Reference only; verify against the target FastGPT release and deployment documentation.
What Data for This Category Looks Like
Data for this category comes from social insurance management systems, tax social insurance collection systems, market entity registration information, and shared data from public security and civil affairs departments. The update frequency is daily.
Daily processed documents include enrollment registration forms, payment base declaration forms, personnel change forms, social insurance payment verification slips, and personal benefit records.
Most document fields include personnel identity identifiers, insured unit information, payment base, verified amount, and personnel status.
Each document focuses on a single batch of enrollment or payment verification matters, with no cross-batch redundant associated fields.
Constraints on Knowledge Base Retrieval and Recall
Daily data updates require knowledge base retrieval to support incremental synchronization or full refresh, to avoid retrieving outdated enrollment or payment information.
Heterogeneous data from multiple sources requires the retrieval link to support cross-system field matching. For example, associate payment records from the tax system with enrollment registration information from the social insurance management system.
The feature that each document focuses on a single batch of matters requires retrieved results to be aggregated by batch or unit, to avoid scattered fields interfering with ledger aggregation.
Data association rigor requires retrieved results to include original data source identifiers. These identifiers support subsequent reconciliation and traceability, reducing the risk of underpayment, overpayment, or incorrect personal benefit records.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirement for core fields such as personnel identity and payment base, to avoid retrieving irrelevant historical records |
limit | Top 10 | The number of documents for a single batch of enrollment or payment verification matters is limited. Excessive retrieval results increase the processing burden of ledger aggregation |
searchMode | Vector Retrieval + Keyword Retrieval | Balances structured field matching for enrollment registration documents and unstructured business description content, to improve retrieval comprehensiveness |
usingReRank | Enabled | Reorders retrieved results by business relevance, prioritizing documents directly related to current personnel changes or payment verification |
datasetSearchUsingExtensionQuery | Enabled | Expands query terms to cover synonymous expressions for enrollment registration and payment verification, such as matching "personnel changes" with "personnel movements" |
collectionIds | Specify the dedicated collection ID for human resources and social security ledgers | Isolates knowledge base data from other categories, to avoid retrieving benefit verification documents that interfere with the current task |
Three Common Configuration Errors
- Phenomenon: Generated ledger reports do not include content source labels, making it impossible to confirm whether the content comes from the configured dedicated human resources and social security knowledge base. Cause: Knowledge base traceability configuration is not enabled, or the prompt does not require returning source fields.
- Phenomenon: Retrieval returns a large number of non-enrollment registration documents, with the number of results far exceeding the set
limitvalue. Cause:collectionIdsis not configured correctly, causing the retrieval scope to cover knowledge base collections from other categories. - Phenomenon: Retrieval requests trigger parameter format errors, making it impossible to retrieve documents normally. Cause: The value of
similarityis outside the reasonable range, or a non-specified configuration item name is used incorrectly.
How to Verify Configuration Correctness
- Submit a test query, such as "2024 Q1 [a unit] payment base declaration form", check whether the source of the returned results is the configured dedicated knowledge base collection.
- View the similarity score of the retrieval results, confirm that the score falls within the set
similarityrange, and adjust the threshold if it does not match. - Generate a ledger aggregation report, check whether the report includes the source identifier of each retrieval result, and confirm that the traceability function works properly.
- Simulate uploading a new personnel change form document, verify whether the retrieval results include the newly uploaded content, and confirm that no old data is missed.
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.