Build domain-aware behavior

Define a Domain from the public task interface, then let live evidence determine episode-specific behavior.

Give Adapt-1 the public task contract your application already has, then send live observations and outcomes. The returned result can be consumed directly without generated language.

Default path: define only the public vocabulary, observable inputs, available actions or policies, native outcomes, and result contract the workflow needs. Omit learner settings unless a measured integration requirement justifies an override.

The Domain returned by create or read is the resolved contract. It can include defaults filled in by Adapt-1. Send only the Domain fields your integration needs; returned defaults do not need to be copied into the next create request.

Domain contract

A Domain can declare whichever parts of the stable task grammar the workflow uses: entities, relations, signals, event and outcome semantics, action or plan interfaces, and an output contract. Candidate hypotheses, custom rules, query templates, and specialized learner mappings are optional.

When a workflow intentionally configures structured transition learning, choose which request fields are observations, which field identifies the action, and what observable result should be predicted. Configured paths, types, grouping, event boundaries, and support controls define that learner view.

A Domain does not contain the current episode’s hidden mapping, correct procedure, target coordinates, solution path, or answer-bearing episode information. Core can begin with no task history inside the declared scope and form task-local structure from the observations and outcomes it receives.

For example, an application can translate an external observation into entities, relations, and measurements, then request a contract-defined relation, compatibility judgment, location, or other structured result. The adapter supplies observable facts; it does not supply the interpretation Core is expected to form.

A stable Domain can declare public meanings and available operations. Live evidence should determine the current roles, useful action, supported hypothesis, and result.

Structured transitions

Keep three data layers separate:

Layer	Role
Operational record	Store all observable data from before and after the event. Also store provenance.
Learner view	Use observations, the proposed intervention, and permitted derived features available before the observable consequence.
Application-only context	Keep privileged state, preferred actions, solver output, and downstream judgments outside transition inputs and observation-only targets.

The target must represent all observable distinctions that Adapt-1 must predict. The target must not state that an action was correct. Use grouping only to separate incompatible transition behavior. Use one joint target if separate fields can make an invalid consequence.

The API can store a Domain event that the transition learner does not accept. Check learner admission and the applicable learner version. Do this before you interpret an unchanged prediction.

Design Domains for transition learning covers event boundaries, value geometry, action_path, group_by_paths, target structure, support, abstention, hypotheses, and read-only calls.

Causal event schema

A Domain can declare the event schema needed to preserve an exact trace:

state observation -> action commitment -> exact execution -> transition -> native outcome

It can also declare that a native outcome such as advancement is positive evidence when that meaning is part of the task contract. It cannot predeclare that a particular episode action causes advancement. Adapt-1 can form and reuse that mapping from observed transitions, scoped outcomes, contrasts, and later states.

Default operating loop

declare domain -> add observations/events -> query or select -> execute externally -> record transition/outcome -> add feedback -> query again

Routes

Step	Route	Purpose
Create the task contract	POST /api/v1/domains	Declare the domain and its schema
Extend the schema	POST /api/v1/domains/{domain_id}/schemas	Add domain schema information
Add rules	POST /api/v1/domains/{domain_id}/rules	Register domain rules
Record an event	POST /api/v1/domains/{domain_id}/events	Add domain-shaped evidence/state
Query	POST /api/v1/domains/{domain_id}/query	Reason over the available domain state
Explain	POST /api/v1/domains/{domain_id}/explain	Inspect support, missing evidence, and conflicts using the route contract
Submit feedback	POST /api/v1/domains/{domain_id}/feedback	Send an outcome signal

Domain definition and lifecycle

A Domain is owner-scoped. The authenticated identity owns its definition, evidence, decisions, and learner state; a body session_id does not grant access to another owner’s Domain.

Lifecycle step	UAT endpoint
List owned Domains	GET https://rei-neuroadapt-api.reilabs.org/api/v1/domains
Read one owned definition	GET https://rei-neuroadapt-api.reilabs.org/api/v1/domains/{domain_id}
Reset feedback-policy learned state	POST https://rei-neuroadapt-api.reilabs.org/api/v1/domains/{domain_id}/policy/reset
Delete the Domain, its evidence, and learned state	DELETE https://rei-neuroadapt-api.reilabs.org/api/v1/domains/{domain_id}

POST /schemas replaces the supplied schema section and POST /rules replaces the supplied hypotheses, templates, or learning section. Send the complete desired section rather than a partial patch. Domain routes already scope retrieval to the path domain_id, so do not duplicate it in a metadata filter.

Use the reset route only when the intended boundary is feedback-policy learning. Use deletion for a cold Domain evaluation; see Operations and retries for the reset-scope matrix.

Response fields

Choose the field that answers the application question. Depending on the route and domain, that may be items, ranked_hypotheses, missing_evidence, predicted_observations, core_support, or policy_scores.

These fields expose the behavioral result of the current domain state. They do not reveal or define all of Core’s internal machinery.

Domain events and adaptive-policy events

POST /api/v1/domains/{domain_id}/events records a normal domain event. POST /api/v1/domains/{domain_id}/adapt/events updates an online adaptive policy and returns prediction.<output_key>.

Use the route whose output contract matches the desired result.

Domain reuse

Reuse one domain across episodes or deployments that share the same public task semantics. Record any adapter, action dictionary, outcome translation, or workflow hint as part of the information condition. Changing those semantics changes what Adapt-1 is being asked to reason over.