What the data for this category looks like
This category’s data comes from national public resource trading platforms, local public resource trading platforms, engineering approval systems, and bidding and tendering supervision systems. The update frequency is daily. Daily generated documents include bidding announcements, pre-qualification documents, bidding documents, bid submissions, bid evaluation reports, successful candidate announcements, and bid result announcements. Document fields cover core business information such as section division, qualification requirements, bid evaluation rules, bill of quantities, and bid results. Document formats vary slightly across different sources.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
Daily updated data sources require the retrieval link to support incremental synchronization configuration. This avoids system resource consumption from full data pulls. Differences in attributes across multiple types of business documents require matching different retrieval priorities. For example, bid evaluation reports have higher authority than standard bidding announcements. They must be displayed first during the recall stage. Structured fields such as sections, qualification requirements, and bill of quantities require retrieval to precisely match business keywords. This prevents generalized recall of irrelevant non-section documents. The need to integrate multi-source data requires the retrieval link to support binding datasets by data source group. This prevents cross-platform data from interfering with retrieval result accuracy.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Match precise qualification requirements and section information for engineering bidding, filter low-relevance recall results |
limit | Top 8 entries | Cover core information across multiple document types including bidding documents, bid evaluation reports, and bid result announcements, avoid exceeding the 12000 token model context window |
searchMode | Precise matching | Target structured fields such as sections and qualifications, prioritize returning documents that align closely with business needs |
usingReRank | Enabled | Reorder initial recall results by relevance, improve ranking priority for high-value documents such as bid evaluation reports and bid result announcements |
datasetSearchUsingExtensionQuery | Enabled | Automatically expand business-related keywords such as qualification requirements and bill of quantities, expand retrieval coverage |
collectionIds | Configured by data source group | Differentiate datasets from national public resource trading platforms, local trading platforms and other sources, prevent cross-source data interference |
Three Common Configuration Errors
- Retrieval results only display 1 knowledge base reference. The interface shows the prompt "Knowledge Base Reference (1 entry)". The cause is that the
limitparameter is set to 1. This does not align with the business requirement for multi-document recall in engineering bidding. - The default dataset selection is empty after importing the knowledge base. Manual dataset selection must be triggered to proceed. The cause is that the
collectionIdsparameter is not configured. No exclusive dataset for the corresponding data source is bound. - An abnormal status marker appears after enabling result reordering in the knowledge base. The cause is a configuration conflict. The
usingReRanksetting does not match the reordering switch of the model plugin.
How to Confirm the Configuration is Correctly Set
- Access the knowledge base retrieval test page in FastGPT v4.17.0.
- Input a section number or qualification requirement keyword. Verify that the number of returned results matches the configured
limitvalue. - View the dataset binding list. Confirm that
collectionIdsis associated with the data source of the corresponding public resource trading platform. - Enable
usingReRank. Verify that the reordering switch status is normal, with no abnormal markers. - Initiate a retrieval request containing bill of quantities keywords. Check whether relevant business terms are automatically expanded. This matches the configuration effect of
datasetSearchUsingExtensionQuery.
The values provided are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-28.