What data looks like for this category
Financial lease financial statement analysis data mainly comes from internal lease project ledgers, audited annual/quarterly financial statements, and special regulatory submission reports. Data update cadence follows monthly project ledger updates, quarterly interim reports, and annual audited financial statements. Single document structure includes project details, lease asset module under balance sheet, and lease income line items in cash flow statement. Core fields include net finance lease receivables, lease principal balance, weighted average lease term, and annualized lease interest rate. Units are mostly ten thousand yuan or hundred million yuan. Some segmented items use yuan as the pricing unit.
What constraints do these characteristics impose on knowledge base retrieval and recall
Financial lease financial statements have many specialized fields and strong relational links. Retrieval and recall must accurately match terms and numerical associations. A single document contains multiple sets of structured data. A single query may need to recall corresponding fields for multiple projects. Too few recalled entries will miss critical information. Data update frequencies vary: monthly ledgers update frequently, annual financial statements have large volume. The system must adapt to document parsing and retrieval efficiency for different scales. Some fields have unit differences. The system must automatically match pricing units during recall to avoid mismatches between values and units. Financial lease financial statements often involve sensitive corporate operating data. Recall results must only match authorized document fragments.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
recall count | Top 8-12 entries | A single financial lease financial statement document contains multiple sets of structured fields. This range covers multiple project details and report items to avoid missing core information due to too few recalled entries |
similarity threshold | 0.72-0.80 | The proportion of specialized terms is high. Precise matching of proper nouns and numerical associations in financial statements is required to avoid recalling irrelevant general financial documents |
chunk length | 800-1200 characters | A single record in a single lease project ledger has moderate length. Too long a chunk will split key field combinations, too short will lose contextual associations |
UPLOAD_FILE_MAX_SIZE | 1000 MB | A single annual financial statement collection may include dozens of sub-project documents. Sufficient storage and parsing space must be reserved for batch uploads |
PARSE_FILE_TIMEOUT_SECONDS | 300 seconds | Parsing large financial statements requires loading multiple table data sets. This setting prevents parsing processes from being interrupted by timeouts |
reranked return count | Top 5 entries | The content finally presented to analysts must focus on core lease asset data to avoid interference from excessive redundant information |
The parameter values provided on this page are common recommended starting points for configuration. Actual values are affected by material format, data volume and business rules. Specific issues require individual analysis. It is recommended to test against your own samples before finalizing.
Three Common Errors
- Phenomenon: The <Reference> field in the System module is empty when calling knowledge base chat. Cause: In FastGPT 4.8.14 and above versions, the knowledge base context splicing configuration is not enabled, or the similarity threshold is set too high, so no matching document fragments are recalled.
- Phenomenon: The file name list in the knowledge base cannot be exported. Cause: The file metadata collection function is not enabled, or the system does not open the interface permission for file list export.
- Phenomenon: Knowledge base capacity statistics do not match actual document usage. Cause: Capacity calculation is based on the compressed package size of uploaded files. The number of parsed plain text characters is not used as the calculation basis, and the occupancy of temporary parsing cache is not included.
How to Confirm Proper Configuration
- Upload a single-project lease financial statement sample, check if the parsed text retains core fields such as net finance lease receivables and lease principal balance.
- Initiate a query containing "2024 Q3 net finance lease receivables", verify that the returned <Reference> fragments include specific values and pricing units of the corresponding report items.
- Attempt to export the current knowledge base file list, confirm that all uploaded financial statement document file names can be obtained.
- Check system logs to confirm that no timeout errors related to
PARSE_FILE_TIMEOUT_SECONDSare triggered when parsing a financial statement larger than 100 MB.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.