Skip to content

version_6_docker_vision_agent

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

version_6_docker_vision_agent

README.md

version_6_docker_vision_agent

🌟 Purpose
This version demonstrates how to dockerize and deploy a Gemini-based Vision Agent using Google Cloud Run. It builds on earlier multi-agent A2A + MCP architectures by showing how to wrap a vision agent into a Docker container and run it locally or remotely on GCP.

🚀 Features

  • Gemini Vision Agent – Accepts text+image input and generates a caption using gemini-2.0-flash
  • Dockerized Agent – Package and run your agent in isolated containers
  • Cloud Run Deployment – Easily deploy to Google Cloud Run
  • Port Mapping & Volume Mounting – Full support for image path routing
  • Modular Codebase – Includes models, server logic, and agent logic in a clean structure

📦 Project Structure

version_6_docker_vision_agent/
├── Dockerfile
├── README.md
├── agents/
│   └── vision_agent/
│       ├── __main__.py
│       ├── agent.py
│       ├── image_samples_ai_generated/
│       │   ├── busstop.png
│       │   ├── butterfly.png
│       │   ├── computer.png
│       │   └── rainbow.png
│       └── task_manager.py
├── main.py
├── models/
│   ├── agent.py
│   ├── json_rpc.py
│   ├── request.py
│   └── task.py
├── pyproject.toml
└── server/
    ├── server.py
    └── task_manager.py

🛠️ Prerequisites

  • Python 3.11+
  • Docker installed locally
  • Google Cloud account with gcloud CLI configured
  • Valid GOOGLE_API_KEY for Gemini models

🧪 Local: Build, Run, Test

🔨 Build Docker Image Locally

docker build --platform linux/amd64 --no-cache -t vision_agent_docker .

▶️ Run the Docker Container Locally

docker run --rm -p 10003:10003 \
  -v /absolute/path/to/image_samples_ai_generated:/root/image_samples_ai_generated \
  vision_agent_docker

🔁 Send a Test Request

curl -X POST http://localhost:10003/ \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": "test-1",
    "method": "tasks/send",
    "params": {
      "id": "task-123",
      "sessionId": "session-abc",
      "message": {
        "role": "user",
        "parts": [
          {
            "type": "text",
            "text": "Describe the image || /root/image_samples_ai_generated/computer.png"
          }
        ]
      },
      "historyLength": null,
      "metadata": null
    }
  }'

☁️ Cloud: Build, Push, Deploy, Test

🧱 Build Docker Image with Tag for GCP

docker build --platform linux/amd64 --no-cache -t gcr.io/YOUR_PROJECT_ID/vision-agent-docker .

📤 Push Docker Image to Google Container Registry

docker push gcr.io/YOUR_PROJECT_ID/vision-agent-docker

🚀 Deploy to Google Cloud Run

gcloud run deploy vision-agent-docker \
  --image gcr.io/YOUR_PROJECT_ID/vision-agent-docker \
  --platform managed \
  --region asia-south1 \
  --port 10003 \
  --allow-unauthenticated

🔒 Note: --allow-unauthenticated is used for demo access. Once tested, revoke or delete this service for security.

🌐 Test Deployed Vision Agent

curl -X POST https://YOUR_CLOUD_RUN_URL \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": "test-1",
    "method": "tasks/send",
    "params": {
      "id": "task-123",
      "sessionId": "session-abc",
      "message": {
        "role": "user",
        "parts": [
          {
            "type": "text",
            "text": "Describe the image || /app/agents/vision_agent/image_samples_ai_generated/computer.png"
          }
        ]
      },
      "historyLength": null,
      "metadata": null
    }
  }'

🧹 Delete Service and Clean Up

🚫 Delete the Cloud Run Service

gcloud run services list --region asia-south1 
gcloud run services delete vision-agent-docker --region asia-south1

🗑️ Delete the Container Image from GCP

gcloud container images list-tags gcr.io/YOUR_PROJECT_ID/vision-agent-docker
gcloud container images delete gcr.io/YOUR_PROJECT_ID/vision-agent-docker@sha256:YOUR_DIGEST_ID --force-delete-tags

✨ More Resources