Configure an adaptive policy contract

Adaptive Policies update online from structured observations and return the current structured prediction immediately.

Route sequence

POST /api/v1/domains
POST /api/v1/domains/{domain_id}/adapt/reset
POST /api/v1/domains/{domain_id}/adapt/events
POST /api/v1/domains/{domain_id}/adapt/predict

The result is read from:

prediction.<output_key>

Do not read this result from predicted_observations, POST /api/v1/domains/{domain_id}/query, or POST /api/v1/domains/{domain_id}/explain. Those belong to the Domain Reasoning and audit contracts.

The adaptive template

The currently exposed interval-policy integration is configured under the exact key:

query_templates.adaptive_interval_policy

The template declares how to read the application’s observation and how to name the returned rows:

Template field	Role
`items_key`	List of observed items inside `values`
`output_key`	Key created under `prediction`
`index_key`	Ordered observation index
`baseline_key`	Optional baseline used by the policy
`bounds`	Optional lower and upper range fields
`input_fields`	Mapping from application fields to policy inputs
`output_fields`	Mapping from policy values to application-facing names

The field mapping translates application schemas into policy inputs and names the returned fields.

Generic integration pattern

caller object
  -> values and input_fields mapping
  -> POST /api/v1/domains/{domain_id}/adapt/events
  -> response.prediction[output_key]
  -> caller response schema

An output adapter can select the response fields required by the downstream schema. Adapt-1 updates the policy state before returning that response.

State and ordering

The policy is online and stateful per domain/session. Send observations in order and use one writer unless interleaving is intentional. Call POST /api/v1/domains/{domain_id}/adapt/reset to start a new policy history. To rebuild from a known record, reset first and replay its observations in order through /adapt/events. Use POST /api/v1/domains/{domain_id}/adapt/predict to read current state without adding an observation. Policy thresholds determine when an item appears in the output; the observation count depends on the selected template configuration.

Scope

adaptive_interval_policy is one structured-output contract exposed by Adapt-1. This page covers its observation mapping, interval outputs, and policy state.

Get started

Concepts

Guides

Planning

Discovery

Operations

API reference

Route sequence

The adaptive template

Generic integration pattern

State and ordering

Scope

​Route sequence

​The adaptive template

​Generic integration pattern

​State and ordering

​Scope

Route sequence

The adaptive template

Generic integration pattern

State and ordering

Scope