What this category of data looks like
This category of data mainly comes from flow, pressure, temperature type analytical instrument manufacturers and metrology laboratories. Daily documents include assembly and commissioning sheets, calibration procedures, original calibration records, error curves, metrology certificates, and software version records. Data sources cover design documents, standard instrument and environmental monitoring data, calibration software, laboratory information systems, and customer range requirements. The update frequency is after each calibration of every device. Data fields include calibration date, error value, standard instrument number, measured range and corresponding units (such as Pa, ℃, MPa). A single calibration record document usually contains multiple sets of measurement point data.
What constraints these characteristics impose on the knowledge base retrieval and recall link
Since the data update frequency is after each calibration of every device, the knowledge base must support high-frequency incremental synchronization to avoid recalling expired calibration records. Documents contain multiple sets of measurement data with units. Retrieval must match both fields and unit information; keyword matching alone is insufficient to cover business needs. Data sources cover multiple types of systems, so retrieval must be compatible with calibration files and certificates in different formats. Calibration records are core deliverables. The traceability and accuracy of retrieval results must meet metrological audit requirements, so there are high constraints on recall precision.
Configuration Settings
The following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Documents in this category require matching precise fields such as calibration date and range unit. This interval filters irrelevant general instrument documents and retains highly relevant calibration records |
limit | Top 10 entries | Calibration records for a single device usually do not exceed 5 sets. Recalling 10 entries covers associated data from multiple devices and avoids result redundancy |
searchMode | embedding | Data in this category relies on semantic and field matching. Vector search accurately recalls documents containing error values and calibration procedures |
usingReRank | Enabled | Documents with multiple sets of measurement point data use reranking to improve relevance sorting, ensuring the most relevant calibration records appear first |
datasetSearchUsingExtensionQuery | Enabled | Supplementary customer range requirements and standard instrument information as extended queries improve retrieval coverage for specific scenarios |
collectionIds | Specified according to the dedicated collection for calibration documents | Data in this category must be isolated to a dedicated knowledge base collection to avoid mixing documents from other categories and ensure precise retrieval scope |
Three Common Misconfigurations
- Phenomenon: Retrieval works normally in debug preview, but no retrieval results or a large number of irrelevant documents are returned after deploying to a custom page. Cause: The
collectionIdsparameter is not configured correctly. The retrieval scope is not limited to the dedicated collection of calibration documents for this category, resulting in matching irrelevant knowledge base content. - Phenomenon: Returned documents include expired calibration records and do not show the latest calibration data. Cause: The incremental update mechanism of the knowledge base is not enabled. Newly generated calibration records and certificates are not synchronized at the frequency of each calibration of every device.
- Phenomenon: When the knowledge base has no matching documents, the workflow still triggers the large model to generate content. Cause: The termination logic for empty retrieval results is not configured. The large model call is not skipped when the number of retrieval results is 0 as required by the business.
How to Verify Proper Configuration
- Enter a query containing a specific calibration date and range unit, and check whether retrieval results include calibration records and certificates with corresponding fields.
- View the knowledge base update log to confirm that new documents generated after each calibration of every device have been synchronized to the dedicated knowledge base collection.
- Test the query scenario where no matching documents exist, and confirm that the workflow does not trigger the large model to generate content.
- Check the sorting of retrieval results, confirm that the most relevant calibration records appear first, and verify that the reranking function is active.
The values provided are common starting points and should be evaluated against relevant in-use samples.