What Data for This Category Includes
Five data sources are used for this category: unified budget management system, project declaration materials, fund payment data, business system ledgers, and satisfaction surveys. Daily processed documents include performance target forms, self-assessment forms, performance evaluation reports, supporting material lists, and rectification reports. Data updates occur annually. Single document length varies widely, from hundreds of characters for performance target forms to tens of thousands of characters for evaluation reports. Most fields use fixed structured formats, including project code, fund amount, performance indicator completion values, and similar content.
Constraints on Tool Calling and Plugin Workflows
Annual data updates require the knowledge base refresh cycle for tool calling to match the annual rhythm. Frequent synchronization is unnecessary. Multiple data sources require tool calling to adapt to interface formats of different systems. Cross-data-source retrieval of budget, payment, satisfaction and other data must be supported. Wide variation in single document length requires tool calling context processing capability to adapt to long texts. Critical information is not truncated. Fixed structured fields require tool call return results to strictly match preset field formats. Data verification accuracy is ensured.
Configuration Settings
This configuration guidance is developed for FastGPT v4.17.0, and covers two core parameters: toolChoice and functionCall.
| Configuration Item | Recommended Value | Basis for This Setting |
|---|---|---|
toolChoice | Set to auto | Adapt to scenarios requiring cross-system data verification, trigger tool calling only when external business system data retrieval is needed |
| Tool call timeout period | 600 seconds | Performance evaluation reports and supporting material verification require processing multiple documents, a longer processing duration prevents mid-run interruptions |
| Number of returns per single tool call | Top 3 entries | Fiscal data requires precise matching, too many return results may introduce irrelevant information and reduce verification accuracy |
| Tool call retry count | No more than 2 times | Avoid business system interface rate limiting caused by repeated calls, meet stability requirements for fiscal data verification |
functionCall format verification | Enable mandatory verification | Ensure tool call return fields comply with fixed structured format requirements of performance target forms and self-assessment forms |
| Knowledge base refresh cycle | Set to annual | Match the annual update frequency of fiscal expenditure performance evaluation, avoid using expired project and fund data |
Three Common Misconfigurations
- Phenomenon: API calls return performance target matching results that differ significantly from online chat, with higher accuracy in online chat. Cause: Online chat enables context optimization and result rearrangement by default, while the API has no corresponding parameters configured, leading to insufficient relevance between retrieved supporting materials and performance targets.
- Phenomenon: Tool calling returns empty fields and fails to retrieve budget payment data. Cause: Tool call data source permissions are not configured correctly, and the interface parameter format of the unified budget management system is not adapted.
- Phenomenon: Tool call results cannot generate required pie charts. Cause: Markdown rendering support is not enabled in tool call result formatting configuration, and the adapted format for chart generation is not specified.
How to Verify Correct Configuration
- Initiate a test call, pass a query request for a single performance self-assessment form, verify that tool calling triggers only when external data retrieval is required.
- Check the API return result fields, confirm they match the fixed field formats of performance target forms and self-assessment forms.
- Compare identical query results from online chat and API calls, confirm their retrieval logic and result consistency meet expectations.
- Trigger a timeout test, verify that tool calling completes within the set duration or returns a reasonable timeout prompt.
The values listed here are common starting points and should be evaluated against applicable 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.