Confirm the data shape — related words

erDiagram
    SENSE_CARD ||--o{ RELATION_ITEM : "owns"
    RELATION_TARGET ||--o{ RELATION_ITEM : "is mentioned by"
    RELATION_TARGET ||--o{ TARGET_GLOSS : "general meaning, one row per language"
    RELATION_ITEM ||--o{ ITEM_GLOSS : "sense-specific meaning, one row per language"
    RELATION_TARGET ||--o{ RELATION_EXAMPLE : "usage, Indonesian"
    RELATION_EXAMPLE ||--o{ EXAMPLE_TEXT : "translation, one row per language"
    RELATION_ITEM ||--o{ ITEM_NOTE : "context note, one row per language"
    RELATION_TARGET }o..o| FULL_CARD : "link when a full card exists"
    RELATION_TARGET }o..o| ENRICH_QUEUE_ROW : "one queue row per target"
    LANGUAGE ||--o{ TARGET_GLOSS : "lang"
    LANGUAGE ||--o{ ITEM_GLOSS : "lang"
    LANGUAGE ||--o{ EXAMPLE_TEXT : "lang"
    LANGUAGE ||--o{ ITEM_NOTE : "lang"
    RELATION_TARGET {
        uuid target_id PK
        text l2_language FK
        text text_norm "unique with l2_language"
        text text_display
        text kind "word | phrase — authored"
        text register_intrinsic "nullable; slang etc. when true of the word itself"
        text ipa
        text respelling
        text audio_normal_url
        text audio_slow_url
        uuid full_card_id FK "nullable"
        text full_card_status "linked | none | queued | refused"
        bigint enrich_queue_id FK "nullable"
        jsonb provenance "model, version, authored_at — opaque, stored whole"
    }
    RELATION_ITEM {
        uuid item_id PK
        uuid source_card_id FK "the owning CT2 sense"
        uuid target_id FK
        text relation_kind "similar | opposite | informal | related_compound"
        int ordinal
        text register "formal | informal | slang | neutral — of the relationship"
        bool is_idiom
        text tag_raw "jenis as authored, kept for the record"
        jsonb provenance
    }
    TARGET_GLOSS {
        uuid target_id FK
        text lang FK
        text text
        text state "machine | llm | human | locked"
    }
    ITEM_GLOSS {
        uuid item_id FK
        text lang FK
        text text
        text state
    }
    RELATION_EXAMPLE {
        uuid example_id PK
        uuid target_id FK
        int ordinal
        text l2_text
    }
    EXAMPLE_TEXT {
        uuid example_id FK
        text lang FK
        text text
        text state
    }
    ITEM_NOTE {
        uuid item_id FK
        text lang FK
        text text
        text state
    }
    

flowchart TD
    TAP["Learner taps the standard add chip on an item"] --> Q{"target_status?"}
    Q -- linked --> ADD["POST /api/learning-list/add with sense_id"]
    ADD --> T310["t310 row, user_status = priority: next session"]
    Q -- "no_card" --> ENQ["POST /api/learning-list/add with the item text"]
    ENQ --> QUEUE["t_enrich_queue row: queue_type headword, status pending, requested_by = profile"]
    QUEUE --> ITEM["item.target_status = queued"]
    QUEUE --> DAEMON["enrichment daemon"]
    DAEMON -- word --> CARD["catalog card created"]
    DAEMON -- phrase --> PP{"phrase path: corpus phrase?"}
    PP -- yes --> CARD
    PP -- no --> REF["refused today: eight corpus-grounded steps refuse a non-corpus phrase (desk item G #334)"]
    CARD --> LINK["item.target_card_id filled, status linked"]
    LINK --> AUTO{"auto-add to the requester's list?"}
    AUTO -- "built" --> T310
    AUTO -- "not built today" --> PEND["learner sees pending; must tap again"]
    Q -- queued --> WAIT["chip shows pending"]
    Q -- refused --> HIDE["chip hidden or greyed"]