What the Data for This Category Looks Like
Core data for customer requirement pool management comes from ticketing systems, customer interview records, CRM customer profiles, and product roadmap documents. Regularly maintained documents include requirement pool lists, requirement evaluation forms, customer feedback records, and version planning notes. Data is updated weekly or per version release cycle. Data fields include requirement source, customer tier, priority tags, associated version number, and evaluation cost and value dimensions. Individual document content focuses on specific customer feature requests and improvement suggestions, with no unified short-text format.
Constraints Imposed by These Characteristics on Knowledge Base Retrieval and Recall
Since data sources are scattered and update frequency is high, retrieval must be limited exclusively to customer requirement pool-related documents to avoid mixing in business data from other categories. Data includes both structured tags and natural language descriptions, so both semantic matching and precise field filtering must be supported. Individual document content is lengthy, so the length of content recalled in a single operation must be controlled to avoid exceeding the context window. Questions in pre-sales and technical support scenarios mostly revolve around requirement priority and scheduling planning, so retrieval results must be sorted by business priority. Cross-system requirement data must be regularly synchronized to the knowledge base to ensure retrieval content aligns with the latest version planning. Outdated matching content will appear otherwise.
Configuration Settings
These configurations are based on the knowledge base retrieval and recall logic of FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Semantic similarity threshold for matching customer requirement descriptions; balances recall precision and coverage, adapts to the long-text characteristics of requirement pool documents |
limit | Top 8 entries | Number of documents recalled in a single operation; adapts to scenarios where single requirement pool entries are lengthy, avoids context overload |
searchMode | Vector retrieval + keyword hybrid | Combines structured fields and natural language descriptions of requirement pools, improves matching accuracy for cross-source data |
usingReRank | Enabled | Re-ranks recall results by requirement priority and associated customer tier, aligns with the business evaluation logic of pre-sales and technical support |
datasetSearchUsingExtensionQuery | Enabled | Expands user pre-sales or support questions into requirement pool-related search terms, improves matching coverage for short questions |
collectionIds | Specify requirement pool knowledge base collection ID | Limits retrieval scope to only customer requirement pool-related documents, avoids interference from irrelevant business content on retrieval results |
Three Common Mistakes
- Phenomenon: A code:514 error is returned when calling the knowledge base retrieval API. Cause: The
collectionIdsparameter was not specified correctly, or access permissions for the corresponding knowledge base collection were not configured. - Phenomenon: The number of retrieval results does not match the configured
limit. Cause: The automatic retrieval switch for irrelevant knowledge bases was not turned off, or thesimilaritythreshold was set too high, causing valid results to be filtered out. - Phenomenon: Retrieval results do not include associated version planning content. Cause:
datasetSearchUsingExtensionQuerywas not enabled, and user pre-sales questions were not expanded into search terms matching requirement pool version fields.
How to Confirm Proper Configuration
- Run a curl test to call the specified knowledge base retrieval interface, check that the similarity scores of returned results fall within the 0.75–0.85 configuration range.
- View the source field of retrieval results, confirm that only requirement pool-related knowledge base collections are included, with no other irrelevant business documents.
- Submit a test question that includes a requirement priority description, check that returned results are sorted from highest to lowest priority.
- Verify that when
datasetSearchUsingExtensionQueryis enabled, short test questions are expanded into search terms matching requirement pool fields to improve recall coverage.
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.