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 design philosophy of "everything is a plugin"
• Exploration of Web UI, CLI, and headless modes

Installation and Initial Setup

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

Model Configuration

• Configuring the DeepSeek API key
• Selecting and switching between models
• Integrating OpenAI-compatible endpoints

Workspaces and Sessions

• Selecting and switching between workspaces
• Understanding the session composer and conversation flow
• Reading, editing, and generating files

Executing Coding Tasks

• Summarizing repositories and navigating codebases
• Executing commands and delegating tasks
• Maintaining structured task plans

Permissions and Sandbox Environment

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

Integrated Tools and Protocols

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

Subagents and Task Delegation

• Spawning and forking subagents
• Product providers for other coding agents
• Distributing work across a session

The Cordis Plugin Architecture

• Plugins defined as services, events, and effects
• Shared context and lifecycle management
• Reasons for the absence of a privileged core

Profiles, Bundles, and Configuration

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

Core Subsystems

• The append-only session log mechanism
• Assembly of system prompts and tool schemas
• Agent loop mechanics and turn flow

Events and Extension Points

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

Plugin Development

• Registering services on context keys
• Incorporating tools and model adapters
• Publication and discovery of plugins

SDK, Headless Operation, and Automation

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