What this category of data looks like
Data for this category primarily comes from state-owned assets supervision information systems, enterprise asset management systems, property right exchange systems, evaluation agency reports, and financial statements. The update frequency is monthly. Daily processed documents include property registration forms, asset evaluation reports, property right transfer plans, on-site transaction announcements, and asset disposal approval forms. Data fields cover asset ownership information, evaluation benchmark values, transaction status, approval nodes, and more. Single document length varies widely, with some containing structured tables and unstructured text.
What constraints these characteristics impose on the knowledge base retrieval and recall link
Monthly updated data requires regular synchronization to the knowledge base to prevent expired information from interfering with inquiry responses. Mixed storage of multiple document types requires the retrieval link to adapt to retrieval rules for both structured tables and unstructured text. Query demands for precise fields such as ownership and transaction status require the retrieval mode to support a combination of exact matching and semantic recall. The sensitivity of asset data requires the recall scope of retrieval results to be strictly limited to specified permission sets, while preventing sensitive information leaks. Wide variation in single document length may cause context breaks after chunking, so reasonable retrieval and sorting parameters must be set to ensure complete response content.
How to set the configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches precise fields such as asset ownership and transaction status, avoids mixing low-relevance results, while retaining a sufficient recall range |
limit | Top 10 entries | Adapts to the response needs of a single administrative inquiry, avoids excessive results exceeding the context window |
searchMode | hybrid | Combines semantic recall and exact matching to cover query demands for both structured fields and unstructured text |
usingReRank | Enabled | Performs secondary sorting on initial recall results, improves the ranking priority of high-value assets such as evaluation reports and property right transfer plans |
datasetSearchUsingExtensionQuery | Disabled | The data fields of this category are relatively fixed, no need to extend query terms additionally, avoids introducing irrelevant information |
collectionIds | Specified according to business permissions | Limits the retrieval scope to the asset set of the corresponding state-owned entity, ensures data access security |
Three common mistakes
- The number of asset data entries returned in retrieval results is less than the content in the uploaded Excel file. The cause is failing to correctly configure the
collectionIdsparameter to limit the retrieval set, or setting thesimilaritythreshold too high, filtering out some entries that meet the matching requirements. - Unable to accurately retrieve on-site transaction announcements from a specified property right transaction institution. The cause is not using the hybrid retrieval mode with
searchModeset to hybrid; relying only on semantic recall cannot match structured transaction institution fields. - Duplicate asset registration entries appear in retrieval results. The cause is failing to clean duplicate documents during upload, or failing to bind a unique business collection via the
collectionIdsparameter, leading to duplicate recall of cross-batch data.
How to confirm the configuration is correct
- Upload a single asset evaluation report, enter a query term containing the report number, check whether the title of the returned result matches the query term as expected.
- Enter a query term containing a specific ownership status, check whether the recalled result fields include this status information, verify that the hybrid retrieval of
searchModeis effective. - Switch between different business permission sets, check that the retrieval results only include asset data within the corresponding set, verify that the
collectionIdsparameter is configured correctly. - Adjust the
similarityparameter range, observe the change in the number of recalled results, confirm that the threshold setting meets the precision requirements of business queries.
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 2026-09-28.