What data does this category include?
Core data for this category comes from the national public resource trading platform, local trading platforms, project approval systems, and bidding supervision systems. The update frequency is daily. Commonly generated documents include bidding announcements, pre-qualification documents, bidding documents, tender documents, bid evaluation reports, public notices of successful bidders, and successful bid result announcements. Professional fields involved include qualifications, performance, bill of quantities, construction period, and others. Service subjects include development and reform departments, industry competent departments such as housing and urban-rural development, transportation, water resources, public resource trading centers, and bidding agencies.
What constraints do these characteristics impose on the tool calling and plugins link?
Daily updated data sources require tool calling to be configured with a scheduled synchronization mechanism to avoid returning expired data. Multiple types of document structures require plugins to adapt to parsing rules for different documents, accurately extracting professional fields such as qualifications and bill of quantities. The specificity of professional fields requires tool calling to limit scope, preventing result deviations caused by generalized calls. The docking needs of multiple service subjects require plugins to support permission verification and unified format processing for multiple data sources.
Reference only; verify against the target FastGPT release and deployment documentation.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
toolChoice | auto | Adapts to the demand for on-demand calling of multiple tools in bidding and tendering scenarios, automatically matching tool types for business scenarios |
functionCall | Enabled, limit call scope to qualification condition analysis, bid section splitting | Focuses on core business links of bidding and tendering, avoiding interference from irrelevant tool calls on processes |
| Single tool call upper limit | First 3 entries | Controls returned data volume, preventing context overload that affects generation efficiency |
| Tool call timeout period | Calibrated based on actual testing (30-90 seconds) | Adapts to timeliness requirements for bidding and tendering data synchronization, preventing process interruptions |
| Reference format cleaning | Enabled | Filters annotation formats returned by the knowledge base, avoiding redundant markup in output |
Three Common Misconfigurations
- Phenomenon: The string "Reference Mark: [1]" appears in the output content. Cause: The reference format cleaning configuration is not enabled, and the original returned content from the knowledge base is directly spliced after tool calling.
- Phenomenon: The custom interface is not called after the specified keyword is triggered. Cause: The callback parameters of
functionCallare not correctly bound, and the call path and permission verification of the custom interface are not configured. - Phenomenon: The number of results returned by tool calls exceeds expectations. Cause: A reasonable single tool call upper limit is not set, and the default configuration returns excessive redundant data.
How to Verify Successful Configuration
- Submit a test request containing "pre-qualification documents", check whether the tool automatically calls functions related to qualification condition analysis.
- Review the generated content, confirm that redundant format characters such as "Reference Mark: [1]" are not present.
- Simulate the scenario triggered by the configured specified keyword, check whether the bound custom interface is successfully called and corresponding business content is returned.
- View tool call logs, confirm that the number of single returned results meets the configured upper limit requirements.
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.