Data Overview for This Category
The core data for this category originates from database monitoring, audit logs, SQL execution plans, and ticketing systems. Update frequency follows a daily schedule or is triggered by each database change. Daily generated documents include database change orders, slow SQL reports, backup records, and permission application forms. Slow SQL reports contain numeric fields such as execution duration and scanned row counts. Backup records include backup size and backup timestamp. Permission application forms include authorized database instance names and account scope. The field structure of each document varies slightly based on the specific operation and maintenance workflow.
Constraints for Traceability Workflow
Data sources are distributed across multiple platforms including database monitoring, audit logs, and ticketing systems. During traceability, unique identifiers from different systems must be associated to avoid mixing up operation and maintenance documents for different instances. The field structure of daily documents varies slightly by workflow, so precise matching of specific fields such as SQL fingerprints and change order numbers is required, rather than generalized full-text matching. Update frequency is triggered by database changes, so the knowledge base must support incremental synchronization configuration to maintain the timeliness of traceability data. Operation and maintenance scenarios have strict accuracy requirements; incorrect traceability may lead to errors in slow query localization, so retrieved documents must strictly match the core characteristics of the query.
Configuration Settings
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
quoteMaxToken | 8000–12000 token | Matches the average length of single database operation and maintenance document content, to avoid truncating key traceability information in slow SQL reports and change orders |
responseData | Enabled | Used to return metadata of original documents, including source system, generation time, and associated database instance ID |
detail | Full | Displays the full path and associated fields of traceability documents, such as node numbers of SQL execution plans |
| Similarity Threshold | 0.6–0.7 | Adapts to the text characteristics of slow SQL reports and change orders, reducing false recall rates for irrelevant documents |
| Single Recall Limit | Top 3 entries | Prevents excessive traceability information from interfering with operation and maintenance personnel's ability to quickly locate core issues |
Common Configuration Mistakes
- Unable to adjust the knowledge base citation limit in the interface: The configuration item for the citation limit is grayed out or has no input box. The cause is that the knowledge base advanced configuration switch for FastGPT v4.17.0 has not been enabled.
- Unable to display answer sources when calling corpus via API only: No metadata fields for traceability documents are shown at the end of the AI response. The cause is that the
responseDataparameter has not been enabled. - Unable to retrieve required operation and maintenance documents: The matching results are unrelated to the current query's slow SQL or change order. The cause is that the similarity threshold is set too low or
quoteMaxTokenis set too small, resulting in truncation of key matching fields.
How to Verify Successful Configuration
- Based on FastGPT v4.17.0, initiate a query targeting a specific slow SQL, and check if the system identifier and associated fields of the source document are displayed at the end of the AI response.
- Enter the knowledge base configuration page, confirm that the value of
quoteMaxTokenfalls within the preset range. - Call the FastGPT API interface, check if the returned
responseDatafield includes document metadata information. - Adjust the similarity threshold and initiate a test, confirm that the matching degree between the retrieved documents and the query content meets expectations.
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 2026-10-19.