Reference only; verify against the target FastGPT release and deployment documentation.
What This Category of Data Looks Like
Typical data for this category comes from the national public resource trading platform, natural resource trading systems, property right exchange systems, and state-owned asset supervision systems. The platform updates data daily. Daily documents include transfer announcements, trading instructions, bid application materials, margin vouchers, transaction confirmation documents, listing result announcements, and property right transfer specifications. Core fields cover transaction subject ownership, restriction conditions, margin amount, and transaction number. Supported units include square meters, ten thousand yuan, and parcels.
What Constraints Do These Characteristics Impose on the "Knowledge Base Retrieval and Recall" Link
Data for this category comes from multiple dispersed sources and updates frequently. The retrieval link must support incremental synchronization across multiple systems to avoid data lag. Document types vary widely, and core information sits in unstructured fields such as ownership and restriction conditions. Prioritize matching content with strong semantic relevance. Errors in transaction subjects cause serious consequences, so retrieval results need high confidence to avoid irrelevant or incorrect transaction rule data. Party building learning Q&A also requires linking compliance requirements. Expand queries to cover the combination points of policies and transaction scenarios.
How to Configure Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | This category requires precise matching of core transaction fields such as ownership and restriction conditions. A too-low threshold introduces irrelevant subject data, while a too-high threshold may miss key compliance notes. This setting is calibrated via actual testing. |
limit | Top 10 entries | Core transaction information for this category concentrates in a small number of highly relevant documents. Too many retrieved entries exceed the context window, while too few may miss key ownership or margin requirements. |
searchMode | vector | This category contains large volumes of unstructured transaction rules and ownership descriptions. Vector retrieval ensures semantic relevance. Combining keyword retrieval supplements precise matching. |
usingReRank | true | Document structures for this category vary widely. A reranking model filters low-relevance announcement documents and focuses on key information such as ownership and margin amounts. |
datasetSearchUsingExtensionQuery | true | Party building learning Q&A links compliance policies. Query expansion supplements party building-related compliance requirements to meet compliance review needs for transaction scenarios. |
collectionIds | Configure by business domain grouping | Data for this category comes from multiple independent systems. Grouped configuration accurately retrieves documents for the corresponding business and avoids cross-domain data interference. |
Three Common Misconfigurations
- Phenomenon: After setting
usingReRanktotrue, the reranking field in retrieval results returnsfalse. Cause: The corresponding reranking model is not enabled in platform settings, or the reranking model is not bound to the current knowledge base collection. - Phenomenon: The variable reference dropdown menu for the knowledge base search node is empty, and dynamically specified knowledge bases cannot be selected. Cause: A variable of type "Knowledge Base ID" has not been created in global variables, or the variable has not been authorized to the current workflow.
- Phenomenon: The number of retrieved documents is far lower than the configured
limitvalue. Cause:collectionIdsis not configured correctly, or the data source has not completed the daily incremental synchronization, resulting in insufficient available retrievable valid documents.
How to Confirm Proper Configuration
- Initiate a test query that includes core ownership keywords, and check whether the
similarityscores of retrieval results fall within the configured range. - Trigger the reranking operation, and confirm that the returned results are sorted according to the priority of core fields such as ownership and margin requirements.
- Switch to test the dynamic variable, and confirm that the reference variable dropdown menu of the knowledge base search node can properly load the configured global variables.
- Check the data source synchronization log, and confirm that the daily updated transaction announcements have been included in the retrievable scope.
The values provided here 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.