What the data for this category looks like
Data for this category mainly comes from the national petition information system, local petition systems, the 12345 citizen hotline platform, grassroots governance platforms, and historical petition archives. The update frequency is daily. Daily documents include petition registration forms, acceptance notice letters, transfer and assignment letters, handling opinion letters, review and re-examination applications, resolution plans, and more. Data fields cover requester subject, matter content, handling milestones, responsible units, handling results, and other items. The volume of a single entry of data gradually increases as the matter process advances.
What constraints these characteristics impose on the knowledge base retrieval and recall link
Mass daily updated petition data requires the retrieval and recall link to support incremental synchronization and high-frequency refresh, to avoid data lag affecting the timeliness of ledger summary. Mixed structured and unstructured documents of multiple types require parsing rules adapted to different document formats, to ensure key information of petition matters is accurately extracted. Data associated with responsible units, handling processes and other fields requires filtering recall results by business attributes during retrieval, to avoid irrelevant petition matters interfering with ledger organization. Coexistence of historical archives and real-time data requires distinguishing retrieval priorities between current handling matters and past pending cases, to ensure information access efficiency for grassroots handling staff. Data errors or delays may lead to repeated petitions, petitions submitted beyond authorized hierarchy, and other consequences, so the accuracy requirement for retrieval and recall is relatively high.
How to Configure
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Adapts to the large differences in wording of petition matters, filters low-correlation recall results, and ensures matching accuracy of core demands |
limit | Top 8–12 entries | Controls the number of single recall results, avoids information overload during ledger summary, and aligns with the information reading habits of grassroots handling staff |
searchMode | hybrid | Combines keyword and semantic retrieval, adapts to scenarios where petition matters include both fixed-format documents and free-text demands |
usingReRank | Enabled | Performs secondary sorting on initial recall results, prioritizes matching key fields such as handling process and responsible units of petition matters |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands related expressions based on user questions, covers synonymous demands and policy terms common in petition matters |
collectionIds | Specified by petition business collection | Limits the retrieval scope to only knowledge base collections related to petition matters, avoids interference from cross-category data |
Three Common Mistakes
- Initial retrieval fails to recall target petition matters, but normal results are returned during the second retrieval. The cause is that
datasetSearchUsingExtensionQueryis not enabled, and no expansion is performed for synonymous expressions of petition demands, leading to semantic matching deviation. - No matching results are returned after importing HTTP response data as a knowledge base reference. The cause is that the response data is not organized into the standard document format for petition matters, and key business fields such as requester subject and handling milestones are not included.
- Retrieval results include non-petition business data. The cause is that the retrieval scope is not limited via
collectionIds, and document content from other business collections is mixed in.
How to Verify Proper Configuration
- Import a piece of historical petition archive data, retrieve a question containing the key demands of the matter, and check whether the matching degree and quantity of recall results meet the configuration requirements.
- View the retrieval log to confirm that the
searchModeandusingReRankparameters have taken effect, and that the retrieval scope only covers the specified petition business collection. - Simulate daily incremental data synchronization, and check whether the retrieval system can promptly recall newly added petition matter documents.
- Test the question expansion function, input a simplified petition demand, and confirm that the system can recall corresponding results after automatically expanding related expressions.
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.