What data for this category includes
Core documents for this category typically include basic government service item catalogs, implementation lists, service guides, standard material directories, and flowcharts. Data is primarily sourced from the national basic government service item catalog, administrative licensing item list, power and responsibility list, and local government service item databases, with a monthly update cycle. Most documents are structured tables and standardized texts, with fixed fields including item codes, implementation levels, acceptance conditions, and application material lists. No custom unit-type fields are present.
Constraints on knowledge base retrieval and recall
Regular synchronization tasks must be configured for the knowledge base to accommodate monthly updated data sources, preventing data lag from disrupting business operations. Structured document formats require retrieval processes to balance matching accuracy for table fields and text content, avoiding field misalignment caused by fuzzy matching. Multi-source cross-level data must be separated using dataset grouping for national, provincial, municipal, and county-level catalogs, preventing mixing of item entries from different administrative levels. Fixed fields require retrieval and recall results to prioritize matching core identifiers such as item codes and implementation levels, ensuring alignment between retrieval results and business needs.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.72–0.80 | Aligns with field matching accuracy for structured government documents, avoiding recall of irrelevant item entries |
limit | Top 8 entries | Covers associated documents for a single item (catalogs, guides, material lists, etc.) while controlling context length |
searchMode | Hybrid retrieval | Matches both text keywords and structured fields, adapting to the multi-dimensional content of government documents |
usingReRank | Enabled | Performs secondary sorting on initial recall results to correct priority deviations from keyword matching |
datasetSearchUsingExtensionQuery | Enabled | Generates extended query terms based on the original question, covering synonymous expressions in policy documents |
collectionIds | Grouped by administrative level | Distinguishes national, provincial, municipal, and county-level datasets, preventing cross-level data mixing |
Three Common Misconfigurations
- Symptom: Retrieval results display document download links in debug mode, but links disappear after application publishing. Cause: Document reference permissions are not enabled in the publishing configuration, or the dataset does not have public access permissions configured.
- Symptom: Vectorization takes excessive time after uploading government service item PDF documents. Cause: Document chunk optimization is not enabled, or documents larger than the system processing limit are used.
- Symptom: The reference limit set in the knowledge base search configuration does not match the limit set in the workflow search configuration. Cause: Failure to note that default retrieval parameter configurations differ between the application end and workflow end, and failure to manually unify configuration item values.
How to Verify Successful Configuration
- Run a debug mode retrieval, confirm that recall results include government service item documents from the corresponding administrative level, and that matching fields meet expectations.
- Trigger a monthly synchronization task, confirm that the knowledge base automatically pulls the latest item catalog and guide documents after the data source updates.
- View retrieval logs, confirm that similarity scores of recall results fall within the configured range, and that sorted results have a reasonable ranking.
- Test queries for items at different administrative levels, confirm that datasets corresponding to different
collectionIdsonly return documents from the matching administrative level.
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.