Example Prompt Chaining

A TypeScript implementation demonstrating a sophisticated prompt chaining system with specialized AI agents.

Architecture

This system implements a multi-agent architecture where each agent specializes in a specific task:

Agent Flow

graph TD
    A[User Query] --> B[User Query Receiver]
    B --> C[Summarizer]
    C --> D[Pain Identifier]
    D --> E[Plan Creator]
    E --> F[AI Spawner]
    F --> G[Specialized Agents]
    G --> H[Responder]

    C -.Summary.-> H
    D -.Pain Points.-> H
    E -.Plan.-> H
    F -.Execution Results.-> H

    H --> I[Final Response to User]

    style A fill:#e1f5ff
    style H fill:#fff4e1
    style I fill:#e8f5e9

Loading

Components

1. User Query Receiver: Entry point that receives user queries and coordinates the chain
2. Summarizer: Extracts key findings from the user query
3. Pain Identifier: Identifies top pain points from the summary
4. Plan Creator: Creates a step-by-step plan using available tools
5. AI Spawner: Dynamically spawns specialized agents based on the plan
6. Responder: Aggregates results and responds to the user

Installation

npm install

Configuration

Create a .env file use env.example for reference.

Usage

Interactive Mode (Chat with the System)

Start the interactive terminal chat:

npm run dev
# or
npm start

This starts an interactive CLI where you can chat with the system. Available commands:

• help - Show available commands
• clear - Clear screen
• reset - Reset the system
• history - Show chain execution history
• exit - Exit the program

Demo Mode (Watch Example Processing)

Run the built-in demonstration:

npm run demo

This processes example queries and shows the complete chain execution.

Project Structure

src/
├── types/           # TypeScript interfaces and types
├── agents/          # Specialized agent implementations
├── tools/           # Available tools for agents
├── orchestrator/    # Chain orchestration logic
├── cli.ts           # Interactive CLI (npm start/dev)
└── index.ts         # Demo examples (npm run demo)

How It Works

1. User submits a query
2. Query is summarized to extract key information
3. Pain points are identified from the summary
4. A plan is created based on available tools
5. Specialized agents are dynamically spawned to execute the plan
6. Results are aggregated and returned to the user