What this category of data looks like
Data sources for ready-mixed concrete production include engineering orders, inspection records for sand, gravel, cement and admixtures, aggregate moisture content measurement data, mixing station control system logs, and laboratory test reports. Updates follow each vehicle production task, each batch of raw material delivery, and daily laboratory batch testing. Daily documents include mix proportion notices, opening identifications, production and transport orders, raw material inspection reports, test block records, and strength reports. Mix proportion notices contain sand usage, cement usage, and mixing water usage. Opening identifications contain measured slump values and mixing duration. Test block records contain molding dates and curing conditions. There is no unified mandatory format for data fields across all sources.
What constraints do these characteristics impose on the knowledge base retrieval and recall link?
The high-frequency, multi-source update rhythm requires the retrieval link to prioritize recalling the latest data, and avoid using expired raw material inspection or mix proportion data. Mixed storage of multiple types of documents requires filtering by document type, to ensure recalled mix proportion notices and raw material inspection reports match the current production task. Differences in data fields across sources require retrieval to support cross-source association, such as binding aggregate moisture content data with mix proportion data for matching. The number of documents involved in a single production batch is limited, so the number and relevance of recall results must be controlled to avoid redundant information interfering with subsequent reviews. The accuracy of core indicator matching directly affects production quality.
How to Configure the Setup
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirements for strongly correlated business fields such as mix proportions and slump in ready-mixed concrete production, and avoids recalling irrelevant old batch data |
limit | Top 8 entries | Covers multiple types of documents involved in a single production batch, including raw material inspections, mix proportions, and opening identifications, while controlling context length |
searchMode | fullTextRecall | The query examples emphasize identifiers or specialist terms; test this mode with representative data, not as a product default. |
usingReRank | Enabled | Reorders recalled results by business relevance, and prioritizes returning same-day moisture content measurement data and the latest mix proportion data |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands retrieval terms based on user questions. For example, expands "slump correction" to "aggregate moisture content correction water" to match professional documents |
collectionIds | Knowledge base collections associated with production tasks | Binds exclusive knowledge base collections for the current project’s mixing station and laboratory, avoiding recalling irrelevant documents from other projects |
Three Common Configuration Errors
- When calling the knowledge base search, the prompt "No matching content found" appears. The cause is that
collectionIdsis not configured to bind the exclusive knowledge base collection for the corresponding mixing station, and the global default knowledge base is mistakenly called, resulting in failure to retrieve project-specific documents. - The retrieval results include expired raw material inspection data. The cause is that the incremental update synchronization mechanism is not enabled, and raw material inspection reports from old batches older than 24 hours are recalled.
- When assigning a knowledge base to a global variable, the correct
collectionIdsis not passed, and the knowledge base name is passed instead. This causes the system to fail to match the target knowledge base collection.
How to Confirm the Configuration Is Correct
- Initiate a test query containing "same-day aggregate moisture content", and check whether the recall results include the latest same-day moisture content measurement documents.
- View the retrieval log to confirm that the number of returned results matches the
limitsetting of top 8 entries. - Switch the
collectionIdsparameter for the test project, and verify that the retrieval results only return documents from the current project, and do not include irrelevant data from other projects. - After enabling
usingReRank, check whether documents related to the same day’s production appear at the top of the sorted results.
The values provided are common starting points and should be measured against the reader's own samples.