Advanced GenAI Applications

This brings together concepts from Domain 2.

A sophisticated architecture could look like:

                         User
                           │
                           ▼
                    API Gateway
                           │
                           ▼
                    Agent Application
                           │
                ┌──────────┼──────────┐
                ▼          ▼          ▼
             Agent 1    Agent 2    Agent 3
                │          │          │
                └──────────┼──────────┘
                           ▼
                    Tool / MCP Layer
                           │
             ┌─────────────┼─────────────┐
             ▼             ▼             ▼
          Lambda         APIs           Data
                           │
                           ▼
                      Bedrock FMs
                           │
                           ▼
                       Response

Useful for AWS-native agentic applications.

Agent
 ↓
Plan
 ↓
Tools
 ↓
Actions
 ↓
Results

Useful when multiple specialized agents need to cooperate.

Coordinator
 ├── Research Agent
 ├── Analysis Agent
 └── Response Agent

Useful when explicit workflow orchestration is required.

Step 1
 ↓
Step 2
 ↓
Condition
 ├── Path A
 └── Path B
 ↓
Step 3

Prompt chaining means connecting multiple model/prompt steps.

Prompt 1
   ↓
Output
   ↓
Prompt 2
   ↓
Output
   ↓
Prompt 3
   ↓
Final Result

Useful when one large prompt is less suitable than several controlled stages.