DeepSeek Harness: Building and Extending AI Coding Agents Training Course

Introduction to DeepSeek Harness

• Overview of DeepSeek Harness and its role in the ecosystem
• The "everything is a plugin" architectural philosophy
• Web UI, CLI, and headless operational modes

Setup and Initial Launch

• Node.js dependencies and package installation
• Starting the Web UI using npx
• Compiling and running from the source code

Model Setup

• Configuring the DeepSeek API key
• Choosing and switching between models
• Incorporating OpenAI-compatible endpoints

Workspaces and Sessions

• Selecting and changing active workspaces
• Using the session composer to manage conversation flow
• Performing file reading, editing, and generation tasks

Executing Coding Tasks

• Summarizing repositories and navigating codebases
• Running commands and delegating specific tasks
• Maintaining and updating task plans

Permissions and Sandbox Environment

• Defining approval and permission policies
• Restricting file and command access
• Utilizing local and remote sandbox backends

Native Tools and Integrations

• Filesystem, terminal, and shell utilities
• Support for the Model Context Protocol
• Integration with the Language Server Protocol

Subagents and Task Delegation

• Creating and forking subagents
• Using product providers for external coding agents
• Distributing work across a session

The Cordis Plugin Architecture

• Plugins defined as services, events, and effects
• Shared context and lifecycle management
• The absence of a privileged core component

Profiles, Bundles, and Configuration

• Layering profiles and bundles
• Modifying configuration via cordis.patch.yml
• Reviewing the boot tree structure

Core Subsystems

• The append-only session logging mechanism
• Assembly of system prompts and tool schemas
• The agent loop and turn-based processing flow

Events and Extension Points

• Session, agent, and capability events
• Waterfall and interceptor event handling
• Linking behaviors to specific extension points

Plugin Development

• Registering services using context keys
• Incorporating new tools and model adapters
• Publishing and locating available plugins

SDK, Headless Mode, and Automation

• Controlling the harness from external processes
• Utilizing TypeScript and Python SDKs
• Implementing CLI and headless automation workflows