What Data Looks Like for This Category
Daily documentation for cosmetics production includes formula process sheets, weighing and feeding records, emulsification or filling records, cleaning and site clearance records, semi-finished and finished product inspection reports, and label review documents. Data sources include registered or filed formulas, raw material safety materials, MES/ERP systems of production equipment, laboratory test data, and packaging material version archives. Data updates are triggered with each production batch and each line change. Document fields include raw material names, restricted substance content, production parameters, batch numbers, packaging material claim content, and more. The length of individual documents varies. Filing-type documents must strictly match officially published content.
Constraints Imposed on Source Citation and Traceability
Because data is updated with each production batch, it is necessary to accurately associate documents corresponding to the current production batch to avoid cross-batch citation errors. There are multiple document types that must strictly meet filing requirements. The traceability link must clearly distinguish sources of raw material filing information, production records, and inspection data. Individual filing documents are relatively long, so key content must be fully cited to support process review and exception reports. Priority should be given to tracing sensory and stability-related production records, which matches the core business needs of cosmetics production.
Configuration Settings
The following configuration is based on FastGPT v4.17.0.
| Configuration Item | Suggested Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Cosmetics filing process documents are relatively long individually, and key content such as raw material restricted substances and formula ratios must be fully retained |
responseData | Enable full traceability mode | The original number and source path of cited documents must be displayed in process documents and exception reports |
detail | Enable batch-associated retrieval | Production batch numbers must be used as core retrieval keywords to match production and inspection records for the current batch |
| Single search limit | Top 5 entries | The number of core documents associated with a single production batch usually does not exceed 5, to avoid redundant content interfering with report generation |
| Knowledge base chunk length | 800–1000 characters | Adapt to the paragraph structure of cosmetics filing documents to ensure that key parameters can be fully matched during retrieval |
Three Common Mistakes
- After calling the FastGPT chat interface, the returned chat results do not include the cited knowledge base document ID and source path. The cause is that the full traceability configuration of
responseDatais not enabled, so the interface does not return source association fields. - The raw material information cited in the generated exception report does not match the current production batch, and cross-batch citation of historical data occurs. The cause is that production batch numbers are not used as retrieval keywords, resulting in recall of document data from non-current batches.
- The content of the cited raw material filing document is truncated, and the filing number and safety material details of restricted substances cannot be fully displayed. The cause is that the
quoteMaxTokensetting value is too small, exceeding the citation length limit for individual documents.
How to Confirm the Configuration Is Correct
- Initiate a query about the raw material filing information for a specific production batch of cosmetics, and check whether the returned results include the original document number and source path.
- Check the
quoteMaxTokenconfiguration value to confirm that the length of a single cited content can cover the key information of the filing document. - Call the chat interface to verify that the returned
responseDatafield includes detailed entries of the cited source. - Test the exception report generation function to confirm that the cited cleaning records and inspection data are all associated with the correct production batch numbers.
The values provided here are common starting points and should be evaluated against the reader's own samples.