What the data for this category looks like
Production data for this category comes from water source and raw material test reports, formula systems, online sugar and acidity monitoring, filling equipment logs, visual inspection results, and LIMS systems. Daily documents include formula sheets, water treatment records, blending records, sterilization or aseptic filling records, packaging material inspection certificates, and finished product test reports. Data is updated per production batch and shift, with some monitoring data collected continuously. Document fields include raw material batch number, sugar and acidity values, filling liquid level parameters, sealing pass rate, label batch number, and shelf life period. Units include mg/L, %, mm, hours, and others.
What constraints these characteristics impose on knowledge base retrieval and recall
Production data for this category is updated in real time or per batch. The knowledge base must support high-frequency incremental synchronization to avoid retrieving expired water treatment parameters or formula information. Document fields contain professional test values and production parameters. Retrieval must match exact fields to avoid over-matching with generic keywords, which can lead to matching errors for feeding parameter deviations. Bid document and quotation preparation requires linking to test data for specific batches. Recall results must be limited to documents from the corresponding production cycle to avoid mixing cross-batch data. Additionally, individual documents are lengthy. Reasonable chunking and recall range control is needed to ensure retrieval results match bid requirements.
Configuration settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | This category’s documents contain professional test values and batch identifiers. A threshold that is too low will introduce irrelevant data, while a threshold that is too high may miss valid matching results |
limit | Top 8–12 results | Bid documents need to cover multi-dimensional production and test data. Too many results will exceed the context window, while too few will not support a complete quotation basis |
searchMode | fullTextRecall | The query examples emphasize identifiers or specialist terms; test this mode with representative data, not as a product default. |
usingReRank | Enabled | Long document chunks may have scattered semantics. Reranking improves the retrieval priority of key feeding parameters and water treatment data |
datasetSearchUsingExtensionQuery | Enabled | Bid requirements often include complex conditions. Extended queries can supplement implicit needs such as associated water treatment and filling parameters |
collectionIds | Bind the collection ID of the corresponding production dataset | Retrieval scope must be limited to the factory’s exclusive production documents to avoid cross-factory data interfering with bid preparation |
Three common mistakes
- Phenomenon: When uploading water treatment records or formula sheets, the interface prompts that the file size exceeds the limit, but the actual file is only about 1 MB. Cause: The knowledge base’s file upload verification rules are incorrectly configured, or the collection’s storage quota is exhausted, leading to unintended size limit triggers.
- Phenomenon: After upgrading from FastGPT v4.17.0 to 4.8.18, previously matchable finished product test documents cannot be retrieved. Cause: Default parameters for some retrieval configurations were adjusted after the version update, and the precise matching requirements for this category were not synchronized and adapted.
- Phenomenon: When attempting to create or modify a single piece of data in a collection via the API, local QA synchronization cannot be achieved. Cause: The mapping relationship between the collection ID and the local data source is not correctly associated, or the incremental synchronization configuration for the dataset is not enabled.
How to confirm the configuration is correct
- Upload a test formula sheet and water treatment record, check that the upload succeeds with no abnormal prompts.
- Enter a search term that includes specific sugar level parameters or label batch numbers, verify that the recall results include documents from the corresponding batch, and that the number of results matches the configured range.
- Compare retrieval results before and after the upgrade, confirm that the retrieval mode and reranking function configuration are adapted to the current version’s retrieval logic.
- Call the API to create or modify a single piece of data in the collection, check that synchronization between the local data source and the knowledge base takes effect.
The values provided are common starting points and should be measured against the reader's own samples.