What This Category's Data Looks Like
Core data for this category comes from the national land and space planning one-map platform, project approval and supervision platform, land use status database, and ecological protection red line data, with irregular update cycles. Daily processed documents include land pre-examination and site selection opinions, construction land planning permits, construction engineering planning permits, red line drawings, planning conditions, and more. Red line drawings contain spatial coordinate fields. Planning conditions include structured content such as land scale and control indicators. Approval files include combined documents such as project declaration forms and prior approval opinions.
Constraints Imposed on Knowledge Base Retrieval and Recall
Scattered data sources and irregular update cycles require the retrieval system to support incremental synchronization of multi-source heterogeneous data, avoiding resource consumption from full retraining. Documents include structured fields such as spatial coordinates and control indicators, as well as unstructured content such as vector drawings and approval texts. This requires the system to support both text semantic matching and structured attribute retrieval. Project site selection verification and red line conflict review are core time-consuming steps. Recalled results must accurately associate planning control rules with the current project status, otherwise approval errors may occur.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirements for planning control rules, avoids low-relevance results interfering with review |
limit | Top 10 entries | Balances retrieval efficiency and information completeness, covers the recall scope of core approval basis |
searchMode | Hybrid retrieval | Supports both text semantic and structured field matching, adapts to multi-type document characteristics |
usingReRank | Enabled | Performs secondary sorting on initial recall results, improves the priority of key content such as red line conflicts and control requirements |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands query terms, covers synonymous expressions of professional terms such as project numbers and planning indicators |
collectionIds | Exclusive knowledge base collection ID for this category | Isolates approval data of different categories, avoids cross-category retrieval interference |
Three Common Mistakes
- Phenomenon: Retrieval results for structured approval tables cannot match land scale values within cells. Cause: Multi-vector support for table data is not enabled, and vectors are only generated for full table text, making accurate matching of cell-level professional fields impossible.
- Phenomenon: A 404 status code is returned when calling the API to pull the document library. Cause: The
collectionIdsparameter is not configured correctly, pointing to a non-existent knowledge base collection. - Phenomenon: Retrieval results for long-text planning conditions lose contextual association and cannot match complete control requirements. Cause: Model-based paragraph recognition is not enabled during chunking, with maximum paragraph depth set to 3 and maximum chunk size set to 1000 characters, leading to incorrect splitting of long text.
How to Confirm Proper Configuration
- Upload a test document containing structured approval tables and planning condition text, verify that the knowledge base chunking results retain cell-level fields and control clause context.
- Initiate a test query containing keywords such as land scale and ecological protection red line, check that the sorting of retrieval results conforms to the secondary sorting configuration logic.
- Call the FastGPT v4.17.0 retrieval API, pass the specified
collectionIdsparameter, and confirm that only knowledge base content for this category is returned. - Check the knowledge base synchronization logs, confirm that incremental update tasks are automatically triggered according to the irregular update schedule.
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.