Feature preview
The rows that need to be filtered, reused, and batch imported in the product parameter table should be made into custom fields, and descriptive paragraphs should be left in the rich text of the main text. Fields are responsible for structured data, and templates are responsible for rendering and default handling. When batch importing, the column names of the table header and fields correspond one to one. Whether the parameter table can be understood by the machine depends on whether the division of labor is clear in this case.
Can the custom content model bear the product parameter table, depending on whether the division of labor is appropriate: the rows that need to be filtered, reused, and batch imported should be made into custom fields, and the content that needs to be explained in segments should remain in the rich text of the main text. The division of labor between fields and templates is also here - fields are responsible for storing what, and templates are responsible for how to display it and what to display when there are missing values. Whether the parameter table can be understood by search engines and large models depends on whether this division is clear, not on how beautiful the table is drawn.
The criteria are three questions: Will this line of data be used for filtering, will it appear repeatedly in multiple products, and will it need to be imported in bulk?
Parameters such as model, material, size, net weight, warranty period, and compatible models, if all three conditions are met, should be made into custom fields. They should appear in the filtering conditions on the list page, as well as on the detail page, comparison page, and search result cards, and be reused in multiple places.
While contents such as 'installation instructions' and 'usage precautions' are not suitable: they expand according to each product, without any filtering value or need for page-to-page reuse, they are more natural in the rich text body. Making them into fields, where they still store entire paragraphs of text, is like using a structured container for unstructured content, which makes it impossible to process the template values.
AnQiCMS supports custom content models and can create different models according to product types: single product model, series model, document model, each with its own set of fields. The boundary should be defined according to the data form, not according to the page sections.
A rule that is easy to follow: information that can be written as 'key-value pairs' should be entered in fields, while information written in 'sentences or paragraphs' should be entered in the text. Values with units (such as material being aluminum alloy, net weight 1.8) should especially be entered in fields, because when the parameter table is referenced, the other party captures this value and its unit, not the entire sentence containing it.
Avoid repeating parameter values in the text. If the same value is stored in two places, changing one and missing the other will quickly result in discrepancies between the list page and the detail page. The field is a single storage location, and the text needs to be rendered through a template when it needs to be referred to.
Parameters are usually taken from tables, AnQiCMS supports two batch import methods: ZIP compressed packages and Excel spreadsheets. The correspondence between columns and fields should be set before import, not after.
Prepare four things before import: the column names in the table header should be consistent with the field names, whether the columns with units should only contain numbers or both numbers and units should be unified, the format of the date column is fixed, and non-mandatory fields are allowed to be empty but cannot contain placeholders. ZIP is suitable for scenarios where images and documents are imported together, while Excel is suitable for batch establishment of pure parameter tables.
After import, check a few articles to see if the fields are in place. The most common mistake in batch writing is that two columns with the same name but different meanings are merged into the same field, and this situation occurs frequently when supplementing data.
The work of the template is to render the field values to the page location and decide what to do with empty values. Three conventions can eliminate most of the rework:
Define default text for each parameter, display 'As per actual' instead of an empty label when the field is empty; only the fields used for filtering are taken on the list page, and the full details are expanded on the detail page to avoid spreading the entire parameter table on the list page; the unit of the same field is added uniformly by the template, and only the value is stored in the field, so no data needs to be reloaded when changing the unit.
When providing data externally, the parameter table can also be reused by other front-end interfaces. The interface document describes the calls to the document list and details, categories and tags, and the field values are returned together with the detail interface, so the mini-program or App terminal does not need to parse the text on the page.
| content | Belonging | Reason |
|---|---|---|
| Model, material, size, net weight | Custom field | Need to filter and reuse in multiple places |
| Warranty period, compatible models | Custom field | Need to import in bulk and structure references |
| Installation instructions,注意事项 | Main text rich text | Continuous narrative, no filtering value |
| Empty value default text | Template | Display rules, not part of the data |
| Unit | Template | Unified changes to avoid repeated storage |
Can existing plain articles be changed to a content model with fields?It can be reclassified according to the model, but the field set must be finalized first. If fields are changed after the launch, old documents need to be filled in item by item, which is more costly than when the final draft is made.
What are the consequences of writing the same number repeatedly in the parameter table and the text?When the two sides are inconsistent, contradictory parameters will appear on the page, and the referencing party will be unable to determine which one is the standard. Let the field be the single storage location, and the text is obtained through a template.
Will the draft status affect the acceptance of the parameter table?It does not affect. The document status is divided into formal, draft, to be published, and recycle bin, the draft can be previewed using a link with the preview parameter, the field values are rendered the same in the preview, and confirmed correctly before becoming formal.
The imported Excel has dozens of columns, does it need to be made into fields?No. Only the columns that need to be filtered and reused should be made into fields, the rest can be directly discarded or merged into the text, the more fields, the greater the template maintenance.
Preview
AI mass production of articles needs to be divided into two processes: generation layer and closure layer: automatic tagging, AI painting matching and TDK keyword description belong to the generation layer, improving efficiency; sensitive word filtering and keyword replacement, content review belong to the closure layer, determining risk. Manual review focuses on the four items most likely to cause problems - tag drift, description inconsistent with the text, matching picture not consistent with the theme, and hitting sensitive words but still entering the queue - and fixes the closure before publication, batch scenarios are inspected by category.
Preview
Parameter binding handles the injection risk of input being executed as a statement, sensitive word filtering handles text content violations, and the two have different points of action, judgment criteria, and consequences. AnQiCMS entrusts the query to the GORM and Iris framework layer convergence, and sets up sensitive word filtering, keyword replacement, and review process on the content side, and the two lines cannot be replaced for each other.
Preview
When AI-generated content and manual review run simultaneously, quality control is handled by three layers: the generation layer is responsible for producing content and automatic tagging, the rule layer filters out definite issues using sensitive word filtering and keyword replacement, and the manual layer relies on draft previews and pending publication status. The sequence is that the rule layer takes effect before the manual layer, and humans only handle the judgment issues left by the rule layer, and mass production can only sustain it.
