What Data for This Category Looks Like
Typical data sources for this category include CDN consoles, DNS resolution records, user access logs, and node monitoring alerts. Daily produced documents include domain access forms, cache rule configuration forms, refresh and preheat operation records, and node fault analysis reports. Fields covered include node IP addresses, cache TTL values, access bandwidth thresholds, error code types, and more. Data update frequency is either daily sync or real-time update when configuration changes. Document structures typically combine structured configuration items and unstructured fault troubleshooting steps.
Constraints Imposed on Knowledge Base Retrieval and Recall by These Characteristics
Since data sources include real-time monitoring and historical logs, the knowledge base must support both incremental synchronization and full data recall modes to avoid recalling expired cache rules or node information. The mixed document structure requires retrieval to balance keyword matching and semantic understanding. For example, it must accurately match domain name parameters while understanding technical expressions such as "origin pull timeout". High-frequency time-consuming steps such as domain record verification and cache rule debugging require retrieval results to prioritize matching target documents, reducing manual screening costs. Additionally, errors directly impact public network access experience, so retrieval result accuracy must be strictly controlled to avoid incorrect configuration suggestions.
How to Set Configurations
This configuration is based on FastGPT v4.17.0
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | CDN-related inquiries mostly focus on technical details. This range balances precise matching and recall coverage, avoiding missed edge node troubleshooting inquiries |
limit | Top 10 entries | CDN operation and maintenance issues involve multi-dimensional logs and configuration items. Recalling 10 entries at a time covers the associated documents required for troubleshooting, avoiding redundant results |
searchMode | keyword+vector hybrid mode | Balances explicit keyword matching for domain names and error codes in CDN documents, as well as semantic content recall for fault troubleshooting |
usingReRank | Enabled | CDN issues often have multiple documents with similar expressions. Reranking prioritizes content most relevant to the current inquiry, such as cache rule debugging issues |
datasetSearchUsingExtensionQuery | Enabled | CDN customer inquiries often include shorthand terms, such as "CDN origin pull". Extended queries supplement associated terms to improve recall coverage |
Three Common Misconfigurations
- Symptom: A
timeout of 60000ms exceedederror occurs when deleting a folder containing a large number of CDN cache rule documents in the knowledge base. Cause: No reasonable batch processing parameters were set for bulk document operations, and the number of documents processed in a single operation exceeded the interface's allowed timeout threshold. - Symptom: Some CDN console screenshots fail to load properly in the knowledge base question-and-answer interface. Cause: The image storage domain name of the knowledge base data source was not added to the allowed access whitelist, preventing the front end from loading external link images.
- Symptom: Embedded configuration screenshots in documents cannot be correctly retrieved when connecting a third-party team collaboration knowledge base. Cause: The knowledge base image metadata synchronization configuration was not enabled, so the fault troubleshooting content associated with the images cannot be linked during retrieval.
How to Confirm Configurations Are Correctly Set
- Submit a simulated inquiry containing "cache rule debugging", check if retrieval results prioritize returning the corresponding cache rule documents and that sorting matches expectations.
- Perform a bulk delete operation on a small document folder, confirm no timeout errors occur to verify that the batch processing configuration is effective.
- Upload a 403 error screenshot from the CDN console to the test knowledge base, submit an inquiry containing "403 error", confirm that the retrieval result can link to this screenshot document.
- Configure a public third-party document link as a data source, submit an inquiry containing "document parsing", confirm that the corresponding content can be retrieved normally.
All provided configuration values are common starting points. They should be evaluated against the reader's own specific samples and use cases.
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.