Batch Table

Overview

Stores one project-scoped batch of up to 500 requests. Supported operations are Page Import, Code2Des, and Page Discovery (one homepage per batch).

Table Definition

Logical Name	Physical Name	Column Name	Data Type	Primary Key	Relation	Unique	Nullable	Default Value	Remarks
Batch	batch	id	string	◯					Client UUID / クライアント UUID (varchar(255))
		organization_id	number		organization:id				Organization scope / 組織
		project_id	number		project:id				Project scope / プロジェクト
		type	string						page_import, code2des, or page_discovery (varchar(32))
		request_hash	string						SHA-256 of normalized input or retry source batch / 入力または再試行元バッチの SHA-256 (varchar(64))
		created_at	datetime					CURRENT_TIMESTAMP	Shared audit field / 共通監査項目
		updated_at	datetime					CURRENT_TIMESTAMP	Shared audit field / 共通監査項目
		deleted_at	datetime				◯		Soft delete / 論理削除日時
		created_by	string						Cognito subject (varchar(50))
		updated_by	string						Cognito subject (varchar(50))
		deleted_by	string				◯		Cognito subject (varchar(50))

Relations

• organization_id → organization.id (delete cascade)
• project_id → project.id (delete cascade)
• batch_item.batch_id → batch.id (delete cascade)

Indexes

• PRIMARY KEY (id)
• INDEX batch_org_project_idx (organization_id, project_id)

Notes

• batch and batch_item are shared across operations. Adding an operation requires request validation and a registered job handler, not new batch tables.
• services/batch-operations.ts registers job preparation, status lookup, retry validation, and dead-letter handling. Feature job tables remain the source of processing status and results.
• The existing /page-batches API and internal dispatcher path remain unchanged for the plugin and schedule.
• Aggregate status is derived from child jobs and dispatch metadata; counters are not stored separately.
• Dispatch claims expire after 90 seconds. Row locks prevent concurrent claims. Queue delivery is attempted up to three times using the same child ID and message.
• Workers do not connect to PostgreSQL. Existing webhooks update Page Import and Code2Des completion. Page Discovery publishes status and URL previews to S3; its registered handler reads those results without an additional job table.