Skip to main content
An Entity configuration lives at configs/entities/<name>.yaml in the application checkout. The file stem is its command-line identifier; the name field is its display name.

Use the wizard

The wizard builds an Entity from a personality template and your choices. If you also need inference configuration, start with setup.

Write a small configuration

Create configs/entities/rook.yaml:
This inherits harness defaults, including inference and background behavior. Run it with your configured provider:
The application ignores the local .env and the specific configs/entities/canary.yaml overlay. Arbitrary Entity YAML files are not automatically ignored. Add your private Entity files to your own Git exclusions before storing personal backstory or identifiers.

Configure each concern in its own place

The supported fields and merge rules are in configuration. A field that is absent from the loader will not become a feature simply because it appears in YAML.

Refine by observing behavior

Change one concern at a time. Start with a short identity, a few concrete interests, and a clear voice. Try several conversations, inspect the stored knowledge and relationship documents, then adjust. Use a fresh example Entity when you want an isolated experiment. Changing a display name is not a state migration: default storage paths depend on the configured Entity name.

Personality and voice

Separate writing style from spoken presentation.

Knowledge and journal

Give the Entity stable reference material.