What data falls under this category
Core data for this category comes from the unified budget management system, project declaration materials, fund payment data, business system ledgers, and satisfaction survey results. Regularly produced documents include performance target forms, self-assessment forms, performance evaluation reports, supporting material lists, and rectification reports. Data updates occur on an annual cycle. Document structure is organized around four core modules: performance targets, fund usage, output effects, and rectification status. Fields include project code, budget amount, performance indicator completion rate, supporting material number, and others, and follow the unified standardized format used by fiscal systems.
Constraints on citation sources and traceability
The annual update cycle results in stable data volume but fixed update timelines. Citation and traceability must be permanently linked to project data from the corresponding year to prevent cross-year confusion. Data from multiple sources must be accurately matched using project codes and supporting material numbers. This ensures the correspondence between fund payment data and performance target completion status can be traced. Document fields are numerous and tightly linked. When citing content, it is necessary to locate specific content tied to the relevant performance link. This avoids logical deviations caused by cross-module citations. Fiscal business documents follow unified specifications. Citation formats must align with standardized fields to support subsequent manual verification and compliance audits.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Adapts to the length of a single supporting material section in performance evaluation reports, preventing truncation of core cited content |
responseData | Enabled | Returns the original fields and data paths of citation sources to meet compliance traceability requirements |
detail | Enabled | Displays the full context of cited content, making it easier to verify the correspondence between performance targets and fund usage |
| Single recall limit | Top 6 entries | Covers the primary supporting materials needed for a single performance evaluation report, avoiding interference from redundant information |
| Citation marking rule | By project code + supporting material number | Matches the standardized fields of business documents, supporting subsequent manual verification |
Three Common Configuration Errors
- When calling the citation traceability interface, the interface connection status is normal, but no content field is returned in the response body. This occurs if the
responseDataconfiguration item is not enabled, or ifquoteMaxTokenis set too low, leading to full truncation of core content. - The generated citation content has inconsistent formatting and messy typesetting. This occurs when no unified citation marking rule is configured, and original data fields are directly spliced together.
- Garbled characters appear in the generated citation content. This occurs if the system fails to adapt to the encoding format of business documents, or if non-standard characters in the cited raw data are not preprocessed.
How to Verify Correct Configuration
- Import a single standardized performance self-assessment form, and check if the generated performance evaluation document includes citation markings for the corresponding project.
- Review the response data returned by the interface, and confirm that it includes the field names and data storage paths of the original data.
- Adjust the value of
quoteMaxToken, and verify that the generated content fully retains the core information of the supporting materials without truncation. - Submit multi-source business data for testing, and check if citation markings follow the unified project code + supporting material number rule.
The values provided here are common starting points and should be evaluated 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.