Skip to main content

Digitorn

Build AI agent apps in YAML. You declare agents, tools, security, and UI; Digitorn runs them.

hello.yaml
app:
app_id: hello
name: "Hello"

agents:
- id: assistant
role: assistant
brain:
provider: ollama
model: qwen25-7b-gpu:latest
backend: openai_compat
config:
base_url: http://localhost:11434/v1
api_key: ollama
system_prompt: "Reply with exactly one word: pong."
bash
digitorn install hello.yaml
digitorn chat hello
# -> pong

Use the copy button on code blocks to paste examples into your own files.

Visual builder: Digitorn Studio (/studio): build on the canvas, Install for Test, then Publish.


Where to start​

  1. Install Digitorn on your machine.
  2. Follow the Tutorial from hello-world to a production-shaped app.
  3. Keep Language open while you edit YAML.
  4. Look up tools in Module reference.
  5. Check What you can use in app YAML when you are unsure a name still exists.

Practical recipes: How-tos. Sample shapes: Examples.


The app YAML​

A Digitorn app is one YAML file (or a small bundle directory). The usual top-level blocks:

BlockPurposeReference
app:Identity (id, name, version, …).App config
runtime:Mode, turns, timeouts, hooks.App config
agents:Brains, prompts, multi-agent roles.Agents
tools:Modules, capabilities, channels.Tools
security:Behavior profile, credential schema.Security
ui:Greeting, workspace pane, widgets.Client UI
dev:Variables, skills, local helpers.Bundle / skills
flow:Optional fixed pipelines.Flows

Also valid when you need them: context, documents, docs, templates, requirements.

Validate any file by installing it - the daemon compiles the YAML server-side and reports diagnostics if anything's wrong, so there's no separate dry-run step needed:

bash
digitorn install path/to/app.yaml

Documentation scope​

These docs explain how to write Digitorn applications in YAML. They are not an internal handbook of the runtime process or its HTTP surface. Stay on Language, Tutorial, Modules, and Concepts above.