What This Category of Data Looks Like
The knowledge base data for state-owned capital investment companies mainly comes from group official regulatory documents, official notifications from state-owned assets supervision departments, monthly operating reports of subordinate subsidiaries, internal employee service manuals, and departmental business specification documents. Update frequency follows these rules: State-owned asset policy documents are updated quarterly per regulatory requirements, subsidiary reports are updated monthly, and internal office documents are added or revised on demand as needed. Document structure includes fields such as official document number, effective date, responsible department, and business segment affiliation. Some long documents span dozens of pages, covering statistical units including document page count, data amount, and service volume.
What Constraints These Characteristics Impose on Knowledge Base Retrieval and Recall
Official documents carry effective date and document number fields. Retrieval must match the current valid period. Recalling expired documents will cause compliance risks. Monthly subsidiary reports have long single-file lengths and are classified by business segment. When splitting documents into chunks, the segment affiliation field must be retained. This avoids recalling irrelevant fragments across different segments. State-owned asset policies are updated frequently. Knowledge bases must be synchronized regularly to ensure retrieval results use the latest regulatory requirements. Internal employee service documents include job permission information. Recall results must be filtered based on user roles to protect data security.
How to Set Configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Range | Rationale |
|---|---|---|
similarity | 0.72–0.80 | Matches the precise wording of official documents for state-owned capital investment companies, avoids recalling non-compliant content with overly low relevance |
limit | Top 8 entries | Balances information density and retrieval efficiency in internal office scenarios, prevents excessive documents returned in a single round from causing context overflow |
searchMode | semantic_search | Adapts to the semantic association retrieval needs of state-owned asset policies and internal specifications, supports cross-title matching of core business requirements |
usingReRank | Enabled | Filters irrelevant fragments after splitting long documents via re-ranking, improves the recall priority of compliant documents |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands synonym expressions related to state-owned asset policies, covers terminology usage habits of different departments |
collectionIds | Collection IDs divided by business segment | Stores data classified by group subordinate segments, supports quickly limiting retrieval scope by segment to improve retrieval accuracy |
Three Common Configuration Mistakes
- The symptom is that the number of returned retrieval results is much lower than expected. The root cause is that
collectionIdsare not divided by business segment, and redundant cross-segment data interferes with the precise recall range. - The symptom is that expired state-owned asset supervision documents appear in retrieval results. The root cause is that no regular synchronization update mechanism is configured, and invalid policy documents are not filtered, leading to compliance risks.
- The symptom is that user-initiated policy queries do not return documents with the corresponding document number. The root cause is that
datasetSearchUsingExtensionQueryis not enabled, and synonym variants of policy expressions are not covered, leading to matching failures.
How to Confirm the Configuration Is Correct
- Initiate a query containing a known state-owned asset policy document number. Check whether the returned results include documents within the corresponding valid period.
- Switch the retrieval scope to different business segments. Check whether the recalled results only include documents from the corresponding segment.
- Submit an updated state-owned asset policy document. Wait for synchronization to complete, then initiate a query. Check whether the new document is normally recalled.
- Adjust the
similarityparameter value. Check whether the relevance of the returned results meets the requirements of the business scenario.
The values provided 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 2026-09-28.