT_ID \
--region GCP_LOCATION \
--service_name illustration-agent \
--a2a \
illustration_agent
```
Output:
```
Service [illustration-agent] revision [illustration-agent-00001-xpp] has been deployed and is serving 100 percent of traffic.
Service URL: https://illustration-agent-ProjectNumber.GCP_LOCATION.run.app
```
https://illustration-agent-485230244166.us-central1.run.app/
We use this URL for the agent's url in the card (JSON) files.
</details>
<details open>
<summary>4. Enable another agent to call the illustration agent remotely</summary>
### 4. Enable another agent to call the illustration agent remotely
This second agent creates content for ads. It writes a headline and a couple of sentences of body text, then transfers to the illustration agent to generate an image to illustrate that text.
Launch the UI from the Cloud Shell Terminal:
```
cd ~/music-class-ads
adk web
```
From the 'Select an agent' dropdown on the left, select `ad_content_agent`. Query the agent with an idea for an ad:
```
Create content for an ad about singing classes for beginners
```
Output:
- a headline and body text written by the `ad_content_agent`
- a call to `transfer_to_agent`, indicating a transfer to the `illustration_agent`
- the response from the `illustration_agent` with a link to the new image.
Verify that `ad_content_agent` is using `illustration_agent` as a sub-agent:
<img src="https://github.com/eloiisa-gh/adk_and_a2a/blob/main/img_docs/ad_content_agent_singingbeginners.png?raw" alt="ad agent" width=75% height=75% />
Resulting image:
<img src="https://github.com/eloiisa-gh/adk_and_a2a/blob/main/img_docs/ad_content_agent_generated-images_singingbeginners.png?raw" alt="ad agent result" width=25% height=25% />
</details>
</details>
an example result after updating the style instructions:
<img src="https://github.com/eloiisa-gh/adk_and_a2a/blob/main/img_docs/generated-images_drumming.png?raw" alt="updated results" width=40% height=40% />
Try it yourself: https://ad-content-agent-485230244166.us-central1.run.app
