What data for this category looks like
Data for this category mainly comes from tracking systems, data warehouses, CRM systems, and ticket systems. Most updates run on a daily or weekly schedule. Daily produced documents include usage analysis reports, tracking plans, indicator specification tables, and dashboard requirement forms. Tracking plans contain event identifiers, trigger conditions, and reporting parameters. Indicator specification tables include indicator names, calculation logic, and statistical dimensions. Usage analysis reports cover customer behavior statistics, feature usage frequency, and similar content. Data fields include event names, user identifiers, statistical cycles, indicator values, business dimensions, and more. There is no fixed unified unit format; formats vary based on indicator type.
Constraints on knowledge base retrieval and recall
Multiple data sources require retrieval to support cross-system data association matching, and avoid only searching documents from a single data source. Daily or weekly update cycles require the knowledge base to support incremental synchronization mechanisms, to ensure the timeliness of documents such as indicator specification tables and tracking plans. Diverse document structures require retrieval to use different matching logic for different document types. For example, use exact matching for indicator specification tables, and scenario-based semantic matching for tracking plans. Specific business fields require retrieval to prioritize matching core business content such as indicator specifications and customer dimensions, and avoid retrieving irrelevant documents.
How to configure settings
These configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches professional terminology for product usage analysis, filters low-relevance general content |
limit | Top 10 entries | Adapts to the volume of multi-source retrieval results, avoids returning excessive redundant data |
searchMode | Semantic search | Adapts to unstructured text matching for tracking plans and indicator specification tables |
usingReRank | Enabled | Reorders retrieval results by business relevance, prioritizes returning core content such as indicator specifications |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands query terms to match event aliases and specification aliases in tracking plans |
collectionIds | Specified by business domain | Restricts retrieval scope to the dedicated knowledge base collection for product usage analysis |
Three common configuration errors
- Symptom: An access restricted error appears when importing external web page links, and document content cannot be synchronized. Cause: No web crawling skip login logic is configured, only publicly accessible links are supported for import.
- Symptom: The chat interface prompts that no knowledge base is selected, but retrieval results work normally during the debug preview phase. Cause: The
collectionIdsof the target knowledge base are not bound in the conversation configuration, so the session is not associated with the preset product usage analysis knowledge base. - Symptom: The number of retrieved results is far lower than expected, and core indicator specification content is not covered. Cause: The
limitparameter is set too low, orsearchModeis not switched to semantic search, resulting in insufficient recall scope.
How to verify configurations are properly set
- Enter the test query "product tracking reporting rules", check whether returned results include document fragments related to tracking plans.
- View retrieval logs, confirm that the matching degree of all returned results exceeds the set
similaritythreshold. - Attempt to retrieve content from an unbound knowledge base, confirm that corresponding results cannot be obtained, to verify that the
collectionIdsconfiguration takes effect. - Trigger an incremental update task, check whether daily or weekly updated documents are synchronized to the knowledge base as scheduled.
The values listed here are common starting points, and should be evaluated 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.