Interface IPersonaDefinition

The primary interface defining a complete Persona for a Generalized Mind Instance (GMI).

Interface

IPersonaDefinition

interface IPersonaDefinition {
    id: string;
    name: string;
    label?: string;
    description: string;
    version: string;
    baseSystemPrompt: string | {
        template: string;
        variables?: string[];
    } | {
        content: string;
        priority?: number;
    }[];
    defaultModelId?: string;
    defaultProviderId?: string;
    defaultModelCompletionOptions?: Partial<ModelCompletionOptions>;
    modelTargetPreferences?: ModelTargetPreference[];
    costSavingStrategy?: "always_cheapest" | "balance_quality_cost" | "prioritize_quality" | "user_preference";
    promptEngineConfigOverrides?: Partial<PromptEngineConfig>;
    toolIds?: string[];
    allowedCapabilities?: string[];
    embeddedTools?: ITool<any, any>[];
    allowedInputModalities?: ("text" | "audio_transcription" | "vision_image_url" | "vision_image_base64")[];
    allowedOutputModalities?: ("text" | "audio_tts" | "image_generation_tool_result")[];
    voiceConfig?: PersonaVoiceConfig;
    avatarConfig?: PersonaAvatarConfig;
    personalityTraits?: Record<string, any>;
    moodAdaptation?: PersonaMoodAdaptationConfig;
    defaultLanguage?: string;
    uiInteractionStyle?: "suggestive" | "directive" | "collaborative" | "silent";
    memoryConfig?: PersonaMemoryConfig;
    cognitiveMemoryConfig?: CognitiveMemoryPersonaConfig;
    conversationContextConfig?: PersonaConversationContextConfig;
    metaPrompts?: MetaPromptDefinition[];
    sentimentTracking?: SentimentTrackingConfig;
    contextualPromptElements?: ContextualPromptElement[];
    isCreatorPersona?: boolean;
    isPublic?: boolean;
    activationKeywords?: string[];
    strengths?: string[];
    minSubscriptionTier?: string;
    initialMemoryImprints?: {
        key: string;
        value: any;
        description?: string;
    }[];
    customFields?: Record<string, any> & {
        defaultWorkingMemoryConfig?: any;
        initialUserContext?: Partial<PersonaUserContextDefaults>;
        initialTaskContext?: Partial<PersonaTaskContextDefaults>;
    };
    requiredSecrets?: string[];
}

Properties

id name label? description version baseSystemPrompt defaultModelId? defaultProviderId? defaultModelCompletionOptions? modelTargetPreferences? costSavingStrategy? promptEngineConfigOverrides? toolIds? allowedCapabilities? embeddedTools? allowedInputModalities? allowedOutputModalities? voiceConfig? avatarConfig? personalityTraits? moodAdaptation? defaultLanguage? uiInteractionStyle? memoryConfig? cognitiveMemoryConfig? conversationContextConfig? metaPrompts? sentimentTracking? contextualPromptElements? isCreatorPersona? isPublic? activationKeywords? strengths? minSubscriptionTier? initialMemoryImprints? customFields? requiredSecrets?

Properties

id

id: string

Stable unique identifier for the persona (used for activation & persistence).

name

name: string

Human-readable name surfaced in UI selection lists.

Optional label

label?: string

Optional short label (e.g., abbreviation) for compact UI contexts.

description

description: string

Rich description of goals, domain focus, and behavioral nuance.

version

version: string

Semantic version of persona definition; bump on behavioral / config changes (e.g., '1.2.0').

baseSystemPrompt

baseSystemPrompt: string | {
    template: string;
    variables?: string[];
} | {
    content: string;
    priority?: number;
}[]

Base system prompt (or structured template) establishing foundational directives. Supports:

  • Raw string
  • Templated object { template, variables[] }
  • Ordered array of prompt fragments with priority for deterministic merging.

Type declaration

  • template: string
  • Optional variables?: string[]

Optional defaultModelId

defaultModelId?: string

Default model id to target for this persona's typical tasks (can be routed or overridden).

Optional defaultProviderId

defaultProviderId?: string

Preferred provider if same model family exists across vendors (helps routing heuristics).

Optional defaultModelCompletionOptions

defaultModelCompletionOptions?: Partial<ModelCompletionOptions>

Baseline completion option overrides (temperature, maxTokens, presence penalties, etc.).

Optional modelTargetPreferences

modelTargetPreferences?: ModelTargetPreference[]

Ordered preferences describing desired model traits (cost/perf/latency) for dynamic selection.

Optional costSavingStrategy

costSavingStrategy?: "always_cheapest" | "balance_quality_cost" | "prioritize_quality" | "user_preference"

High-level cost strategy guiding router decisions.

Optional promptEngineConfigOverrides

promptEngineConfigOverrides?: Partial<PromptEngineConfig>

Partial overrides merged onto global PromptEngine configuration for persona specialization.

Optional toolIds

toolIds?: string[]

Referenced tool identifiers persona is permitted to invoke.

Optional allowedCapabilities

allowedCapabilities?: string[]

Abstract capability flags enabling conditional UI / workflow features (e.g., 'web_search').

