What this category of data looks like
Data sources for this category include government portal websites, document management systems, government information disclosure directories, policy document repositories, and press conference materials. The update frequency is daily. Daily documents include policy documents, policy interpretation drafts, public disclosure directories, information release approval forms, press releases, and website column update records. Data fields include disclosure attribute, confidentiality level, release time limit, affiliated column, approval status, and some documents include release approval process records and multi-channel synchronization markers.
What constraints these characteristics impose on the "knowledge base retrieval and recall" link
Daily updated data sources require the retrieval link to support incremental synchronization and incremental recall, to avoid excessive resource consumption from full scans. Structural differences across multiple document types require retrieval configurations to adapt to the field characteristics of different documents. For example, policy documents should prioritize matching document numbers and clauses, while public disclosure directories should match columns and release times. The presence of disclosure attribute and confidentiality level fields requires retrieval results to automatically filter content that does not meet public disclosure standards, to prevent sensitive information from being leaked. Multi-channel synchronization markers require recalled results to match the column rules of the target release channel, ensuring that retrieved content complies with the requirements of the corresponding release scenario.
Reference only; verify against the target FastGPT release and deployment documentation.
How to set configurations
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Filter low-match content to avoid irrelevant policies or interpretations from mixing into retrieval results, meeting the accuracy requirements of government affairs content |
limit | Top 8 entries | Control the number of single-recall results, adapt to the quick review needs of internal office scenarios, while avoiding excessive content consuming context |
searchMode | hybrid | Combine semantic and keyword matching, balancing precise clause matching for policy documents and keyword retrieval needs for public disclosure directories |
usingReRank | Enabled | Re-rank initial recall results to raise the display priority of highly relevant content, adapting to differences in the importance of policy content |
datasetSearchUsingExtensionQuery | Enabled | Automatically expand retrieval terms, supplement synonyms and abbreviations related to policies, adapting to the diversity of government affairs terminology |
collectionIds | Configured by business grouping | Differentiate between different datasets such as public disclosure directories and policy documents, ensuring the retrieval scope is limited to corresponding business documents |
Three common mistakes
- The number of retrieval results returned does not match the configured
limitparameter. The symptom is that the number of results displayed on the interface exceeds or falls below the preset value. The cause is failure to synchronously update the parameter configuration of the retrieval node in the workflow, or confusion between global and single-node parameter settings. - CSV file upload fails to split content using a custom delimiter. The symptom is that file content is not split into independent text blocks as expected, with entire rows remaining unsplit. The cause is failure to correctly check the custom delimiter option, or the input delimiter contains invisible characters.
- User authentication for retrieval nodes does not take effect. The symptom is that unauthorized internal employees can still access confidential or undisclosed policy documents. The cause is failure to bind authentication rules to the corresponding knowledge base collection, or the scope of the authentication configuration does not cover the dataset of the current retrieval node.
How to confirm the configuration is complete
- Upload a test policy document and public disclosure directory, perform a retrieval operation, and verify that the number of returned results matches the preset value of the
limitparameter. - Simulate a retrieval request for a confidential document, confirm that this type of document is not included in the results, to verify that the filtering rule is effective.
- Upload a CSV test file using a custom delimiter, check whether the text block splitting results meet expectations.
- Check the authentication configuration of the retrieval node, confirm that only users within the authorized scope can trigger the retrieval operation.
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.