Open Autonomous Intelligence Initiative

This post defines shared abstract types used by OAII Base Model object definitions.

Normative keywords MUST, SHOULD, MAY are used as in OAII object definitions.

---

1. Identifier Types

All identifiers MUST be locally unique within their scoping context and SHOULD be globally unique when exported beyond the originating edge system.

• WorldId: Identifier for a World
• EventId: Identifier for an Event
• SignalId: Identifier for a Signal
• DeviceId: Identifier for a Device
• SensorId: Identifier for a Sensor
• AgentId: Identifier for an Agent
• AxisId: Identifier for a ContextualAxis
• MechanismId: Identifier for a mechanism implementation (recognition, continuity, fusion, etc.)
• PolicyId: Identifier for a Policy
• LogId: Identifier for a Log stream or record set
• InterfaceId: Identifier for an Interface

Representation (Non‑Normative): Identifiers may be UUIDs, ULIDs, hash-derived IDs, or compound IDs (e.g., device_id + local_counter).

---

2. Parameter and Result Containers

OAII uses generic containers to describe attributes and methods in an implementation‑agnostic manner.

2.1 ParameterSet (Normative)

A ParameterSet is a keyed collection of parameters that:

• MUST be serializable;
• MUST support explicit typing or type tagging;
• SHOULD include versioning when used for mechanisms or policies.

Minimum fields:

• params_version: version indicator for the parameter schema
• params: keyed values

2.2 ResultSet (Normative)

A ResultSet is a keyed collection of outputs that:

• MUST be serializable;
• MUST include status information;
• SHOULD include diagnostic or trace data when safe to disclose.

Minimum fields:

• status: {SUCCESS, PARTIAL, FAIL}
• result_version: version indicator for result schema
• result: keyed values
• error_code (optional)
• error_detail (optional)

---

3. Confidence and Uncertainty

3.1 Confidence (Normative)

Confidence expresses a degree of certainty for recognition, classification, or inference.

Minimum fields:

• confidence_value: numeric value in [0,1]
• confidence_model: identifier for the model or interpretation of confidence

3.2 UncertaintyModel (Normative)

An UncertaintyModel describes how uncertainty is represented.

Minimum fields:

• uncertainty_model_id
• uncertainty_params: ParameterSet

---

4. Privacy and Access Classes

4.1 PrivacyClass (Normative)

A PrivacyClass indicates sensitivity of data or representation.

Minimum categories for MVP:

• PUBLIC: safe to export broadly
• LOCAL: intended to remain on-device / on-prem
• SENSITIVE: export only with explicit authorization
• RESTRICTED: export prohibited by default; may require strong controls

4.2 AccessClass (Normative)

AccessClass indicates permitted operations:

• READ_ONLY
• READ_WRITE
• WRITE_ONCE

---

5. ContextualAxis

5.1 ContextualAxis (Normative)

A ContextualAxis defines a dimension along which interpretation occurs within a World.

Minimum fields:

• axis_id: AxisId
• axis_type: AxisType
• label: human-readable name
• value_domain: ValueDomain

Optional fields:

• unit
• frame_ref
• resolution
• uncertainty_model: UncertaintyModel
• privacy_class: PrivacyClass

5.2 AxisType (Normative)

AxisType is an extensible enumeration.

Minimum AxisType values for the Edge‑primary Aging‑in‑Place MVP:

• TEMPORAL: time-of-day, duration, routine intervals
• SPATIAL: rooms/zones/regions, adjacency, thresholds
• ACTIVITY: observable actions/interactions (non-medical)
• SOCIAL: presence/identity tokens, known/unknown participant context
• DEVICE_STATE: device and sensor operational state
• POLICY_CONTEXT: notification and escalation context

Implementations MAY define additional AxisTypes; additional values MUST be namespaced or otherwise uniquely identifiable when exported.

---

6. ValueDomain

6.1 ValueDomain (Normative)

A ValueDomain defines the allowable values and ordering for an axis or attribute.

Minimum forms:

• DISCRETE_SET: set of allowed values
• ORDERED_SET: ordered categories
• CONTINUOUS_RANGE: numeric range with bounds

Minimum fields:

• domain_type: one of the above
• domain_spec: ParameterSet describing set elements or range bounds

---

7. Mechanism Bundles (Attributes and Methods)

OAII supports consistent representation of mechanisms using an ID/params/result pattern.

7.1 Mechanism Bundle (Normative)

A mechanism bundle consists of:

• _id: MechanismId
• _params: ParameterSet
• _result (optional): ResultSet

Examples:

• continuity_mechanism_id / continuity_mechanism_params / continuity_mechanism_result
• recognition_mechanism_id / recognition_mechanism_params / recognition_mechanism_result
• fusion_mechanism_id / fusion_mechanism_params / fusion_mechanism_result

7.2 Method Representation (Normative)

Methods are represented as:

• method_name
• method_params: ParameterSet
• method_result: ResultSet

Methods SHOULD document:

• idempotency
• preconditions / postconditions
• failure modes

---

Status: Draft v0.1 (Shared Types and Conventions)