What the data for this category looks like
Data for this category mainly comes from the national land spatial planning unified map, project approval and supervision platform, land use status database, and ecological protection redline data. The update frequency is irregular. Daily generated business documents include land pre-examination and site selection opinions, construction land planning permits, construction engineering planning permits, redline drawings, planning conditions, and more. Documents contain fields such as project site selection scope, land scale, planning control requirements, and redline coordinates. Some drawing documents include spatial coordinates and geographic boundary information.
What constraints do these characteristics impose on the "knowledge base retrieval and recall" link
Data is scattered across multiple government data sources. This requires the retrieval link to support cross-dataset associated calls to avoid incomplete information from a single data source. Irregular updates require the retrieval system to support incremental synchronization mechanisms to ensure recalled content aligns with the latest control requirements. Documents contain both text rules and spatial coordinate information, so both semantic retrieval and spatial boundary matching must be supported. Business processes are linked to preconditions from multiple departments. The retrieval scope must be limited based on the current project stage, only recalling compliance requirements for the corresponding approval node to improve retrieval accuracy.
How to set the configuration
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Filter low-relevance planning control content to avoid interference from land use rules of non-target projects on retrieval results |
limit | Top 8–12 entries | There are many documents related to land pre-examination. Limiting the number of recalled entries can compress the context length while covering core control requirements |
searchMode | Semantic Retrieval + Spatial Association | Documents in this category contain both text rules and spatial coordinates, so both semantic and geographic boundary conditions need to be matched |
usingReRank | Enabled | Reorder recalled results by project relevance to optimize the display priority of content related to preconditions |
datasetSearchUsingExtensionQuery | Enabled | Automatically expand retrieval keywords based on project pre-processes to cover pre-approval requirements from relevant departments |
collectionIds | Collection of dataset IDs associated with the project | Only retrieve planning and approval datasets related to the current land project to avoid irrelevant knowledge base content being included |
Three common mistakes
- Phenomenon: Retrieval results include ecological protection redline or planning content from other regions unrelated to the current land project. Cause: Failed to use
collectionIdsto limit retrieval to only planning and approval datasets associated with the project, resulting in recalled data from cross-regional or cross-category sources. - Phenomenon: Generated responses exceed the scope of the knowledge base and introduce external common knowledge not retrieved. Cause: Failed to enable generation constraints based on retrieval results, or failed to correctly configure retrieval scope filtering rules.
- Phenomenon: Unable to export all uploaded business document details. Cause: Failed to configure file export permissions for the dataset, or failed to correctly associate the corresponding data collection.
How to confirm the configuration is correct
- Upload core business documents of this category, initiate a query including "Does the project site selection comply with ecological protection redline requirements", and check whether the recalled results include corresponding redline data and planning conditions.
- Adjust the
similarityparameter, observe changes in the relevance of retrieval results, and confirm that the threshold setting meets the matching accuracy requirements of business content. - Initiate a general question that does not appear in the knowledge base, check whether the system outputs a preset custom response, and confirm that the generation constraint configuration takes effect.
- View the dataset management interface, confirm that
collectionIdshas been configured as the collection of dataset IDs associated with the current project, and no irrelevant data sources are included in the retrieval scope.
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.