What this category of data looks like
This category of data originates from the national petition information system, local petition systems, the 12345 citizen hotline platform, grassroots governance platforms, and historical petition archives. Daily processed documents include petition registration forms, acceptance notices, transfer and assignment letters, handling opinions, review and reexamination applications, resolution plans, and more. The document structure is designed around the petition handling workflow. Fields cover acceptance scope definitions, handling procedure explanations, review and reexamination rules, and related content. The data update frequency is daily, requiring synchronization of the latest policy adjustments and handling records.
Constraints Imposed on Knowledge Base Retrieval and Recall
Multiple data sources require isolating different knowledge bases by business scenario during retrieval, to avoid mixing content across systems. The daily update frequency requires regular synchronization of the latest policies and archives to ensure the timeliness of retrieved content. The specific document structure and fields must match the core demands of petition handling. During retrieval, content related to acceptance scope and compliance verification must be prioritized for recall. Retrieval errors directly relate to issues such as repeated petitions and petitions submitted beyond the appropriate hierarchy. As a result, accuracy and relevance requirements for retrieval and recall are higher. Low-relevance irrelevant content must be filtered out to ensure the compliance of handling opinions.
Configuration Settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
collectionIds | Configure dedicated collection IDs classified by petition business | Match the multi-source data sources of petition matter handling, and distinguish different business documents such as policy and regulation libraries, historical archive libraries, and compliance verification libraries |
similarity | 0.75–0.85 | Ensure that retrieved content matches the core expressions of policy basis, filter low-relevance auxiliary content, and ensure the accuracy of petition handling judgments |
limit | Top 8 entries | Control the number of returned results, adapt to the scale of policy entries that need to be referenced once during petition handling, and avoid information overload |
usingReRank | Enabled | Optimize ranking for semantic differences in policy texts, and improve the display priority of core content such as compliance verification and acceptance scope determination |
datasetSearchUsingExtensionQuery | Enabled | Supplement synonymous extended expressions of policy terms, match colloquial or variant expressions common in petition handling, and improve recall coverage |
Three Common Misconfigurations
- Retrieval results are not displayed classified by business knowledge base. Phenomenon: Returned results mix policy documents of different business types, and cannot distinguish between business systems and compliance verification content. Cause: The dedicated collection ID is not specified via the
collectionIdsparameter, and the retrieval scope of different knowledge bases is not isolated. - Large document upload tasks are blocked and cannot be manually intervened. Phenomenon: When uploading petition archives larger than 100 MB, the task remains unresponsive for a long time and there is no pause control option. Cause: Scheduling logic that prioritizes uploads by document size is not configured, and user-controllable upload pause functions are not enabled.
- Retrieval results do not prioritize matching core policy content. Phenomenon: Auxiliary explanation content accounts for a high proportion in the recall results, and core policy basis is ranked at the end. Cause: The
usingReRankreranking function is not enabled, and a reasonablesimilaritythreshold is not set to filter low-relevance content.
How to Verify Correct Configuration
- Retrieve specified petition policy keywords, and check whether the returned results only include the knowledge base content corresponding to the configured
collectionIds. - Upload test documents of different sizes, confirm that scheduling prioritizing uploads by document size can be triggered, and that upload tasks can be manually paused and resumed.
- Initiate a query related to the petition acceptance scope, and check whether the reranked results place core policy basis at the top.
- Enter colloquial expressions of policy questions, verify that the corresponding official documents can be recalled, and confirm that the
datasetSearchUsingExtensionQueryfunction is active.
The values provided here 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.