What this category of data looks like
Daily documents for this category include land pre-review and site selection opinions, construction land planning permits, construction engineering planning permits, red line drawings, and planning conditions. Data sources include the national spatial planning one-map system, project approval and supervision platform, current land use database, and ecological protection red line data. Document structure falls into three categories: structured approval forms, vector drawing attachments, and text approval opinions. Core fields include project number, land area (unit: hectare), planning control zone, red line coordinates, and others. Data updates follow no fixed cycle, and are triggered as individual projects progress or regional planning adjustments are made.
What constraints these characteristics impose on the knowledge base retrieval and recall link
Data for this category comes from multiple official systems, with significant format differences. This requires the retrieval process to adapt to mixed retrieval scenarios of structured forms, vector drawing metadata, and plain text approval opinions. Data updates have no fixed rhythm, so flexible configuration of incremental synchronization and full updates is needed to avoid retrieving outdated planning control data. Documents contain professional geospatial fields and approval terminology, so precisely matched content must be prioritized for retrieval to avoid irrelevant information from generalized searches. Daily core work involves checking conflicts between project site selection and red lines, so retrieval must quickly locate relevant control data to support efficient review for internal office work.
How to set the configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Match professional terms related to land pre-review and planning control, filter low-relevance non-business data |
limit | Top 10 entries | Adapt to the information demand of single approval in internal office work, avoid excessive redundant content interfering with review |
searchMode | Precise matching | Prioritize retrieving precisely matched business content for strongly related documents such as red line drawings and planning conditions |
usingReRank | Enabled | Perform secondary ranking on initially retrieved results, improve the priority of results related to red line conflicts and planning control zones |
datasetSearchUsingExtensionQuery | Enabled | Supplement retrieval coverage for structured fields such as project number and land area, improve recall rate of cross-source data |
collectionIds | Fill in according to the actually bound natural resource approval dataset | Limit the retrieval scope, avoid irrelevant business data interfering with the core approval process |
Three common mistakes
- The symptom is that a result is returned only after entering the same question twice, with no result returned on the first retrieval. The cause is that the query extension optimization of
datasetSearchUsingExtensionQueryis not enabled for the first retrieval, or thesimilaritythreshold is set too high, filtering eligible professional documents. - The symptom is that the specific configuration value of
collectionIdscannot be obtained. The cause is that the unique identifier of the bound collection was not viewed in the corresponding knowledge base details page of the dataset management page in FastGPT v4.17.0. - The symptom is that no matching results are returned after connecting an external knowledge base. The cause is that field mapping was not completed according to the format requirements of national spatial planning data, or the correct data source synchronization update rules were not configured.
How to confirm the configuration is properly set
- Enter a professional question that includes project number, land area, and planning control zone, verify that matching business documents are returned on the first retrieval.
- Check the number of returned retrieval results, confirm that the quantity does not exceed the upper limit set by
limit. - Check the sorting logic of retrieval results, confirm that after
usingReRankis enabled, results related to red line conflicts and planning control are ranked first. - Verify the configuration of
collectionIds, confirm that only natural resource approval-related datasets are bound, with no irrelevant business data mixed in.
The values provided here are common starting points and should be evaluated 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.