Files
product-import-ai/sql/local-scrape-db.sql
T

476 lines
18 KiB
SQL

CREATE DATABASE IF NOT EXISTS `catalog_scrape`
CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;
USE `catalog_scrape`;
CREATE TABLE IF NOT EXISTS `ps_product_catalog_source` (
`id_product_catalog_source` INT UNSIGNED NOT NULL AUTO_INCREMENT,
`source_key` VARCHAR(64) NOT NULL,
`source_name` VARCHAR(128) NOT NULL,
`source_type` VARCHAR(32) NOT NULL DEFAULT '',
`base_url` VARCHAR(512) NOT NULL,
`enabled` TINYINT(1) NOT NULL DEFAULT 1,
`priority` INT NOT NULL DEFAULT 100,
`search_primary_key` VARCHAR(64) NOT NULL DEFAULT 'ean',
`search_fallback_keys` VARCHAR(255) NOT NULL DEFAULT '',
`fallback_source_type` VARCHAR(32) NOT NULL DEFAULT 'none',
`search_url_template` VARCHAR(1000) NOT NULL,
`fallback_url_template` VARCHAR(1000) NOT NULL DEFAULT '',
`search_context_keys` VARCHAR(255) NOT NULL DEFAULT '',
`access_mode` VARCHAR(32) NOT NULL DEFAULT 'browser',
`browser_engine` VARCHAR(32) NOT NULL DEFAULT 'default',
`browser_headless` TINYINT(1) NOT NULL DEFAULT 0,
`wait_after_load_ms` INT UNSIGNED NOT NULL DEFAULT 3000,
`note` TEXT NULL,
PRIMARY KEY (`id_product_catalog_source`),
UNIQUE KEY `uq_product_catalog_source_key` (`source_key`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `ps_product_catalog_manufacturer_source` (
`id_product_catalog_manufacturer_source` INT UNSIGNED NOT NULL AUTO_INCREMENT,
`id_manufacturer` INT UNSIGNED NOT NULL,
`id_product_catalog_source` INT UNSIGNED NOT NULL,
`enabled` TINYINT(1) NOT NULL DEFAULT 1,
`priority` INT NOT NULL DEFAULT 100,
`note` TEXT NULL,
PRIMARY KEY (`id_product_catalog_manufacturer_source`),
UNIQUE KEY `uq_product_catalog_manufacturer_source` (`id_manufacturer`, `id_product_catalog_source`),
KEY `idx_product_catalog_manufacturer_source_source` (`id_product_catalog_source`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_sources` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`legacy_catalog_source_id` BIGINT UNSIGNED NULL,
`source_key` VARCHAR(64) NOT NULL,
`source_name` VARCHAR(128) NOT NULL,
`source_type` ENUM('supplier', 'manufacturer', 'backup', 'other') NOT NULL DEFAULT 'supplier',
`base_url` VARCHAR(1024) NOT NULL,
`enabled` TINYINT(1) NOT NULL DEFAULT 1,
`priority` INT NOT NULL DEFAULT 100,
`search_primary_key` VARCHAR(64) NULL,
`search_fallback_keys` VARCHAR(255) NULL,
`fallback_source_type` VARCHAR(64) NOT NULL DEFAULT 'none',
`search_url_template` VARCHAR(2048) NULL,
`fallback_url_template` VARCHAR(2048) NULL,
`search_context_keys` VARCHAR(255) NULL,
`access_mode` VARCHAR(64) NOT NULL DEFAULT 'browser',
`browser_engine` VARCHAR(32) NOT NULL DEFAULT 'default',
`browser_headless` TINYINT(1) NOT NULL DEFAULT 0,
`wait_after_load_ms` INT UNSIGNED NOT NULL DEFAULT 3000,
`scraper_config` JSON NULL,
`request_delay_ms` INT UNSIGNED NOT NULL DEFAULT 3000,
`last_health_status` SMALLINT UNSIGNED NULL,
`last_health_checked_at` DATETIME NULL,
`note` TEXT NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
`updated_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
UNIQUE KEY `uq_scrape_sources_legacy` (`legacy_catalog_source_id`),
UNIQUE KEY `uq_scrape_sources_key` (`source_key`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_manufacturer_sources` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`manufacturer_id` INT UNSIGNED NOT NULL,
`source_id` BIGINT UNSIGNED NOT NULL,
`enabled` TINYINT(1) NOT NULL DEFAULT 1,
`priority` INT NOT NULL DEFAULT 100,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
`updated_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
