What Data for This Category Includes
Core data for precast concrete components comes from BIM detailed design, rebar and concrete batch ledgers, mold table and curing kiln operation records, laboratory test reports, and QR code traceability systems. Daily documents include component detailed drawings, rebar cutting lists, mold inspection records, pouring and vibration records, curing records, and demolding and factory inspection reports.
Data is updated per individual component, per batch of materials, and per curing kiln cycle. Document structures include fields such as component ID, rebar specification, embedded part location, and strength parameters. Most units use standard engineering units including millimeters, megapascals, and cubic meters.
Constraints on Knowledge Base Retrieval and Recall
Retrieval for precast concrete components must link dynamic data for individual components, batches, or kiln cycles. Retrieval logic must support precise matching of corresponding documents by component ID.
Supported document types include drawings, structured ledgers, and operation records. Retrieval must accommodate mixed-format text and structured data. Fields include engineering-specific units, so retrieval must retain unit-associated matching to avoid ambiguity.
Error consequences are significant, so recalled results must have high confidence, and irrelevant documents must be filtered. Unlike ready-mixed concrete categories, retrieval must match individual components corresponding to lifting positions. It cannot use the retrieval logic for bulk fluid delivery.
Configuration Settings
This document is written based on the configuration logic of FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Match precise fields for precast components such as embedded part location and strength parameters, avoid low-relevance general engineering documents |
limit | Top 8 entries | Cover associated documents for individual components including cutting lists and inspection records, avoid excessive redundant data interfering with review |
searchMode | fullTextRecall | Use full-text retrieval for text content. Field, metadata, batch, and document-type constraints require separate filters and are not enforced by this mode. |
usingReRank | Enabled | Reorder initially recalled results by component ID and batch number, prioritize matching the target review object |
datasetSearchUsingExtensionQuery | Enabled | Automatically supplement extended query terms such as batch and kiln cycle associated with component IDs, improve retrieval precision |
collectionIds | Configured per target factory’s component document collection | Limit retrieval scope to only the knowledge base related to precast concrete components for that factory, avoid interference from cross-category data |
Three Common Configuration Errors
- The interface continuously displays the "Retrieving" status with no returned results. The cause is failure to limit the
collectionIdsscope, retrieving the full cross-factory knowledge base, with excessive data volume causing a timeout. - Knowledge base question optimization only supports GPT-3.5 configuration. The cause is failure to enable custom model access in system settings, only the default GPT-3.5 model chain is active.
- The number of retrieved results does not match expectations, with a large number of irrelevant ready-mixed concrete documents. The cause is failure to enable structured field matching in
searchMode, only using full-text retrieval without filtering documents for non-precast concrete components.
How to Verify Correct Configuration
- Enter the target component’s ID, check if retrieved results prioritize returning associated documents such as the component’s detailed drawing and rebar cutting list.
- Check if the value of
similarityin the configuration items falls within the 0.75–0.85 interval. This can be verified by comparing recalled results across different threshold values. - Call the interface to query two precast concrete component knowledge bases under the same account, confirm that returned results include file IDs and document content for the corresponding collections.
- Check if retrieval latency meets business requirements. This can be optimized by adjusting the
limitparameter.
The values provided are common starting points and should be measured against the reader's own samples.