memory-taxonomy

name	memory-taxonomy
description	Home domain memory taxonomy for service-provider entity resolution, subjects, predicates, permanence, tags, and example facts.
version	1.0.0
tools_required	["memory_entity_resolve","memory_entity_create","memory_store_fact"]

Home Memory Taxonomy Skill

Purpose

Load this skill when storing home-domain memory facts, especially facts about service providers that must be anchored to resolved entities.

Service Providers: Resolve Before Storing

When the user mentions a home service provider (plumber, electrician, HVAC technician, cleaning company, etc.), resolve or create a transitory entity before storing facts about them. Never store facts with only a raw string subject for external organizations or people.

Entity type inference for home domain:

Home entity	entity_type
Plumber, electrician, HVAC tech, contractor	person or organization (use organization if a company name; person if an individual)
Cleaning service, pest control, landscaping company	organization
Appliance manufacturer or brand	organization
Individual tradesperson (e.g., "Mike the plumber")	person

Resolve-or-create pattern for service providers:

# "Called Mike's Plumbing to fix the leaking pipe under the kitchen sink"
candidates = memory_entity_resolve(name="Mike's Plumbing", entity_type="organization")
# → zero candidates: create transitory entity
try:
    result = memory_entity_create(
        canonical_name="Mike's Plumbing",
        entity_type="organization",
        metadata={
            "unidentified": True,
            "source": "fact_storage",
            "source_butler": "home",
            "source_scope": "home"
        }
    )
    provider_entity_id = result["entity_id"]
except ValueError:
    candidates = memory_entity_resolve(name="Mike's Plumbing", entity_type="organization")
    provider_entity_id = candidates[0]["entity_id"]

memory_store_fact(
    subject="Mike's Plumbing",
    predicate="service_provider",
    content="plumbing — fixed kitchen sink leak; reliable, called for emergencies",
    entity_id=provider_entity_id,
    permanence="stable",
    importance=6.0,
    tags=["service-provider", "plumbing", "maintenance"]
)

The entity appears in the dashboard "Unidentified Entities" section for the owner to confirm. Never fall back to a bare string subject for a service provider.

Room, device, and scene subjects (e.g., "bedroom", "thermostat", "movie-night") are internal identifiers; they do not require entity resolution.

Home Domain Taxonomy

Subject:

• For room-specific knowledge: room name (e.g., "bedroom", "living-room", "kitchen"), no entity required
• For device-specific knowledge: device identifier (e.g., "thermostat", "front-door-lock"), no entity required
• For scene knowledge: scene name (e.g., "movie-night", "bedtime"), no entity required
• For user preferences: "comfort_preference", "energy_preference", no entity required
• For service providers: company/person name; it MUST be resolved to an entity (see above)

Predicates:

• comfort_preference: User's temperature, humidity, lighting, or air quality preferences
• comfort_deviation: Detected deviation from user's comfort preferences (temporary alert)
• scene_preference: User's preferences for scene timing, trigger conditions, or modifications; also used when a scene is created or modified
• automation_schedule: A scheduled automation linked to a scene or recurring action
• schedule_pattern: Observed patterns in room usage or device activation (e.g., "living room always used 7-10pm")
• device_issue: Known device problems, quirks, maintenance needs, or firmware history (use tags to distinguish: battery, offline, firmware, quirk, maintenance)
• energy_baseline: Typical energy consumption by device or time period (used for anomaly detection)
• energy_spike: Anomalous energy consumption detected above baseline (volatile)
• energy_pattern: Observed patterns in energy consumption over time (standard)
• usage_pattern: Observed patterns in how user interacts with devices or scenes
• service_provider: Known home service providers such as plumbers, electricians, cleaners, and contractors (fact anchored to service provider entity)

Permanence levels:

• stable: Long-term preferences that persist across seasons and living patterns (e.g., "user prefers bedroom at 68°F at night")
• standard: Current preferences and typical patterns (e.g., "user usually activates movie night at 7pm on weekends")
• volatile: Temporary states, immediate issues, or time-sensitive alerts (e.g., "basement sensor battery at 15%", "HVAC firmware update available")

Tags: Use tags like temperature, humidity, lighting, energy, comfort, scene, device, maintenance, urgent, seasonal, service-provider

Example Facts

# From: "I like the bedroom cooler at night around 68 degrees"
memory_store_fact(
    subject="bedroom",
    predicate="comfort_preference",
    content="user prefers 68°F (67-69°F range) at night for sleeping",
    permanence="stable",
    importance=8.0,
    tags=["temperature", "comfort", "bedroom", "night"]
)

# From: observing user activates movie night every Friday at 7pm
memory_store_fact(
    subject="movie-night-scene",
    predicate="usage_pattern",
    content="user typically activates movie night scene on Friday evenings around 7pm",
    permanence="standard",
    importance=6.0,
    tags=["pattern", "scene", "movie-night", "weekend"]
)

# From: device status check showing basement sensor battery at 15%
memory_store_fact(
    subject="basement-sensor",
    predicate="device_issue",
    content="basement sensor battery at 15% — needs replacement soon",
    permanence="volatile",
    importance=7.0,
    tags=["maintenance", "battery", "urgent"]
)

# From: analyzing energy consumption data
memory_store_fact(
    subject="hvac",
    predicate="energy_baseline",
    content="HVAC typically uses 40% of daily energy in winter, 25% in summer. Peak usage 7-9am and 6-8pm.",
    permanence="standard",
    importance=6.0,
    tags=["energy", "hvac", "baseline"]
)

# From: "I like it bright in the kitchen during the day"
memory_store_fact(
    subject=,
    predicate=,
    content=,
    permanence=,
    importance=,
    tags=[, , , ]
)