What data for this category looks like
Data for this category comes from fault timelines, system logs, monitoring metrics, and code commit records. Daily output documents include RCA reports, incident tickets, review minutes, and rectification tracking sheets. Data updates occur every major incident or once per month. Document fields include unique incident identifier, root cause classification, associated test case number, rectification measures, closed-loop status, and more. Units are mostly timestamps, character counts, and event counts. There are no unified fixed numerical units; only event, character, and timestamp are used as measurement identifiers.
What constraints do these characteristics impose on tool calling and plugins
Low data update frequency and large single data volume require tool calling to support batch query and incremental synchronization. This avoids repeated pulling of full historical data. Most document fields are structured identifier content. Tool plugins must adapt to field parsing and association matching to enable quick association of corresponding test cases during root cause analysis. Incident occurrences have strong temporal correlation. Tool calling must support data filtering by time range, and retain complete log traceability links to facilitate root cause hypothesis verification. The closed-loop status of rectification tracking sheets requires real-time synchronization. Tool calling must support callback triggering after status updates to ensure the timeliness of root cause analysis.
How to configure settings
This document applies to FastGPT v4.17.0. The following are tool calling configuration recommendations for this category:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
toolChoice | auto | Adapts to multi-tool linkage scenarios during root cause analysis, automatically selecting matching tools to process test case-associated data |
| Tool calling timeout | 600 seconds | Matches the time span of single incident data pulling, preventing timeout interruptions due to large data volume |
| Plugin batch query limit | First 20 items | Adapts to the monthly updated incident data volume, avoiding excessive redundant information returned by a single call |
functionCall | Triggered by field matching | Accurately calls corresponding tools to parse root cause descriptions and associated test cases in RCA reports |
| Result post-processing switch | Enabled | Filters irrelevant log fragments returned by tools, focusing on content related to root causes |
| Associated test case matching threshold | 0.75–0.85 | Calibrated through actual testing, ensuring association accuracy between test cases and incident root causes |
Three common misconfigurations
- Tool calling returns empty results or malformed format. The cause is that the
toolChoiceparameter is not configured, preventing the AI from automatically matching the corresponding tool to process log data associated with test cases. - Database plugin reports an error when executing multi-line SQL statements. The cause is that batch execution support for the plugin is not enabled. The default setting only supports single-statement execution, and does not adapt to multi-step rectification SQL associated with RCA reports.
- Tool calling results do not include expected referenced content. The cause is that the trigger rule for
functionCallis not configured, causing the online tool to fail to correctly capture associated log fragments in the fault timeline.
How to confirm correct configuration
- Manually trigger tool calling, check whether associated logs and test case data for the specified incident can be automatically pulled, and confirm that the
toolChoiceparameter is effective. - Test the database plugin by executing multi-line SQL statements, confirm there are no errors, then verify that the returned results match the expected rectification tracking sheet data.
- Enable the result post-processing function, check whether irrelevant fragments in the content returned by the tool are filtered, and adjust the filtering rules according to actual business needs.
- Simulate an incident update scenario, confirm that the plugin can synchronously update the closed-loop status of the rectification tracking sheet and trigger subsequent root cause analysis processes.
The values provided here 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 2026-10-19.