What Data for This Category Looks Like
This category’s data covers animal quarantine certificates, slaughter and cutting records, ingredient marinating records, thermal processing records, cold chain temperature records, microbial test reports, and more. Data sources include animal quarantine information, supplier batch archives, production traceability ledgers, real-time temperature collection data, metal detection results, and laboratory test reports. Updates follow production batches and shift-level operation cycles. Cold chain temperature data is updated via continuous collection. Document fields include batch numbers, quarantine numbers, temperature thresholds, test indicators and results, and some records must be bound to unique identifiers for corresponding raw materials or finished products.
Constraints for Source Citation and Traceability
Data for this category must be bound to unique batch identifiers. When citing for traceability, ensure each cited item can be traced back to the original record of the corresponding production batch. This requires precise matching of the association between sources and cited content. Continuously collected cold chain temperature data has a high update frequency, so support for real-time or near-real-time retrieval is needed to avoid citing expired data. The strict binding between raw material quarantine and production records requires prioritizing association of multiple document types from the same batch during retrieval, to prevent traceability errors from cross-batch citations. The production process of separating raw and cooked materials also requires distinguishing between raw material and finished product record categories during citation, to prevent traceability confusion related to cross-contamination. Data field specificity—such as temperature values and test results—must accurately correspond to original collection nodes, and generalized citations are not allowed.
Configuration Settings
This configuration is based on the functional logic of FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 8000–12000 token | The total length of documents related to a single batch for this category is relatively high, requiring coverage of complete content across multiple record types including quarantine, production, and testing |
responseData | Includes batch number, source document type, collection time | Matches batch binding and time verification requirements for traceability in this category |
detail | Full | Retain field details of original records for subsequent traceability verification |
| Single recall limit | Top 10 entries | Balances retrieval efficiency and coverage of multiple document types required for traceability |
| Retrieval scope filter | Execute based on the current production batch | Prevent irrelevant cross-batch documents from being included in citation scope |
Three Common Configuration Mistakes
- Phenomenon: The number of cited returns retrieved does not match the number of context window entries displayed on the interface. For example, the interface shows 30 context entries but actual retrieval returns 310. Cause: The retrieval scope is not correctly bound to the current production batch, leading to irrelevant cross-batch documents being included in the context.
- Phenomenon: After setting a citation limit, document chunks in the knowledge base that exceed the limit are still cited. Cause: Document chunks are not split by production batch, so single chunks exceed the limit without being split according to the corresponding batch dimension.
- Phenomenon: Cited content cannot be associated with specific quarantine or production records. Cause: Unique identifier fields such as batch numbers are not included in returned data, making it impossible to match original records during traceability.
How to Verify Correct Configuration
- Trigger a process document drafting request, and verify that returned citation sources include all relevant document types for the current production batch.
- View the retrieval log to confirm that the number of returned citations matches the set context window, with no irrelevant cross-batch content.
- Export the citation traceability report, and verify that each cited item includes necessary identification fields such as batch number and collection time.
- Simulate an incident report drafting scenario to verify that cited cold chain temperature data matches the continuously collected record for the corresponding time period.
The values provided are common starting points and should be measured against the reader's own samples.