What this category of data looks like
This category of data mainly comes from local public data open platforms, integrated big data platforms, departmental business databases, and electronic certificate libraries. The update frequency is monthly. Daily documents include public data resource catalogs, data opening lists, data item descriptions, desensitization rules, interface documents, and update frequency descriptions. Data fields cover data item names, opening scope, desensitization requirements, interface addresses, update cycles, authorization boundaries, and more. All are structured or semi-structured openable data assets.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
The monthly update feature requires the knowledge base to complete full data synchronization at fixed intervals, to avoid recalling expired open data items or updated interface information. The mainly structured document structure requires retrieval to accurately match data fields, reducing interference from irrelevant results. The existence of desensitization rules and authorization boundaries requires the recall link to automatically filter data content that is not open or beyond the authorization scope. The scenario where multiple types of documents exist in parallel requires clearly limiting the retrieval scope to the document set related to public data catalogs, avoiding mixing in other government public content, and ensuring the accuracy and compliance of ledger summary.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Public data catalog fields require precise matching. This interval adapts to the retrieval accuracy requirements for structured fields, avoiding irrelevant data item recall |
limit | Top 10 entries | The monthly updated document batches for this category are usually manageable in scale. 10 entries can cover the core data required for a single ledger summary, avoiding context overflow |
searchMode | Precise matching | Most data in this category uses structured fields. Precise matching mode reduces recall of non-target data, meeting the rigor requirements for ledger summary |
usingReRank | Enabled | The monthly updated document set contains a large number of similar fields. Reranking improves the sorting priority of highly relevant data, ensuring content accuracy for report generation |
datasetSearchUsingExtensionQuery | Disabled | The field definitions for public data catalogs are clear, so no additional query term expansion is needed, avoiding introducing redundant content unrelated to data items |
collectionIds | Bind a dedicated collection for public data catalogs | Ensure the retrieval scope only covers documents from this category, avoiding mixing in other government documents, and meeting data governance boundary requirements |
Three common mistakes
- Phenomenon: After adding the public data catalog knowledge base, calling the retrieval interface returns an error message, or the locally deployed large model cannot respond normally after adding the knowledge base. Cause: The dedicated collection is not correctly bound, the retrieval scope covers non-target documents, or the
similaritythreshold is set unreasonably, causing recalled content to exceed compliance scope. - Phenomenon: The number of retrieval results does not match the configured
limitparameter, with either too many or too few results. Cause:collectionIdsis not set correctly, the retrieval scope includes documents from other collections, or the reranking step filters some qualified results. - Phenomenon: The generated ledger report contains undesensitized data item content or opening descriptions beyond the authorization scope. Cause: The desensitization rules and authorization boundary documents are not associated during the retrieval and recall link, or
searchModeis set to fuzzy matching, recalling data items without compliance annotations.
How to confirm the configuration is complete
- Check the knowledge base collection binding status, confirm that it only contains documents related to public data catalogs, verify that the
collectionIdsparameter matches the actual collection ID, and refer to the FastGPT v4.17.0 collection configuration specifications. - Initiate a test retrieval, enter a data item name keyword, check if the returned result fields match the category document structure, and confirm that the
similaritythreshold meets business accuracy requirements. - Trigger a ledger summary task, check if the generated report content only includes open data items and corresponding interface descriptions, with no unauthorized or undesensitized content appearing.
- Wait until the monthly synchronization cycle ends, verify that the retrieval results include the latest updated data items, and confirm that the synchronization process is running normally.
The values given 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.