UNIQUE KEY `uq_scrape_manufacturer_source` (`manufacturer_id`, `source_id`),
KEY `idx_scrape_manufacturer_source_source` (`source_id`),
CONSTRAINT `fk_scrape_manufacturer_source_source` FOREIGN KEY (`source_id`) REFERENCES `scrape_sources` (`id`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_source_capabilities` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`source_id` BIGINT UNSIGNED NOT NULL,
`capability` ENUM('search', 'pictures', 'texts', 'features') NOT NULL,
`enabled` TINYINT(1) NOT NULL DEFAULT 1,
`priority` INT NOT NULL DEFAULT 100,
`parser_key` VARCHAR(128) NOT NULL,
`parser_module` VARCHAR(255) NULL,
`browser_engine` VARCHAR(32) NULL,
`browser_headless` TINYINT(1) NULL,
`wait_after_load_ms` INT UNSIGNED NULL,
`min_width` INT UNSIGNED NULL,
`min_height` INT UNSIGNED NULL,
`min_longest_side` INT UNSIGNED NULL,
`default_target_strategy` ENUM('unknown', 'color', 'size', 'combination', 'manual') NOT NULL DEFAULT 'unknown',
`selector_config` JSON NULL,
`extraction_config` JSON NULL,
`quality_config` JSON NULL,
`note` TEXT NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
`updated_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
UNIQUE KEY `uq_scrape_source_capability_parser` (`source_id`, `capability`, `parser_key`),
KEY `idx_scrape_source_capabilities_lookup` (`source_id`, `capability`, `enabled`, `priority`),
CONSTRAINT `fk_scrape_source_capabilities_source` FOREIGN KEY (`source_id`) REFERENCES `scrape_sources` (`id`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_manufacturer_source_capabilities` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`manufacturer_source_id` BIGINT UNSIGNED NOT NULL,
`source_capability_id` BIGINT UNSIGNED NOT NULL,
`enabled` TINYINT(1) NOT NULL DEFAULT 1,
`priority` INT NOT NULL DEFAULT 100,
`target_strategy_override` ENUM('unknown', 'color', 'size', 'combination', 'manual') NULL,
`config_override` JSON NULL,
`note` TEXT NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
`updated_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
UNIQUE KEY `uq_scrape_manufacturer_source_capability` (`manufacturer_source_id`, `source_capability_id`),
KEY `idx_scrape_manufacturer_source_capabilities_capability` (`source_capability_id`),
CONSTRAINT `fk_scrape_manufacturer_source_capabilities_mapping` FOREIGN KEY (`manufacturer_source_id`) REFERENCES `scrape_manufacturer_sources` (`id`),
CONSTRAINT `fk_scrape_manufacturer_source_capabilities_capability` FOREIGN KEY (`source_capability_id`) REFERENCES `scrape_source_capabilities` (`id`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_products` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`manufacturer_id` INT UNSIGNED NULL,
`id_product` BIGINT UNSIGNED NULL,
`supplier_reference` VARCHAR(255) NULL,
`ean13` VARCHAR(32) NULL,
`color` VARCHAR(255) NULL,
`source_id` BIGINT UNSIGNED NOT NULL,
`source_url` VARCHAR(2048) NOT NULL,
`match_method` ENUM('ean', 'reference', 'name', 'manual') NOT NULL DEFAULT 'ean',
`source_product_key` VARCHAR(255) NULL,
`product_name` TEXT NULL,
`raw_data` JSON NULL,
`parsed_data` JSON NULL,
`status` ENUM('found', 'parsed', 'error', 'manual') NOT NULL DEFAULT 'found',
`error_message` TEXT NULL,
`fetched_at` DATETIME NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
`updated_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
KEY `idx_scrape_products_ean` (`ean13`),
KEY `idx_scrape_products_reference` (`supplier_reference`),
KEY `idx_scrape_products_source` (`source_id`),
CONSTRAINT `fk_scrape_products_source` FOREIGN KEY (`source_id`) REFERENCES `scrape_sources` (`id`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_assets` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`scrape_product_id` BIGINT UNSIGNED NOT NULL,
