| # Phase 0 data model |
|
|
| Entity-relationship diagram for the Supabase schema in [`migrations/20260819200000_init.sql`](migrations/20260819200000_init.sql). |
|
|
| ## ER diagram |
|
|
| ```mermaid |
| erDiagram |
| cats ||--o{ images : has |
| cats ||--o{ evaluation_tasks : has |
| invites ||--|| sessions : "one per link" |
| evaluation_tasks ||--|{ task_candidates : contains |
| images ||--o{ task_candidates : "shown as" |
| sessions ||--o{ feedback : submits |
| evaluation_tasks ||--o{ feedback : "for task" |
| images ||--o{ feedback : "rates" |
| |
| cats { |
| uuid id PK |
| text slug UK |
| text display_name |
| timestamptz created_at |
| } |
| |
| invites { |
| uuid id PK |
| text token UK |
| text evaluator_label |
| boolean is_active |
| timestamptz created_at |
| } |
| |
| sessions { |
| uuid id PK |
| uuid invite_id FK |
| timestamptz created_at |
| } |
| |
| images { |
| uuid id PK |
| uuid cat_id FK |
| text local_path |
| text storage_url |
| text prompt |
| text negative_prompt |
| text base_model |
| text lora_version |
| bigint seed |
| jsonb generation_settings |
| text batch_id |
| timestamptz created_at |
| } |
| |
| evaluation_tasks { |
| uuid id PK |
| uuid cat_id FK |
| text test_id |
| text batch_id |
| timestamptz created_at |
| } |
| |
| task_candidates { |
| uuid task_id PK |
| uuid image_id PK |
| smallint slot UK |
| } |
| |
| feedback { |
| uuid id PK |
| uuid session_id FK |
| uuid task_id FK |
| uuid image_id FK |
| smallint accepted |
| timestamptz submitted_at |
| } |
| ``` |
|
|
| Note: `sessions.invite_id`, `cats.slug`, and `invites.token` are UNIQUE in SQL; Mermaid allows only one of `PK` / `FK` / `UK` per attribute line, so those uniques are documented here and in the relationships table. |
|
|
| ## Relationships |
|
|
| | From | To | Cardinality | Notes | |
| | --- | --- | --- | --- | |
| | `cats` | `images` | 1:N | Each image belongs to one cat | |
| | `cats` | `evaluation_tasks` | 1:N | Each task evaluates one cat | |
| | `invites` | `sessions` | 1:1 | One session per invite link (`invite_id` UNIQUE) | |
| | `evaluation_tasks` | `task_candidates` | 1:4 | Up to four slots per task | |
| | `images` | `task_candidates` | 1:N | Same image can appear in multiple tasks later | |
| | `sessions` + `evaluation_tasks` + `images` | `feedback` | N:1 each | UNIQUE `(session_id, task_id, image_id)` | |
|
|
| ## Review flow |
|
|
| ```text |
| invite (private link) |
| → session (1 per invite) |
| → feedback (4 rows per submit: accepted 0/1) |
| ← tied to evaluation_task + image |
| |
| evaluation_task |
| → task_candidates (slot 1–4) |
| → images (with generation provenance) |
| ← cat |
| ``` |
|
|
| Feedback is append-only: each submission writes one row per candidate image. Resubmit for the same session and task is blocked by the unique constraint. |
|
|
