DeepSeek Harness: Building and Extending AI Coding Agents Training Course

Introduction to 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 Execution

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

Model Setup

• Configuring the DeepSeek API key
• Choosing and toggling between models
• Connecting to OpenAI-compatible endpoints

Workspaces and Sessions

• Selecting and changing active workspaces
• Using the session composer for conversation management
• Reading, modifying, and creating files

Executing Coding Tasks

• Summarizing codebases and navigating repository structures
• Running commands and delegating specific tasks
• Tracking and maintaining a structured task plan

Permissions and Sandbox Environment

• Approval workflows and permission policies
• Restricting file and command access
• Local versus remote sandbox backends

Native Tools and Integrations

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

Subagents and Task Delegation

• Creating and forking subagents
• Product providers for external coding agents
• Distributing work across active sessions

The Cordis Plugin Model

• Defining plugins as services, events, and effects
• Managing shared context and lifecycle
• The rationale behind having no privileged core

Profiles, Bundles, and Configuration

• Layering profiles and bundles
• Modifying configuration via cordis.patch.yml
• Reviewing the initialized component tree

Core Subsystems

• The append-only session logging mechanism
• Assembling system prompts and tool schemas
• Understanding the agent loop and turn processing flow

Events and Extension Points

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

Plugin Development

• Registering services on specific context keys
• Incorporating new tools and model adapters
• Releasing and discovering plugins

SDK, Headless Mode, and Automation

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