`asset_type` ENUM('image', 'document', 'other') NOT NULL DEFAULT 'image',
`asset_url` VARCHAR(2048) NOT NULL,
`local_path` VARCHAR(1024) NULL,
`content_hash` CHAR(64) NULL,
`position` INT UNSIGNED NOT NULL DEFAULT 0,
`alt_text` VARCHAR(512) NULL,
`status` ENUM('discovered', 'downloaded', 'error') NOT NULL DEFAULT 'discovered',
`error_message` TEXT NULL,
`fetched_at` DATETIME NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
KEY `idx_scrape_assets_product` (`scrape_product_id`),
CONSTRAINT `fk_scrape_assets_product` FOREIGN KEY (`scrape_product_id`) REFERENCES `scrape_products` (`id`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_image_blobs` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`sha256` CHAR(64) NOT NULL,
`perceptual_hash` CHAR(16) NULL,
`mime_type` VARCHAR(64) NOT NULL,
`width` INT UNSIGNED NULL,
`height` INT UNSIGNED NULL,
`size_bytes` INT UNSIGNED NOT NULL,
`image_data` LONGBLOB NOT NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
UNIQUE KEY `uq_scrape_image_blobs_sha256` (`sha256`),
KEY `idx_scrape_image_blobs_phash` (`perceptual_hash`),
KEY `idx_scrape_image_blobs_size` (`width`, `height`, `size_bytes`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_picture_candidates` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`image_blob_id` BIGINT UNSIGNED NOT NULL,
`source_id` BIGINT UNSIGNED NOT NULL,
`scrape_product_id` BIGINT UNSIGNED NULL,
`manufacturer_id` INT UNSIGNED NULL,
`id_product_catalog` BIGINT UNSIGNED NULL,
`id_product` BIGINT UNSIGNED NULL,
`supplier_reference` VARCHAR(255) NULL,
`ean13` VARCHAR(32) NULL,
`color` VARCHAR(255) NULL,
`size_value` VARCHAR(128) NULL,
`combi` VARCHAR(255) NULL,
`source_page_url` VARCHAR(2048) NULL,
`source_image_url` VARCHAR(2048) NOT NULL,
`source_position` INT UNSIGNED NOT NULL DEFAULT 0,
`target_strategy` ENUM('unknown', 'color', 'size', 'combination', 'manual') NOT NULL DEFAULT 'unknown',
`target_color` VARCHAR(255) NULL,
`target_size` VARCHAR(128) NULL,
`target_combi` VARCHAR(255) NULL,
`target_ean13` VARCHAR(32) NULL,
`target_id_product_catalog` BIGINT UNSIGNED NULL,
`match_confidence` DECIMAL(5,2) NOT NULL DEFAULT 0.00,
`duplicate_status` ENUM('unique', 'exact_duplicate', 'visual_duplicate', 'review') NOT NULL DEFAULT 'review',
`duplicate_of_candidate_id` BIGINT UNSIGNED NULL,
`status` ENUM('draft', 'approved', 'rejected') NOT NULL DEFAULT 'draft',
`review_note` TEXT NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
`updated_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
UNIQUE KEY `uq_scrape_picture_candidate_source_image` (`source_id`, `source_image_url`(512)),
KEY `idx_scrape_picture_candidates_blob` (`image_blob_id`),
KEY `idx_scrape_picture_candidates_source` (`source_id`),
KEY `idx_scrape_picture_candidates_product` (`manufacturer_id`, `supplier_reference`, `ean13`),
KEY `idx_scrape_picture_candidates_target` (`target_strategy`, `target_color`, `target_size`, `target_combi`),
KEY `idx_scrape_picture_candidates_duplicate` (`duplicate_status`, `duplicate_of_candidate_id`),
CONSTRAINT `fk_scrape_picture_candidates_blob` FOREIGN KEY (`image_blob_id`) REFERENCES `scrape_image_blobs` (`id`),
CONSTRAINT `fk_scrape_picture_candidates_source` FOREIGN KEY (`source_id`) REFERENCES `scrape_sources` (`id`),
CONSTRAINT `fk_scrape_picture_candidates_product` FOREIGN KEY (`scrape_product_id`) REFERENCES `scrape_products` (`id`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_picture_target_rules` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`manufacturer_id` INT UNSIGNED NULL,
