What the data for this category looks like
Data sources for this category include the national public resource trading platform, local public resource trading platforms, project approval systems, and bidding supervision systems. The update frequency is daily. Commonly covered document types include bidding announcements, pre-qualification documents, bidding documents, tender documents, bid evaluation reports, successful candidate announcements, and successful bid result announcements. Documents contain dedicated fields such as section division, qualification requirements, bill of quantities, and construction period. The content has both structured and semi-structured characteristics, with no unified universal format.
What constraints these characteristics impose on the knowledge base retrieval and recall link
Daily updated data sources require the retrieval system to support incremental synchronization and scheduled full refresh, to avoid recalling outdated bidding information. Datasets from multiple platforms must be bound separately, to prevent mixing in irrelevant government procurement documents. The mixed structure of documents requires retrieval to balance keyword matching and semantic understanding, to accurately match core business fields such as qualification requirements and bid evaluation reports. At the same time, the consequences of business errors are severe, so retrieval completeness and ranking rationality must be ensured, to avoid missing key basis and triggering compliance risks.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Match the density of professional terminology in bidding documents, filter irrelevant general text, and ensure the relevance of recalled content |
limit | Top 8 entries | Cover the number of core clauses in a single bidding document, while controlling the total context length to comply with large model input limits |
searchMode | hybrid | Combine keyword retrieval and semantic retrieval, adapting to the mixed content structure of structured fields and professional text in bidding documents |
usingReRank | Enabled | Correct deviations from pure semantic ranking, and prioritize core business documents such as qualification requirements and bid evaluation reports |
datasetSearchUsingExtensionQuery | Enabled | Supplement associated terms such as sections, competent authorities, and construction periods, to improve recall completeness across documents |
collectionIds | Bind dedicated bidding datasets | Limit the retrieval scope, avoid mixing in irrelevant government procurement data, and accurately match business scenarios |
Three common misconfigurations
- Insufficient number of recalled results: The number of results returned on the interface is less than the configured
limitvalue. The cause is thatdatasetSearchUsingExtensionQueryis not enabled, and cross-platform dataset synchronization is not configured, resulting in only some locally stored bidding documents being retrieved. - Ranking results do not meet business expectations: Content related to bid evaluation reports appears after bidding announcements. The cause is that
usingReRankis not enabled, and ranking relies solely on semantic similarity, without weighted optimization for core business fields. - Retrieval interface returns timeout errors: The cause is that
collectionIdsis not limited to dedicated bidding datasets, and the retrieval scope covers too many irrelevant documents, causing loading time to exceed thresholds.
How to confirm the configuration is complete
- View the dataset synchronization logs in FastGPT v4.17.0, confirm that the daily incremental update task has been triggered normally, covering the latest successful bid result announcement data.
- Enter the core business question for drafting a bidding document, and verify that the recalled results include specified document types such as pre-qualification documents and bid evaluation reports.
- Adjust the
similaritythreshold, and verify that the relevance of returned results meets the basic requirements of business review. - After enabling
usingReRank, compare the results before and after ranking, confirm that the priority of core business fields has been effectively improved.
The values listed above are common starting points, and should be evaluated 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.