Sequence Flow Experiment With Desktop Commander

🚀 Sequence Flow and Desktop Commander Experiment - Hierarchical Task Network Planning System

A Flutter application for visualizing and managing Hierarchical Task Networks (HTN).

🚀 Quick Start

Sequence Flow is an interactive HTN planning system that enables users to model complex planning problems through guided AI conversations. It offers a visualized directed graph to display planning tasks and their complex dependencies, supports hierarchical compression of sub - graphs to manage complexity while retaining cross - boundary references.

✨ Features

• Interactive Modeling: Handle complex planning problems through guided AI conversations.
• Directed Graph Visualization: Present planning tasks in the form of a directed graph, including complex dependencies.
• Hierarchical Compression: Compress sub - graphs to manage complexity while retaining references.
• Three - Column Interface: Consists of an AI chat, an interactive graph, and a hierarchical task list.
• Iterative Problem Solving: Dynamically select problems based on the current planning state.
• External Persistence: Prevent data loss due to LLM context limitations.

📦 Project Structure

sequence_flow/
├── lib/
│   ├── behaviors/      # Implement HTN planning behaviors
│   ├── models/         # Application data models
│   │   └── entities/   # Core entity definitions
│   ├── screens/        # Application interfaces
│   ├── services/       # External service integrations
│   ├── utils/          # Utility functions and auxiliary features
│   └── widgets/        # Reusable UI components
├── assets/             # Static resources (icons, images)
└── test/               # Unit tests and widget tests

📦 Installation

Prerequisites

• Flutter (>= 3.6.1)
• Dart (>= 3.6.0)
• An IDE (VS Code, Android Studio, or IntelliJ)

Steps

1. Clone the repository:

   git clone https://github.com/yourusername/sequence_flow.git
   cd sequence_flow
   

2. Install dependencies:

   flutter pub get
   

3. Run the application:

   flutter run
   

📚 Core Components

Entities

• Task: The basic building block of HTN planning.
• Method: Defines how a task is decomposed.
• TaskDecomposition: The hierarchical relationship between tasks.
• TaskDependency: The non - hierarchical dependency between tasks.
• CrossBoundaryReference: References that cross the boundaries of compressed sub - graphs.
• WorldState: Represents the world state during the planning process.
• Plan: The planning solution to solve the problem.
• Conversation: AI - guided planning conversations.

Behaviors

• HierarchicalCompression: Manages the compression and expansion of sub - graphs.
• AIGuidedPlanning: Controls AI - driven planning conversations.
• GraphStateManagement: Manages the persistence and history of the graph state.

User Interface

• ThreeColumnLayout: The main layout of the application, consisting of three panels.
• AIChat: The left - hand panel for conversations with the AI.
• DirectedGraph: The middle panel for displaying the directed graph.
• TaskList: The right - hand panel for displaying the hierarchical task view.

📚 Development

Code Style

This project follows the official Dart Style Guide and uses the recommended Flutter code specifications.

Build

To build the web version:

flutter build web

📄 Acknowledgments

• This project is based on HTN planning technology for modeling complex domains.
• The UI/UX design is inspired by modern graph visualization tools.