`source_id` BIGINT UNSIGNED NULL,
`id_category_default` INT UNSIGNED NULL,
`catalog_category` VARCHAR(255) NULL,
`product_name_pattern` VARCHAR(255) NULL,
`target_strategy` ENUM('color', 'size', 'combination', 'manual') NOT NULL,
`confidence` DECIMAL(5,2) NOT NULL DEFAULT 0.00,
`examples_count` INT UNSIGNED NOT NULL DEFAULT 0,
`note` TEXT NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
`updated_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
KEY `idx_scrape_picture_target_rules_lookup` (`manufacturer_id`, `source_id`, `id_category_default`),
CONSTRAINT `fk_scrape_picture_target_rules_source` FOREIGN KEY (`source_id`) REFERENCES `scrape_sources` (`id`)
) ENGINE=InnoDB;
CREATE TABLE IF NOT EXISTS `scrape_runs` (
`id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
`source_id` BIGINT UNSIGNED NULL,
`manufacturer_id` INT UNSIGNED NULL,
`started_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
`finished_at` DATETIME NULL,
`status` ENUM('running', 'completed', 'failed', 'cancelled') NOT NULL DEFAULT 'running',
`request_count` INT UNSIGNED NOT NULL DEFAULT 0,
`success_count` INT UNSIGNED NOT NULL DEFAULT 0,
`error_count` INT UNSIGNED NOT NULL DEFAULT 0,
`notes` TEXT NULL,
`created_at` DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
PRIMARY KEY (`id`),
KEY `idx_scrape_runs_source` (`source_id`),
CONSTRAINT `fk_scrape_runs_source` FOREIGN KEY (`source_id`) REFERENCES `scrape_sources` (`id`)
) ENGINE=InnoDB;
INSERT INTO `scrape_source_capabilities` (
`source_id`,
`capability`,
`enabled`,
`priority`,
`parser_key`,
`parser_module`,
`browser_engine`,
`browser_headless`,
`wait_after_load_ms`,
`min_longest_side`,
`default_target_strategy`,
`selector_config`,
`extraction_config`,
`quality_config`,
`note`
)
SELECT
`id`,
'pictures',
1,
10,
'hudy-gallery',
'hudy.picture-gallery',
NULL,
NULL,
NULL,
800,
'color',
JSON_OBJECT(
'colorSwitcherSelector', 'a[id^="product-change-color-"]',
'eanTableSelector', 'table.product-params__params-table',
'topImageSelector', 'p#snippet-productInfo-image img.product-top__img',
'gallerySelector', 'p#snippet-productInfo-image, .pdbox__window'
),
JSON_OBJECT(
'verifyVariantBy', JSON_ARRAY('ean', 'color'),
'pictureSource', 'largeGallery',
'dedupeBy', JSON_ARRAY('sha256', 'perceptual_hash')
),
JSON_OBJECT('minLongestSide', 800, 'allowSquareCrop', true),
'Hudy product gallery pictures. Verify the color variant by EAN before extracting large gallery images.'
FROM `scrape_sources`
WHERE `source_key` = 'hudy'
ON DUPLICATE KEY UPDATE
`enabled` = VALUES(`enabled`),
`priority` = VALUES(`priority`),
`parser_module` = VALUES(`parser_module`),
`min_longest_side` = VALUES(`min_longest_side`),
`default_target_strategy` = VALUES(`default_target_strategy`),
`selector_config` = VALUES(`selector_config`),
`extraction_config` = VALUES(`extraction_config`),
`quality_config` = VALUES(`quality_config`),
`note` = VALUES(`note`);
INSERT INTO `scrape_source_capabilities` (
`source_id`,
`capability`,
`enabled`,
`priority`,
`parser_key`,
`parser_module`,
`browser_engine`,
`browser_headless`,
`wait_after_load_ms`,
`default_target_strategy`,
`selector_config`,
`extraction_config`,
`note`
)
SELECT
`id`,
'texts',
1,
30,
'hudy-product-texts',
'hudy.product-texts',
NULL,
NULL,
NULL,
'manual',
JSON_OBJECT('descriptionSelector', '.product-detail, .product-top, [itemprop="description"]'),
JSON_OBJECT('fields', JSON_ARRAY('shortDescription', 'description', 'marketingText')),
'Hudy text extraction placeholder. User will teach exact text fields before automation is trusted.'
