What the data for this category looks like
Data for this category comes from social insurance handling systems, personal benefit records, labor relations and work injury recognition systems, and the electronic social security card platform. The update frequency is daily. Documents processed daily include benefit application forms, payment period proofs, retirement approval forms, work injury recognition materials, and unemployment benefit application materials. Most documents are a mix of structured forms and unstructured proof materials. Data fields include insured person identity identifiers, payment cycles, benefit calculation parameters, review status, and other items. The proportion of unstructured content without a unified fixed format fluctuates based on business type.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
Daily updated data requires the retrieval pipeline to support incremental synchronization, to avoid resource consumption caused by full reconstruction. The mixed document structure of structured forms and unstructured proof materials requires adaptation to chunking rules and retrieval matching logic for different formats. The requirement for precise matching of the insured person identity identifier field means that retrieval must prioritize associating identity-related parameters, to filter interference from generic keywords. Associated retrieval of review status and benefit calculation parameters requires support for multi-field joint recall, to ensure the review link can quickly retrieve complete associated information for corresponding materials.
How to configure the settings
The following configurations are designed for the knowledge base retrieval module of FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.72–0.80 | Meets the precise matching requirements for core fields such as insured person identity and payment period, and filters low-relevance non-business documents |
limit | Top 8 entries | Covers multiple types of associated documents required for benefit verification, such as payment records, recognition materials, and application forms, and avoids context overflow caused by too many single recall results |
searchMode | vector + keyword | Balances keyword matching for structured fields and semantic vector recall, and adapts to mixed-format business documents |
usingReRank | Enabled | Performs secondary sorting on recall results, corrects deviations from single vector recall, and improves the relevance accuracy of review materials |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands relevant business keywords based on user queries, covering common synonyms and business terms for benefit verification |
collectionIds | Specify the exclusive collection for social insurance benefit verification | Isolates business data from other category data to avoid interference from irrelevant documents on retrieval results |
Three common misconfigurations
- Symptom: A prompt appears during retrieval stating that the file path does not exist, and imported documents fail to load normally. Cause: The local storage path of the knowledge base is not explicitly configured. The system default cache path does not match the business deployment path, and file verification is not enabled, so some abnormal documents cannot be read.
- Symptom: Images in locally imported work injury recognition materials return expired prompts and cannot be displayed normally. Cause: The local file persistent storage rule for the knowledge base is not configured. Temporary cached image links expire when the cache cycle ends.
- Symptom: Retrieved materials do not match the target insured person, and document results from other insured persons appear. Cause:
datasetSearchUsingExtensionQueryis not enabled, and identity identifier association is not used for extended retrieval, leading to generic keyword recall of business materials from other insured persons.
How to confirm the configuration is correct
- View the associated tags of the knowledge base collection, confirm that it only contains documents related to pension, work injury, and unemployment benefit verification, and verify that the
collectionIdsconfiguration is active. - Submit a test query containing an insured person’s ID number, check whether retrieval results prioritize returning that insured person’s payment records and recognition materials, and verify the configuration effects of
similarityandsearchMode. - Import an image of work injury recognition material, retrieve relevant business questions, confirm that the image can be displayed normally and the link has not expired, and verify the correctness of the local file persistent storage configuration.
- View the retrieval log, confirm that the number of recall results matches the
limitsetting, and that the sorting afterusingReRankconforms to business logic.
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.