What this category of data looks like
Typically, data sources for this category include ERP, WMS systems, barcode or RFID equipment, procurement and work order information, storage location records, AGV or conveyor system data, and weighing data. Daily documents include receiving slips, putaway slips, material picking and return slips, inventory transfer slips, inventory checklists, lot ledgers, first-in first-out and expiration date records. Core fields include material codes, lot numbers, storage location numbers, expiration dates, inspection status, and associated work order numbers. Data update frequencies cover real-time scenarios such as real-time AGV positions, daily scenarios such as material picking and return records, and monthly inventory counts such as inventory reconciliation. This page is based on configuration instructions for FastGPT v4.17.0.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
Data for this category is multi-source, heterogeneous, and has highly varied update cycles. Retrieval systems must support field mapping and real-time synchronization across multiple data sources to prevent data delays during lot traceability. Document types are diverse, with core fields focused on precise identifiers such as materials, lots, and storage locations. Retrieval must balance exact keyword matching and semantic understanding to avoid retrieving irrelevant documents. Full-link traceability requirements for associated materials mean retrieved results must cover all process documents from receiving to outbound shipping. Results must be sorted in line with business logic. Errors in this process directly impact production, so retrieval results must have high accuracy and reliability.
How to set the configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Core retrieval fields for this category include precise identifiers such as lot numbers and material codes. This value range filters low-similarity irrelevant warehousing documents |
limit | Top 8–12 entries | Warehousing documents have short individual content. An appropriate number of results can cover all document types required for full-link traceability, including receiving, material picking and return, and inventory check documents |
searchMode | mixedRecall | Balances fields that can be matched via exact keywords such as lot numbers, and content that requires semantic understanding such as expiration dates and logistics statuses |
usingReRank | Enabled | Results from multi-source retrieval must be reordered based on lot association priority to ensure the logical order of the traceability chain is correct |
datasetSearchUsingExtensionQuery | Enabled | Can expand natural language traceability questions such as "receiving records for a specific lot of material" into search terms including lot numbers and storage locations, improving retrieval accuracy |
collectionIds | Specify dedicated dataset IDs for warehousing and lot management | Restrict the retrieval scope to exclude interference from documents of other categories, improving retrieval efficiency |
Three common misconfigurations
- Symptom: A parameter error prompt appears after configuring the warehousing knowledge base and performing retrieval. Cause: Retrieval parameters were not configured according to the field characteristics of this category, and non-specified parameter names or value ranges were incorrectly used.
- Symptom: The number of returned retrieval results is too high or too low, affecting traceability efficiency. Cause: A reasonable
limitparameter was not set, and the value did not match the individual length of warehousing documents and retrieval requirements. - Symptom: Image-associated information in documents cannot be retrieved, only plain text content is obtained. Cause: OCR parsing configuration for document images was not enabled, causing key information such as lot numbers and storage locations in images to not be indexed.
How to confirm the configuration is properly set
- Enter a single lot number for retrieval, and verify that the returned results include all process-related documents for that lot, including receiving, material picking and return, and inventory check documents.
- Adjust the value range of the
similarityparameter, and verify whether the accuracy of retrieved results meets business requirements for lot matching. - Upload a warehousing document containing images such as a receiving slip with a barcode, and verify that text information in the image can be retrieved.
- Check whether the number of retrieved results matches the preset business requirements, and confirm that the
limitparameter is functioning correctly.
The values provided are common starting points and should be evaluated against the samples used in the specific deployment.