Sample Agent Development Kit (ADK)

This project demonstrates how to build an AI-powered agents using Agent Development Kit (ADK) and Gemini APIs as well as other LLM APIs.

---

🚀 Features

• Modular multi-agent architecture
• LLM-powered agenta
• Visual content enhancement
• Auto-formatting with narration, colors, and transitions
• Optional looping agent for refinement
• Built with Google ADK, Gemini API and DeepSeek API

---

🤖 Architecture Overview

Parent Agent (LLM Chatbot)
│
├── 📝 Script Writer Agent
│   └─ Uses Gemini API to fetch & generate content
│
├── 👁️ Visualizer Agent
│   └─ Enhances script with visual cues and layout instructions
│
├── 🎨 Formatter Agent
│   └─ Formats the final output with narration, styles, timing
│
└── 🔁 (Optional) LoopAgent
    └─ Repeats the workflow for improved results

---

🧠 Agent Config Example (YAML)

name: script_writer_agent
model: gemini-1.5-flash
instruction: generate_script_steps.yaml
description: Collects topic-related data and drafts a video script.
output_key: script_draft_output

Instruction File Example

steps:
  - Search Google using Gemini API for topic keyword
  - Collect 5–7 unique content pieces
  - Remove duplicates and unrelated data
  - Save result to output key

---

🛠️ Tech Stack

• Google ADK – Agent framework
• Gemini API – LLM for content generation
• Instruction Files (YAML) – Agent guidance
• Session & Memory – Persistent agent communication

---

📦 Setup

⚠️ This is a conceptual/experimental project. Setup instructions below are generalized.

1. Clone the Repo

   git clone https://github.com/kuahoong/adk.git

2. Install Dependencies Make sure Python and required SDKs are installed.

   pip install google-adk

3. Configure Agents you can view the agent file if any changes needed.

4. Run the Parent Agent

   adk web

---

📈 Use Cases

• Automated content pipelines
• Marketing video scripting
• Explainers and educational video drafts
• Creative prototyping with AI

---

📚 Related Reading

• Google Cloud Blog: Introducing ADK
• Gemini API Overview