What the data for this category looks like
State-owned asset management company policy and regulatory data mainly comes from public policy documents issued by the State-owned Assets Supervision and Administration Commission of the State Council, local state-owned assets supervision institutions, as well as internal state-owned capital operation management measures and asset disposal rules. Updates are synchronized when new policies are released. Internal documents are revised once per quarter. Most documents are clause-based texts with document numbers, issuing units, and effective dates. Fields include regulatory hierarchy, asset category, disposal authority, assessment indicators, and similar metrics. Units involve hundreds of millions of yuan, regulatory levels, and other relevant units.
Constraints imposed by these characteristics on the knowledge base retrieval and recall process
Data sources are scattered across official platforms and internal systems. Multi-source retrieval results must be aggregated to cover the complete policy repository. Documents have long clause structures. Chunking must adapt to clause units to avoid splitting that breaks semantic integrity. Updates have no fixed cycle. A combined update mechanism supporting manual triggers and scheduled synchronization is required. Fields include clear regulatory hierarchy and asset categories. Retrieval must associate matching rules for corresponding dimensions to improve result accuracy.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.72–0.80 | State-owned asset management policies use precise wording. Low similarity will introduce irrelevant regulatory clauses. High similarity will miss approximate expressions. This range is calibrated based on actual testing. |
limit | Top 8 results | State-owned asset management policy entries are numerous. Too many results will exceed the context window. Too few will miss associated cross-clause regulatory requirements. |
searchMode | Semantic + keyword hybrid mode | Policy documents have both standard document number keywords and semantic regulatory requirements. The hybrid mode balances precise matching and semantic recall. |
usingReRank | Enabled | Policy documents contain a large number of similar expressions. Re-ranking can filter low-relevance similar results and improve the ranking priority of core clauses. |
datasetSearchUsingExtensionQuery | Enabled | Policy clauses often use abbreviations. Extended queries can cover matches for both full names and abbreviations. |
collectionIds | Specify the two collection IDs: state-owned asset policies and internal management measures | Retrieval scope must be limited to exclude non-relevant business documents and focus on target datasets. |
Three Common Errors
- The symptom is a code:514 error returned when the recall test API is called. The cause is incorrect configuration of the knowledge base collection ID, absence of a valid kb_id parameter, or non-compliance with FastGPT official API request formats.
- The symptom is a number of retrieval results far lower than expected. The cause is an overly high
similarityparameter setting, which filters out a large number of associated policy clauses. - The symptom is missing content in parsed uploaded PDF policy documents. The cause is the document containing encrypted watermarks or being a scanned image file, without enabling the corresponding parsing mode.
How to Verify Successful Configuration
- Access the knowledge base management page of FastGPT v4.17.0, and verify that the
collectionIdsof the corresponding collection match the configured values. - Manually input a state-owned asset policy-related question, trigger retrieval, and check the similarity score of returned results to confirm they fall within the range set by
similarity. - Call the FastGPT official recall test API, pass valid kb_id and question parameters, and confirm the number of returned results matches the limit set by
limit. - Upload a test non-encrypted PDF policy document, and check that parsed text fully covers the original clause content.
The values provided 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 2026-09-28.