ADR-0024: Polymorphic SoulContext Hierarchy in AISME¶

1. Context¶

Spector's domain model defines a rich, polymorphic sealed soul hierarchy in com.spectrayan.spector.memory.model:

• UserSoul: Human digital twin and persona continuity (PersonaContext, IdiolectProfile, VocalProsodyDNA, EmbodiedKinesicsDNA, identityEmbedding).
• AgentSoul: Autonomous AI assistant persona, tools, and system prompt.
• TenantSoul: Enterprise compliance and domain focus policies.
• OrgUnitSoul: Team/department expertise and guardrails.

However, the initial AISME integration (GenerativeSelfModel, AismeBundle, SpectorMemoryBuilder) was constrained exclusively to AgentSoul.

2. Problem Statement¶

Restricting AISME to AgentSoul prevented UserSoul (digital twin and ancestral continuity) from directly seeding the top-down generative prior \(\boldsymbol{\mu}_0\) of the active inference self-model. Furthermore, enterprise multi-tenant soul stacks (TenantSoul, OrgUnitSoul) could not be exposed in AismeBundle, preventing departmental and tenant-level policies from influencing generative inference and policy selection.

3. Decision Drivers¶

• Polymorphic Generative Priors: All types in the SoulContext hierarchy must be capable of seeding or influencing generative self-model priors.
• Hierarchical Prior Blending: Multi-tier soul configurations (Tenant + Org + Agent + User) must support composite prior synthesis.
• Backward Compatibility: Existing clients and tests depending on AgentSoul accessors in AismeBundle must continue functioning without modification.
• Builder Unification: SpectorMemoryBuilder and AismeBuilder must expose uniform polymorphic methods for soul configuration.

4. Considered Options¶

Option 1: Parallel Model Implementations¶

• Description: Create separate UserGenerativeSelfModel, AgentGenerativeSelfModel, and TenantGenerativeSelfModel classes.
• Advantages: Avoids modifying existing classes.
• Disadvantages: Massive code duplication; fails to support blended multi-soul hierarchies.

Option 2: Polymorphic SoulContext Hierarchy Integration (Selected)¶

• Description: Upgrade GenerativeSelfModel and AismeBundle to operate on the sealed SoulContext interface, supporting polymorphic prior computation and multi-soul blending while providing backward-compatible adaptors.
• Advantages: Universal support for all soul types; supports multi-tier composite blending; 100% backward compatible.
• Disadvantages: Requires updating constructor signatures in internal builders.

5. Decision Outcome¶

Chosen Option: Option 2 (Polymorphic SoulContext Hierarchy Integration).

Architectural Decisions:¶

D1: Polymorphic SoulContext in GenerativeSelfModel¶

• Replace AgentSoul soul with SoulContext soul in GenerativeSelfModel.
• Support polymorphic prior mean initialization via SoulContext.identityEmbedding().
• Add fromSoulsAndProfile(List<SoulContext> soulContexts, CognitiveProfile profile, int dimensions) to compute a composite, blended generative prior \(\boldsymbol{\mu}_0\).

D2: Polymorphic AismeBundle¶

• Update AismeBundle to hold SoulContext primarySoul and List<SoulContext> soulContexts.
• Provide backward-compatible accessor public AgentSoul agentSoul() which returns (AgentSoul) primarySoul if applicable, else null.

D3: Builder Alignment (SpectorMemoryBuilder & AismeBuilder)¶

• Expose soul(SoulContext soul) and soulContexts(List<SoulContext> contexts) in SpectorMemoryBuilder.
• Pass polymorphic soul and soulContexts to AismeBuilder.build(...).

Positive Consequences¶

• Closes the architectural gap for human digital twin persona storage (Homo Digitalis) by allowing UserSoul to directly drive active-inference self-models.
• Supports multi-tenant enterprise and departmental policy integration.
• 100% backward compatible with existing code expecting AgentSoul.

Negative Consequences & Trade-offs¶

• Internal bundle structures hold a collection of souls, requiring null-safe checks when casting to specific concrete types.

6. Pros and Cons of the Options¶

7. Implementation Plan¶

1. Phase 1: Update GenerativeSelfModel to accept SoulContext and implement prior blending.
2. Phase 2: Refactor AismeBundle with primarySoul and soulContexts while retaining agentSoul().
3. Phase 3: Update AismeBuilder and SpectorMemoryBuilder fluent APIs.
4. Phase 4: Add test coverage for UserSoul active-inference initialization.

8. Code Reference & Verification¶

• Primary Module(s): memory/spector-memory
• Key Packages: com.spectrayan.spector.memory.model, com.spectrayan.spector.memory.aisme.fegr, com.spectrayan.spector.memory.builder
• Classes: SoulContext.java, UserSoul.java, AgentSoul.java, GenerativeSelfModel.java, AismeBundle.java, AismeBuilder.java
• Verification Tests: GenerativeSelfModelTest.java, PolymorphicSoulAismeTest.java