FROM `scrape_sources`
WHERE `source_key` = 'hudy'
ON DUPLICATE KEY UPDATE
`enabled` = VALUES(`enabled`),
`priority` = VALUES(`priority`),
`parser_module` = VALUES(`parser_module`),
`selector_config` = VALUES(`selector_config`),
`extraction_config` = VALUES(`extraction_config`),
`note` = VALUES(`note`);
INSERT INTO `scrape_source_capabilities` (
`source_id`,
`capability`,
`enabled`,
`priority`,
`parser_key`,
`parser_module`,
`browser_engine`,
`browser_headless`,
`wait_after_load_ms`,
`default_target_strategy`,
`selector_config`,
`extraction_config`,
`note`
)
SELECT
`id`,
'features',
1,
20,
'hudy-product-params',
'hudy.product-params',
NULL,
NULL,
NULL,
'manual',
JSON_OBJECT('paramsTableSelector', 'table.product-params__params-table'),
JSON_OBJECT('fields', JSON_ARRAY('label', 'value'), 'mergeStrategy', 'review-first'),
'Hudy feature extraction placeholder. Product parameters are kept reviewable before catalog writes.'
FROM `scrape_sources`
WHERE `source_key` = 'hudy'
ON DUPLICATE KEY UPDATE
`enabled` = VALUES(`enabled`),
`priority` = VALUES(`priority`),
`parser_module` = VALUES(`parser_module`),
`selector_config` = VALUES(`selector_config`),
`extraction_config` = VALUES(`extraction_config`),
`note` = VALUES(`note`);
INSERT INTO `scrape_source_capabilities` (
`source_id`,
`capability`,
`enabled`,
`priority`,
`parser_key`,
`parser_module`,
`browser_engine`,
`browser_headless`,
`wait_after_load_ms`,
`min_longest_side`,
`default_target_strategy`,
`selector_config`,
`extraction_config`,
`quality_config`,
`note`
)
SELECT
`id`,
'pictures',
1,
20,
'idealo-splide-gallery',
'idealo.splide-gallery',
'firefox',
1,
3000,
800,
'color',
JSON_OBJECT(
'galleryTrackSelector', 'div.splide__track.container',
'slideImageSelector', 'ul.splide__list li.splide__slide img'
),
JSON_OBJECT(
'urlAttributes', JSON_ARRAY('srcset', 'data-srcset', 'data-large', 'data-original', 'data-src', 'src'),
'pictureSource', 'productGalleryOnly',
'rejectSearchThumbnails', true,
'dedupeBy', JSON_ARRAY('sha256', 'perceptual_hash')
),
JSON_OBJECT('minLongestSide', 800, 'allowSquareCrop', true),
'Idealo is primarily an image backup source. Use product gallery images only, not search-result thumbnails.'
FROM `scrape_sources`
WHERE `source_key` = 'idealo'
ON DUPLICATE KEY UPDATE
`enabled` = VALUES(`enabled`),
`priority` = VALUES(`priority`),
`parser_module` = VALUES(`parser_module`),
`browser_engine` = VALUES(`browser_engine`),
`browser_headless` = VALUES(`browser_headless`),
`wait_after_load_ms` = VALUES(`wait_after_load_ms`),
`min_longest_side` = VALUES(`min_longest_side`),
`default_target_strategy` = VALUES(`default_target_strategy`),
`selector_config` = VALUES(`selector_config`),
`extraction_config` = VALUES(`extraction_config`),
`quality_config` = VALUES(`quality_config`),
`note` = VALUES(`note`);
INSERT INTO `scrape_manufacturer_source_capabilities` (
`manufacturer_source_id`,
`source_capability_id`,
`enabled`,
`priority`,
`target_strategy_override`
)
SELECT
`sms`.`id`,
`sc`.`id`,
`sc`.`enabled`,
`sc`.`priority`,
NULL
FROM `scrape_manufacturer_sources` `sms`
JOIN `scrape_source_capabilities` `sc`
ON `sc`.`source_id` = `sms`.`source_id`
ON DUPLICATE KEY UPDATE
`enabled` = VALUES(`enabled`),
`priority` = VALUES(`priority`);