What This Category of Data Entails
Typical users of this category include judicial administrative departments, departmental legal affairs offices, and policy and legal affairs positions in government offices. Daily processed documents include regulatory document drafts, drafting explanatory notes, legality review opinions, solicited comment summaries, and filing reports. Data sources cover regulatory and rule databases, higher-level law databases, departmental policy document databases, and administrative regulatory document filing systems. Update frequency is irregular. Each document includes fields such as document number, issuing authority, issuance date, core provisions, and review opinions. This category reviews abstract policy rules, which differs from individual administrative law enforcement cases. Its focus lies in formulation authority, review procedures, and the legality of provisions.
Constraints on Knowledge Base Retrieval and Recall
Scattered data sources and irregular updates require the retrieval link to support incremental synchronization and on-demand refresh. This prevents recalling expired or unincluded review references. Daily documents include multiple structure types such as drafts, review opinions, and solicited comment summaries. Some documents contain non-provision auxiliary content. This requires retrieval to filter irrelevant fields, and only return core provisions and policy references related to legality review. This category focuses on reviewing the formulation authority, procedures, and legality of abstract policy rules. It requires strict matching with higher-level laws and relevant authorities. Recalled results must accurately correspond to provision levels and issuing authorities. They must also cover all relevant solicited comments and filing records to ensure no gaps in review.
Configuration Guidelines
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Requires accurate matching of higher-level laws and policy provisions needed for legality review. This range balances recall coverage and precision, and avoids recalling irrelevant content |
limit | Top 10 | Legality review of regulatory documents requires covering multi-dimensional references. Recalling 10 entries at a time ensures coverage of major relevant documents while controlling context length |
searchMode | vector+keyword | Requires matching both semantics and keywords, to ensure retrieval of both semantically relevant policies and explicitly mentioned regulatory provisions, and meet the precision requirements of review |
usingReRank | Enabled | Performs secondary ranking on initial recall results, filters low-relevance entries, improves the quality of final returned references, and fits the rigorous requirements of legality review |
datasetSearchUsingExtensionQuery | Enabled | Generates extended queries based on original questions, covers synonyms and near-synonym expressions, and avoids missing relevant references due to wording differences |
collectionIds | Fill in the dataset IDs of the law databases and filing systems bound to the business | Only retrieves compliance datasets required for the current business, avoids introducing irrelevant content, and ensures the retrieval scope meets review requirements |
Three Common Misconfigurations
- Phenomenon: Retrieval results include administrative law enforcement documents unrelated to legality review. Cause: Did not limit the retrieval scope via
collectionIds, and accidentally included non-regulatory document datasets in the retrieval pool. - Phenomenon: The large model output only summarizes the review conclusion without citing original text fragments. Cause: Did not enable
usingReRankordatasetSearchUsingExtensionQuery. Initial recalled original text fragments have insufficient matching with review requirements, and cannot be effectively associated. - Phenomenon: Uploaded regulatory document drafts are identified as garbled text. Cause: Did not adapt to the encoding and typesetting rules of official document formats, and failed to correctly identify the file content structure during chunking, leading to parsing exceptions.
How to Verify Successful Configuration
- Run a test retrieval, input a legality review requirement for a regulatory document, and check whether the dataset scope of returned results matches the preset
collectionIds. - Check the sorting of retrieval results, confirm that highly relevant higher-level laws and policy provisions are displayed first, to verify that
usingReRankis active. - Trigger an incremental synchronization task, check whether newly added filing files are included in the retrieval scope within a reasonable period, to verify the configuration of the update mechanism.
- After generating an extended query, check whether the retrieval results cover policy expressions not directly mentioned in the original question but relevant, to verify that
datasetSearchUsingExtensionQueryis active.
The values provided here are common starting points, and should be measured against the user’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.