Create things and things_facts tables for object/software/hardware tracking

[NOVA-Openclaw]

Summary

We have entities + entity_facts for people/agents and places + place_properties for locations, but no table for objects/things.

Things = catch-all for any noun that isn't an entity or a place.

Noun Categories

Entities (people, beings, organizations)

• People: John, I)ruid, Satoshi
• Agents: NOVA, Newhart, Coder
• Organizations: corporations, companies, agencies
• Beings: gods, spirits, ghosts, extraterrestrial beings, fictional characters

Places (physical/virtual locations)

• Countries, states, territories
• Cities, towns, villages
• Buildings, facilities, addresses
• Networks, servers, data centers

Things (everything else)

• Software: OpenClaw, signal-cli, bird-x
• Hardware: computers, servers, phones, Hue bridges
• Vehicles: cars, boats, Camper de Verde
• Documents: certificates, contracts, books
• Objects: furniture, tools, rocks, anything physical

Proposed Schema

CREATE TABLE things (
    id SERIAL PRIMARY KEY,
    name VARCHAR(255) NOT NULL UNIQUE,
    type VARCHAR(50),  -- software, hardware, vehicle, document, etc.
    description TEXT,
    owner_entity_id INTEGER REFERENCES entities(id),
    location_place_id INTEGER REFERENCES places(id),
    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

CREATE TABLE things_facts (
    id SERIAL PRIMARY KEY,
    thing_id INTEGER REFERENCES things(id) ON DELETE CASCADE,
    key VARCHAR(255) NOT NULL,
    value TEXT NOT NULL,
    data JSONB,
    data_type VARCHAR(20) DEFAULT 'observation',
    source VARCHAR(255),
    learned_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
    updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

⚠️ Semantic Embedding Triggers (REQUIRED)

New tables MUST auto-queue for semantic embedding on INSERT, like entity_facts does:

-- Trigger for things table
CREATE OR REPLACE FUNCTION queue_thing_embedding() RETURNS TRIGGER AS $$
BEGIN
    INSERT INTO embedding_queue (source_type, source_id, content)
    VALUES ('thing', NEW.id, NEW.name || ': ' || COALESCE(NEW.description, ''));
    RETURN NEW;
END;
$$ LANGUAGE plpgsql;

CREATE TRIGGER trg_embed_thing
AFTER INSERT ON things
FOR EACH ROW EXECUTE FUNCTION queue_thing_embedding();

-- Trigger for things_facts table  
CREATE OR REPLACE FUNCTION queue_thing_fact_embedding() RETURNS TRIGGER AS $$
BEGIN
    INSERT INTO embedding_queue (source_type, source_id, content)
    VALUES ('thing_fact', NEW.id, NEW.key || ': ' || NEW.value);
    RETURN NEW;
END;
$$ LANGUAGE plpgsql;

CREATE TRIGGER trg_embed_thing_fact
AFTER INSERT ON things_facts
FOR EACH ROW EXECUTE FUNCTION queue_thing_fact_embedding();

Also add triggers for existing places table (currently missing):

CREATE TRIGGER trg_embed_place AFTER INSERT ON places ...
CREATE TRIGGER trg_embed_place_property AFTER INSERT ON place_properties ...

Memory Extractor Integration

Update grammar_parser/store_relations.py to route nouns to correct table.

Related Issues

• Add noun classification to route extracted nouns to correct tables #57 (noun classification routing)
• nova-relationships#10 (unified relationships table)