Reference only; verify against the target FastGPT release and deployment documentation.
What data for this category looks like
Data for this category typically comes from the national integrated government service platform's "One Task" special zone, local integrated government service platforms, electronic certificate repositories, and departmental approval systems. Update frequency is daily. Commonly processed documents include joint service item lists, one-time notice sheets, joint application forms, material transfer forms, parallel approval opinions, and completion receipts. Common data fields include cross-departmental item association relationships, application material lists, processing flow nodes, and approval status identifiers. The data uses general government service standard formats, with no custom special units.
What constraints do these characteristics impose on knowledge base retrieval and recall
Daily updated joint initiative data requires retrieval processes to support incremental synchronization, to avoid recalling expired item lists or approval opinions. Cross-departmental combined processing requires retrieval to match cross-departmental item association relationships simultaneously, rather than only recalling single-department independent item content. Diverse document structures require retrieval systems to adapt to notice sheets, transfer forms, and other documents of varying lengths, to avoid losing critical information due to long text truncation. Data from different sources must be categorized by department or special zone, to ensure retrieval scope accurately corresponds to covered departments of joint initiative items and prevent cross-source data confusion.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches cross-departmental association descriptions of joint service items. Avoids introducing irrelevant results if set too low, and missing association relationships if set too high |
limit | Top 8 entries | Covers multi-department item nodes and material requirements, while controlling context length within model processing limits |
searchMode | hybrid | Combines keyword and semantic retrieval, adapts to mixed scenarios of structured item lists and unstructured notice sheets for joint service items |
usingReRank | Enabled | Rearranges recall results by relevance, optimizes association priority of cross-department items, and improves response accuracy |
datasetSearchUsingExtensionQuery | Enabled | Expands user-consulted joint initiative questions, supplements implicit information such as department associations and material reuse |
collectionIds | Specified by covered departments of joint service items | Limits retrieval scope to joint initiative exclusive datasets of corresponding departments, avoids interference from cross-category data |
Three Common Mistakes
- Phenomenon: Importing knowledge base data exported from an older version to a new version results in training data errors. Cause: CSV template fields differ between versions, and the data structure is not adjusted to match the target version's template format.
- Phenomenon: The number of returned results from knowledge base search cannot be adjusted as needed. Cause: The
limitparameter is not configured correctly, or the parameter value does not match the multi-department data coverage requirements of joint service items. - Phenomenon: The reranking function cannot be used in large model calls, but works normally in independent knowledge base search tests. Cause: The
usingReRankoption is not enabled in the large model's retrieval configuration, only enabled during the knowledge base test phase.
How to Confirm Configurations Are Set Correctly
- Upload test documents for corresponding joint service items, perform incremental synchronization, and check if data update times match daily update requirements.
- Input test questions involving cross-departmental item associations, and check if retrieval results cover relevant documents from multiple departments and meet expected quantity requirements.
- After enabling
usingReRank, compare retrieval result rankings before and after enabling, to confirm the reranking logic takes effect. - Call the API to retrieve retrieval corpus, and check if returned results include document metadata, which can be used for subsequent answer source display.
Note: The configuration values listed are common starting points. Adjust them based on your own sample data.
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.