What the data for this category looks like
Data for this category is maintained by SREs, technical leads, and quality teams. Sources include incident timelines, system logs, monitoring metrics, and code commit records. Daily produced documents include RCA reports, incident tickets, post-mortem minutes, and remediation tracking tables. Fields cover incident occurrence time, root cause description, associated log snippets, verification steps, and remediation measures. Data update frequency is once after each major incident, or once per month. Single document content length varies widely, with some reports containing multiple sections of structured and unstructured content.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
This category has low data update frequency, so real-time indexing is not required. However, the wide variation in single document content length requires the retrieval system to support recall of long text snippets. Data sources include multiple types of structured and unstructured content, so retrieval scope must be limited to exclude unrelated monitoring metrics or code commit records. There is a strong correlation between root cause descriptions and remediation measures in documents. Recall should prioritize content related to root causes, and resolve conflicts in root cause descriptions across different documents. When dynamically passing knowledge base retrieval parameters, ensure the associated knowledge base set is correct to avoid failing to retrieve target documents.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75-0.85 | Root cause analysis requires highly relevant documents to filter out non-associated low-interference content |
limit | Top 5-8 results | Control the total length of recalled content to avoid exceeding the large model context window |
usingReRank | Enabled | Perform secondary relevance ranking on recall results to resolve document content conflicts |
datasetSearchUsingExtensionQuery | Enabled | Expand retrieval keywords to cover implicit expressions related to root cause analysis |
collectionIds | Specify the collection ID of the target incident management knowledge base | Limit retrieval scope to exclude non-category-related document data |
Three Common Misconfigurations
- Troubleshooting item: Retrieval results fail to hit target RCA reports. Verify if
collectionIdswas not specified, leading to retrieval scope covering non-incident management knowledge base data. - Troubleshooting item: Conflicting document fragments appear in returned results. Verify if
usingReRankwas not enabled, so no secondary relevance ranking was performed on recall results. - Troubleshooting item: The text understanding model list is empty when creating a knowledge base. Verify if
AIPROXY_API_ENDPOINT(root address of the AI Proxy service, do not append /v1 or other paths, not a single-model address) andAIPROXY_API_TOKEN(AI Proxy administrator token, consistent with server configuration, cannot be empty, not an application/model vendor key) environment variables were not configured on the App/Admin server, causing service startup verification failure.
How to Confirm Proper Configuration
- Check if the App/Admin server has configured
AIPROXY_API_ENDPOINTandAIPROXY_API_TOKENas required by FastGPT v4.17.0, confirm no empty values or format errors. - Verify that
collectionIdspoints to an existing incident management knowledge base set, which can be viewed via the knowledge base management page. - Initiate a simulated retrieval request with preset root cause-related questions, confirm that the
similarityscores of returned results fall within the preset range. - Check retrieval logs to confirm that recall results do not include unrelated monitoring metrics or code commit records, and there are no conflicting content fragments.
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.