Skip to main content

Tutorial

A linear path from "hello world" to a production-shape multi-agent app. Read in order. Each step builds on the previous one and ends with a verified live test.

Prerequisites

  • Python 3.12+
  • A running daemon (digitorn start)
  • Either a local Ollama instance OR an API key for one of the supported providers (Validated provider hints)

Steps

StepWhat you buildWhat you learn
1. Getting startedHello-world chatbotInstall, validate, deploy, chat from the CLI
2. Conversation with memoryAssistant that remembers facts you tell itSystem prompts, the memory module, Remember, auto-recall
3. Add a toolBot that reads files in your workspaceModules, capabilities, the filesystem module
4. Multi-agent teamCoordinator + 2 parallel specialistsThe Agent tool, role-based delegation, isolation
5. Background modeCron-driven monitorTriggers, channels, payload schemas
6. UI surfacesWorkspace pane the agent writes intoui.workspace (renderer), workspace module
7. DeployingProduction-shape app with deny-by-defaultCapabilities, behaviour profile, credential schema

Every step is written and tested live: each YAML deploys against the daemon, every transcript shown is the verbatim output, every tool call count and trigger result was captured by the live event stream.

What you'll have at the end

A multi-agent app that:

  • Spawns specialist sub-agents in parallel
  • Persists working memory across conversations
  • Has a workspace pane the agent writes to in real time
  • Connects to one external service via the credentials vault
  • Has a background trigger that wakes it on a cron schedule
  • Runs sandboxed in production

When to leave the tutorial

The tutorial is opinionated and linear. Once you've finished it, work from Reference and Language directly - those are the canonical surfaces. Come back to Concepts any time the framework's why is unclear.