What this category of data looks like
Typical data for this category originates from ERP/WMS systems, barcode or RFID collection devices, purchase work orders, storage location management ledgers, AGV transportation systems and weighing equipment. Daily documents include receiving slips, putaway slips, material picking and returning slips, warehouse transfer slips, monthly inventory reports, lot tracking ledgers and expiration date control records. Data fields cover material codes, lot numbers, storage location numbers, expiration dates, inventory quantities, inspection statuses and more. Units include pieces, kilograms, cubic meters and others. Update frequencies fall into three categories: real-time (for receiving and putaway operations), daily (for material picking/returning and warehouse transfer operations) and monthly (for full warehouse inventory checks).
What Constraints These Characteristics Impose on the Knowledge Base Retrieval and Recall Link
Multi-source heterogeneous data sources require retrieval systems to support precise matching of structured fields, to avoid returning irrelevant cross-category documents. The high-frequency update feature requires configuring incremental synchronization mechanisms to ensure retrieval results always match the latest storage location, expiration date and lot information. A single business document such as a lot tracking ledger contains multiple core fields, so the retrieval scope must be limited to only warehousing-related collections to prevent invalid recall. The most time-consuming verification step relies on precise matching of materials and lots, so the retrieval threshold must be set to a relatively high level to reduce the probability of false recalls.
Configuration Settings
The following configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Value/Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches precise business fields such as material codes and lot numbers, requiring a high similarity threshold to avoid mixing in irrelevant results |
limit | Top 8 entries | Warehousing-related single-lot document data volume is moderate; 8 entries can cover core information such as storage locations and expiration dates required for complete delivery |
searchMode | fullTextRecall | The query examples emphasize identifiers or specialist terms; test this mode with representative data, not as a product default. |
usingReRank | Enabled | Re-ranks recall results by business relevance, prioritizing valid records that match expiration dates and storage locations |
datasetSearchUsingExtensionQuery | Enabled | Automatically supplements extended queries related to material codes and lot numbers, improving recall coverage across documents |
collectionIds | Exclusive collection IDs corresponding to warehousing and lot documents | Limits the retrieval scope to avoid mixing in knowledge base data from other categories, reducing invalid retrieval time |
Three Common Mistakes
- Phenomenon: Knowledge base retrieval response times out, with single queries taking more than 10 seconds to complete. Cause: The
collectionIdsconfiguration to limit retrieval scope was not set, leading to full traversal of all knowledge base collections. The large volume of warehousing-related documents causes excessively long retrieval times. - Phenomenon: Retrieval results show document download links in debug mode, but these links disappear after the application is published. Cause: The external link generation configuration for knowledge base attachments was not enabled in the production environment, with temporary access permissions only enabled in debug mode.
- Phenomenon: Field matching fails when importing lot ledgers via Excel, and retrieval cannot accurately locate target materials. Cause: No mapping rules were established for core business fields such as material codes and lot numbers during import. Irrelevant fields were confused with core business fields, causing the retrieval matching logic to fail.
How to Verify Successful Configuration
- Run a single material code query, check that the similarity scores of returned results fall within the 0.75–0.85 range, and that storage location and expiration date records for the same lot are returned first.
- After importing a batch of warehousing-related documents, trigger a batch retrieval test to confirm that single query response times meet expectations with no obvious timeouts.
- Switch to the published environment, verify that document download links from debug mode are included in retrieval results, confirming that permission configurations are correct.
- Import an Excel lot ledger file, check the retrieval matching results after field mapping, confirming that core fields such as material codes and lot numbers are matched correctly.
The values provided are common starting points and should be measured against the reader's own samples.