Optional embeddedTools

embeddedTools?: ITool<any, any>[]

Inline tool instances embedded directly (rare; typically tools live in registry).

Optional allowedInputModalities

allowedInputModalities?: ("text" | "audio_transcription" | "vision_image_url" | "vision_image_base64")[]

Whitelisted input modalities persona accepts (driver for validation in interaction layer).

Optional allowedOutputModalities

allowedOutputModalities?: ("text" | "audio_tts" | "image_generation_tool_result")[]

Output modalities persona can produce (text, TTS synthesized, image generation results).

Optional voiceConfig

voiceConfig?: PersonaVoiceConfig

Voice synthesis configuration (preferred voice id, style, speed).

Optional avatarConfig

avatarConfig?: PersonaAvatarConfig

Visual avatar / representation metadata (image URL, animation style).

Optional personalityTraits

personalityTraits?: Record<string, any>

Arbitrary personality trait map (e.g., { humor_level: 0.7, pedagogical_style: 'socratic' }).

Optional moodAdaptation

moodAdaptation?: PersonaMoodAdaptationConfig

Rules controlling adaptive mood shifts & modulation of tone.

Optional defaultLanguage

defaultLanguage?: string

Default output language (BCP‑47) used when user preference unspecified.

Optional uiInteractionStyle

uiInteractionStyle?: "suggestive" | "directive" | "collaborative" | "silent"

High-level interaction posture for UI behaviors (suggestive hints vs directive instructions).

Optional memoryConfig

memoryConfig?: PersonaMemoryConfig

Memory subsystem tuning (retention horizons, summarization cadence, pinning rules).

Optional cognitiveMemoryConfig

cognitiveMemoryConfig?: CognitiveMemoryPersonaConfig

Cognitive memory system per-persona overrides (encoding, decay, working memory capacity, etc.).

Optional conversationContextConfig

conversationContextConfig?: PersonaConversationContextConfig

Conversation context override strategy (message importance heuristics, summarization triggers).

Optional metaPrompts

metaPrompts?: MetaPromptDefinition[]

System or self-reflective prompts guiding meta-cognition, self-correction, or planning loops.

Optional sentimentTracking

sentimentTracking?: SentimentTrackingConfig

Sentiment tracking configuration. Controls whether the GMI analyzes user emotional state and triggers event-based metaprompts (frustration recovery, confusion clarification, etc.). Opt-in: disabled by default. Turn_interval metaprompts (like self-reflection) always work regardless.

Optional contextualPromptElements

contextualPromptElements?: ContextualPromptElement[]

Dynamic contextual prompt elements evaluated per turn for fine-grained adaptation.

Optional isCreatorPersona

isCreatorPersona?: boolean

Allows persona to perform privileged creation/update of other personas.

Optional isPublic

isPublic?: boolean

If true persona is globally discoverable subject to subscription tier gating.

Optional activationKeywords

activationKeywords?: string[]

Natural language keywords enabling auto-activation in multi-persona environments.

Optional strengths

strengths?: string[]

Strength tag list aiding search & recommendation (e.g., ['typescript', 'design_reviews']).

Optional minSubscriptionTier

minSubscriptionTier?: string

Minimum subscription tier required to access persona (e.g., 'pro', 'enterprise').

Optional initialMemoryImprints

initialMemoryImprints?: {
    key: string;
    value: any;
    description?: string;
}[]

Seed working memory imprints establishing initial context (preferences, calibration data).

Type declaration

  • key: string
  • value: any
  • Optional description?: string

Optional customFields

customFields?: Record<string, any> & {
    defaultWorkingMemoryConfig?: any;
    initialUserContext?: Partial<PersonaUserContextDefaults>;
    initialTaskContext?: Partial<PersonaTaskContextDefaults>;
}

Arbitrary extension fields and structured defaults for user/task contexts.

Type declaration

  • Optional defaultWorkingMemoryConfig?: any

    Working memory default config attached at GMI instantiation time.

  • Optional initialUserContext?: Partial<PersonaUserContextDefaults>

    Initial inferred or declared user context values.

  • Optional initialTaskContext?: Partial<PersonaTaskContextDefaults>

    Initial task framing values (e.g., domain, complexity baseline).

Optional requiredSecrets

requiredSecrets?: string[]

Named secrets (API keys, credentials) this persona depends on.

Settings

Member Visibility

Theme

On This Page

idnamelabeldescriptionversionbaseSystemPromptdefaultModelIddefaultProviderIddefaultModelCompletionOptionsmodelTargetPreferencescostSavingStrategypromptEngineConfigOverridestoolIdsallowedCapabilitiesembeddedToolsallowedInputModalitiesallowedOutputModalitiesvoiceConfigavatarConfigpersonalityTraitsmoodAdaptationdefaultLanguageuiInteractionStylememoryConfigcognitiveMemoryConfigconversationContextConfigmetaPromptssentimentTrackingcontextualPromptElementsisCreatorPersonaisPublicactivationKeywordsstrengthsminSubscriptionTierinitialMemoryImprintscustomFieldsrequiredSecrets