What the data for this category looks like
Data for this category comes from internal control systems, ERP systems, financial sharing systems, audit management systems, and procurement contract project systems of central enterprise internal control departments, audit departments, financial departments and their subordinate enterprises, as well as external audit institutions. Daily documents include internal control self-evaluation reports, audit notices, audit working papers, issue lists, rectification ledgers, and rectification closing certificates. Fields cover issue descriptions, rectification measures, supporting material associations, closing status, and more. Data update frequency is annual. Most document structures are business report formats combining structured tables and paragraphs.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
Data sources are scattered across multiple business systems, requiring cross-source integration before importing into the knowledge base. The update frequency is low, so high-frequency incremental synchronization is not needed. Documents contain structured tables and unstructured paragraphs, so retrieval needs to balance precise field matching and full-text semantic recall. Documents such as rectification ledgers and closing certificates are associated with specific issues and responsible entities, so retrieval needs to support targeted recall by associated fields. Original voucher documents such as audit working papers need to retain original links and supporting information, and retrieval results need to be associated with complete business chain information. A single retrieval must cover all audit and rectification documents within the year to avoid missing cross-cycle issue rectification records.
How to configure settings
The following configurations are implemented using the knowledge base feature of FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Aligns with semantic matching accuracy for audit issues, avoiding irrelevant rectification document recalls |
limit | Top 10 entries | Covers the number of issue and rectification records required for a single audit verification, avoiding result overload |
searchMode | hybrid | Combines semantic retrieval and keyword retrieval, adapting to mixed retrieval needs for structured documents and unstructured paragraphs |
usingReRank | Enabled | Reranks recall results to prioritize content strongly associated with audit issues and rectification requirements |
datasetSearchUsingExtensionQuery | Enabled | Expands retrieval terms based on user questions to cover audit-related synonyms and business terminology |
collectionIds | Configured per audit/internal control exclusive dataset | Limits retrieval scope to avoid mixing non-category-specific business documents |
Three common configuration errors
- The number of retrieval results is far lower than expected, with only hundreds of entries returned. Cause:
collectionIdsis not configured to limit retrieval scope, accidentally including the full-platform dataset in retrieval, or complete annual audit and rectification documents have not been imported. - Retrieval results fail to match precise audit issue descriptions, with a large number of irrelevant financial reimbursement documents recalled. Cause:
usingReRankis not enabled or thesimilaritythreshold is set incorrectly, leading to insufficient semantic matching accuracy. - Embedded images in imported audit working papers fail to display properly. Cause: Images were not uploaded to the online image bed supported by the platform, and local file paths were used directly as links, preventing the knowledge base from loading image resources.
How to confirm the configuration is properly set up
- Submit a typical audit issue query, and check whether the retrieval results include corresponding issue list and rectification ledger documents.
- Check the ranking of retrieval results to confirm that highly relevant documents appear at the top.
- Import an audit working paper containing embedded images, and verify that the images load normally.
- Export the knowledge base file, and check that the file size matches the total number of imported documents.
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.