What this category of data looks like
This category of data originates from scheduling systems, fault work orders, monitoring alerts, and organizational contact directories. Daily documents include duty rosters, handover records, escalation contact lists, and daily duty reports. Data is updated daily or weekly. Core fields include duty time periods, handover items, escalation contact information, unresolved fault status, and other relevant details. Scheduling rules and handover templates vary across teams, and data structures adjust based on team operational workflows.
What constraints these characteristics impose on knowledge base retrieval and recall
Scattered data sources and heterogeneous structures require limiting retrieval scope to exclusive business collections, to prevent irrelevant data from interfering with search results. Daily documents contain strongly time-ordered fields such as duty time periods and unresolved items. Retrieval must prioritize matching content from the current duty cycle, and filter expired historical scheduling information. Data updates occur daily or weekly, so regular incremental knowledge base synchronization is required to ensure search results include the latest duty arrangements and handover records. Core scenarios also rely on associated information such as escalation contacts and fault status, so retrieval and recall must cover multi-field associated matching.
Configuration settings
All following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
collectionIds | Specified by business scenario, for example ["duty scheduling collection","handover records collection","escalation contacts collection"] | Limit retrieval scope to exclusive data for this category, exclude irrelevant content from other business knowledge bases |
limit | Top 8-12 entries | This category has relatively few core data entries; excessive results increase screening workload for duty personnel |
similarity | 0.75-0.85 | Balance retrieval precision and recall completeness, avoid missing recently updated scheduling or handover information |
usingReRank | Enabled | This category of data contains strongly time-ordered and associated fields; re-ranking prioritizes returning results matching duty time periods and unresolved items |
datasetSearchUsingExtensionQuery | Enabled | Automatically expand search terms to cover synonymous expressions of scheduling rules and handover processes, improving recall coverage |
searchMode | Semantic retrieval | This category of data includes unstructured handover records and scheduling descriptions; semantic retrieval is better suited to the business scenario than keyword matching |
Three common misconfigurations
- Phenomenon: Search results include non-duty-scheduling business documents unrelated to the current scenario. Cause: The
collectionIdsparameter was not used to limit retrieval scope, or the configured collection ID does not match the actual business collection. - Phenomenon: Retrieved duty documents only show titles and summaries, and complete handover record content cannot be accessed. Cause: Document extended retrieval configuration was not enabled, and content crawling rules for duty documents were not configured.
- Phenomenon: Embedded images fail to load properly after a markdown-formatted duty roster document is uploaded. Cause: Image resources were not synchronized to the knowledge base storage directory during file upload, or image paths in the original markdown file were not replaced with absolute addresses accessible by the knowledge base.
How to verify correct configuration
- Manually enter a recent duty time period keyword, check if search results only include documents such as duty rosters and handover records from this category.
- Adjust the value range of the
similarityparameter, verify whether the precision of search results meets business requirements, and confirm that the threshold setting is reasonable. - Upload a test markdown-formatted duty document, check if embedded images load normally, and confirm that document resource configuration is correct.
- View retrieval logs to confirm that only specified business collections were called, and no irrelevant knowledge base content was retrieved.
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.