What this category's data looks like
Data for this category comes from registered and filed formulas, raw material safety materials, weighing and emulsification equipment, ERP/MES systems, laboratory test records, and packaging material version documents. Daily documents include formula process sheets, weighing and feeding records, emulsification and filling records, cleaning and site clearance records, semi-finished and finished product inspection reports, and label review documents. Data updates are triggered for each production batch or each line change. Document fields include raw material ratios, process parameters, inspection indicators, batch numbers, packaging material information, and more. Field units vary based on document type, with no unified fixed format.
Constraints on the source citation and traceability workflow
High-frequency updates per production batch and line change require traceability data to match the latest version of the current production run. This prevents process deviations caused by citing expired documents. Multi-source, heterogeneous document types require the traceability system to support identification of different document field formats. For example, the raw material ratio field on a formula process sheet and the microbial indicator field on an inspection report must each link to their own dedicated traceability paths. A single production run involves documents across multiple stages. Cited sources must cover the full feeding, emulsification, and inspection workflow. Batch-related information cannot be omitted. Cosmetics filing compliance rules require traceability data to be traced back to official registered and filed documents. This prevents use of non-compliant internal documents.
Configuration Settings
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | The total token count for core process and inspection documents for a single cosmetics manufacturing batch falls within this range. This covers key traceability content and avoids truncation of critical parameters |
responseData | Return citation source details | Display document names, batches, update times, and other traceability fields in responses. This meets compliance review and process traceability requirements |
detail | Enabled | Extract specific document parameters such as emulsification temperature and raw material feeding volume. This supports precise traceability for process Q&A |
| Single search limit | Top 6 entries | Core documents for a single production batch do not exceed 6. Excessive citations cause redundant context and disrupt Q&A logic |
| Maximum context | 2000–3000 characters | Cover associated content across multiple process stages. This ensures coherence of process logic in Q&A |
Three Common Configuration Mistakes
- The symptom is that the citation document list displays in the debug page, but the citation field is empty in the official chat page. The cause is that the return configuration corresponding to
responseDatais not enabled. Traceability data loads only in debug mode, and the setting is not activated in the official environment. - The symptom is that the text content extraction component cannot select knowledge base citations as a data source. The cause is that the component defaults to supporting only native text input. It is not linked to the output fields of the knowledge base traceability module, so content from cited documents cannot be retrieved.
- The symptom is that the number of citations exceeds the preset limit but no truncation is triggered. The cause is that the
quoteMaxTokenparameter is not configured correctly, or the value range does not match the total token count of documents for a single batch. This causes the entry limit to not take effect.
How to Verify Successful Configuration
- Initiate a Q&A regarding a single batch of cosmetics manufacturing processes. Check if traceability information such as associated document names and batch numbers is displayed at the end of the response.
- Adjust the
quoteMaxTokenparameter value. Verify that the number of cited documents in the response changes with the configuration, confirming that the entry limit is active. - Enter debug mode. Initiate a Q&A and check the returned
responseDatafield. Confirm that it contains a complete list of traceability documents. - Use the text content extraction component, select knowledge base citations as the data source. Verify that the component can extract specific field content from cited documents.
The values provided are common starting points. They should be evaluated against the reader's own samples.