Indonesia Guide Agent

A multi-agent AI tour guide for Indonesia, built with Google Agent Development Kit (ADK). Ask about Indonesian tourism destinations, culture, or food and receive friendly, Wikipedia-backed answers delivered in the voice of a real-life tour guide.

How It Works

The agent uses a sequential multi-agent pipeline:

User → tour_guide_greeter → indo_tour_workflow
                                 ├── agent_researcher  (Wikipedia lookup)
                                 └── agent_formatter   (Tour guide response)

1. tour_guide_greeter — Greets the user and saves their prompt to shared state via the add_prompt_to_state tool, then hands off to the workflow.
2. agent_researcher — Queries Wikipedia to gather facts about the requested Indonesian location, culture, or food.
3. agent_formatter — Takes the raw research data and rewrites it in an engaging, tour-guide style for the user.

Prerequisites

• Python 3.10+
• A Google Cloud project with Vertex AI or Gemini API access
• Google Cloud credentials configured locally (gcloud auth application-default login)

Installation

git clone https://github.com/ammarsufyan/IndonesiaGuideAgent.git
cd indonesia_guide_agent

python -m venv .venv
source .venv/bin/activate

pip install -r requirements.txt

Configuration

Create a .env file in the project root:

PROJECT_ID=your-gcp-project-id
PROJECT_NUMBER=your-gcp-project-number
SA_NAME=your-service-account-name
SERVICE_ACCOUNT=your-service-account@your-project.iam.gserviceaccount.com
MODEL=gemini-2.5-flash

Variable	Description
PROJECT_ID	Google Cloud project ID
PROJECT_NUMBER	Google Cloud project number
SA_NAME	Service account name used for Cloud Run deployment
SERVICE_ACCOUNT	Full service account email (name@project.iam.gserviceaccount.com)
MODEL	Model identifier passed to Google ADK (e.g. gemini-2.5-flash)

Running the Agent

Use the Google ADK CLI to serve the agent locally:

adk web

Or run it via the ADK runner directly:

adk run IndonesiaGuideAgent

Dependencies

Package	Version	Purpose
google-adk	1.14.0	Agent framework (Agent, SequentialAgent, LangchainTool)
langchain-community	0.3.27	WikipediaQueryRun + WikipediaAPIWrapper
wikipedia	1.4.0	Wikipedia API backend for LangChain

License

MIT © 2026 Ammar Sufyan