Learning Path
This tutorial site has a lot of content; rather than reading top to bottom, choose a path by your experience level. Each level spells out "what to read, and what you'll achieve".
🟢 New users: first get it running
Goal: set up your key, open a dialog box, and know what's going on.
- Quickstart: configure your key, open the Web UI, first conversation
- Command Line:
dsh web/--profile headless/--dump-config - Configuration: profiles, the patch layer, API keys
- What is DSH: what exactly it is
✅ Afterwards you can: configure a key, open the Web UI, and view --dump-config.
🟡 Everyday use: make it convenient
Goal: know how much it can do, and how to manage permissions and privacy.
- Built-in Tools: what the 21 factory tools can do
- Permissions: presets,
ask/never,DSH_PERMISSION_MODE - Data and Privacy: telemetry,
.anonymous-user-id, session storage - Multi-Model Routing: how to configure providers
- Subagents and Parallelism: handing work out
- Goals, Jobs and Todos: pushing long-running matters forward
- Settings and Credentials: runtime configuration and key management
- Session Search / Attachments: search past sessions, attach images into a session
✅ Afterwards you can: manage permissions, multiple providers, spin up subagents, and use goals to advance long-running tasks.
🔵 Developers: add capabilities to the model
Goal: write your own plugins, tools, services, listen to events, and publish configuration.
- Plugin Anatomy: what a plugin is, patch-layer semantics
- First Plugin: hello world
- Write a Tool: add a tool to the model
- Write a Service: shared logic
- Listen to Events: dispatch modes, event catalog
- Configuration & Publishing:
ctx.settings+ publishing - Skills System / MCP Integration: extend capabilities further
- Feedback / Spill Storage / E2B Cloud Sandbox: advanced capabilities mounted on demand
✅ Afterwards you can: write plugins, add tools, register commands, connect MCP, and publish configuration.
🟣 Architects: dig into the principles
Goal: understand how DSH assembles all of this.
- Architecture Overview: the package-grouping map
- All 11 core-mechanics chapters:
boot-config→context→agent-loop→tools-execution→sandbox-security→session→llm-routing→plugin-anatomy→event-system→web-ui→schedule - Workflows and Ralph: multi-agent orchestration
- Agent Presets and Persona: identity and composition
- Built-in Tools: full picture
✅ Afterwards you can: explain an agent's complete path from startup to running a tool, and pinpoint "which layer to extend at".
💡 Tip: all doc links have no numeric prefixes and can be jumped back; use the right-hand directory for quick navigation; the "Principles Course" works well as a reference manual to query, not something to read in one sitting.