DeepSeek Harness: Building and Extending AI Coding Agents Training Course

Overview of DeepSeek Harness

• Defining DeepSeek Harness and its role in the ecosystem
• The "everything is a plugin" design principle
• Web UI, CLI, and headless operating modes

Setup and Initial Launch

• Node.js prerequisites and package setup
• Starting the Web UI using npx
• Compiling and executing from source code

Model Setup

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

Workspaces and Session Management

• Selecting and navigating between workspaces
• Using the session composer and managing conversation flow
• Viewing, modifying, and creating files

Executing Coding Tasks

• Summarizing repositories and exploring codebases
• Running commands and delegating tasks
• Maintaining an organized task plan

Permissions and Sandboxing

• Managing approval and permission policies
• Restricting file and command access
• Utilizing local and remote sandbox environments

Native Tools and Integrations

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

Sub-agents and Task Delegation

• Creating and forking sub-agents
• Product providers for external coding agents
• Distributing work across sessions

The Cordis Plugin Architecture

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

Profiles, Bundles, and Configurations

• Managing profiles and bundle hierarchies
• Modifying settings via cordis.patch.yml
• Reviewing the initialized component tree

Essential 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 Capabilities

• Session, agent, and capability-related events
• Waterfall and interceptor event patterns
• Aligning specific behaviors with extension points

Plugin Development

• Registering services under context keys
• Including new tools and model adapters
• Distributing and identifying plugins

SDK, Headless Mode, and Automation

• Controlling the harness from external processes
• Utilizing TypeScript and Python SDKs
• Automation via CLI and headless execution