What the data for this category looks like
Daily documents for this category include service guides, application forms, commitment letters, acceptance notices, correction notifications, administrative licensing decision letters, license certificates, and more. Data sources are the national integrated government service platform item database, administrative licensing item list, departmental business approval systems, and electronic certificate database. Update frequency is daily. Documents are bounded by legal licensing items, and include fields such as item setting basis, handling conditions, application material list, approval process, certificate format, and more. There is no unified fixed format, and it adjusts based on item type.
What constraints these characteristics impose on the "knowledge base retrieval and recall" link
The data sources cover cross-departmental systems and real-time updated item and certificate databases. This requires retrieval and recall to cover all relevant data sources and match the daily update rhythm to avoid returning expired information. Documents have no unified fixed format, so adaptation is needed for chunking and semantic recall of multiple types of government documents. This ensures that application materials, decision letters and other content with different structures can be correctly matched. The business rule bounded by legal licensing items requires that retrieval results are strictly bound to the setting basis and handling conditions of the corresponding item. No irrelevant content across items can be returned, to ensure the legality of review.
How to set the configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Meets the rigor requirements for government documents, filters low-relevance non-licensing item content |
limit | Top 10 entries | Controls the number of single recall results, adapting to the item-by-item verification scenario of file review |
searchMode | strict | Strictly limits the retrieval scope to the bound knowledge base set, avoiding cross-item recall |
usingReRank | Enabled | Re-ranks initial recall results, prioritizing content directly related to approval conditions |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands associated item keywords based on user queries, covering synonymous expressions related to approval |
collectionIds | Bind exclusive collections for the current administrative licensing item | Limits retrieval to only the exclusive document library for this item, excluding irrelevant content |
Three common errors
- Content outside the bound knowledge base is returned after triggering retrieval. The cause is incorrect configuration of the
collectionIdsparameter, which causes the retrieval scope to cover other unbound knowledge base sets. - Retrieval results cannot be recognized as valid JSON format. The cause is that format verification for
usingReRankordatasetSearchUsingExtensionQueryis not enabled, leading to abnormal structure of re-ranked or expanded results. - Empty JSON content is returned by retrieval. The cause is that
searchModeis set to strict and no documents matching the corresponding item are found, or the daily update synchronization of data sources is not completed, resulting in no available data.
How to confirm the configuration is properly set
- View the
collectionIdsparameter in the retrieval configuration, confirm that only the exclusive document collection for the current administrative licensing item is bound. - Trigger a test retrieval, verify that returned results only include relevant documents such as the service guide and application materials for this item, with no cross-item content.
- Check the format of retrieval results, confirm that returned content meets preset structured requirements.
- View data source synchronization logs, confirm that the daily update task has executed normally, with no expired data that failed to be synchronized.
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.