What the data for this category looks like
Daily documents for this category include territorial spatial overall plans, detailed plans, special plans, planning implementation evaluation reports, and public opinion summaries. Data sources include the territorial spatial basic information platform, natural resource survey and monitoring data, population and economic statistics data, and remote sensing images. The update frequency is annual. Most documents contain fields such as spatial coordinates, constraint indicators, and regional scope. Some documents involve integration of multi-source spatial data, and the structure varies with planning hierarchy and type, with no unified fixed single-page format.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
Annually updated data means the knowledge base must perform regular synchronization operations to avoid using expired spatial data or planning indicators. Multi-source and heterogeneous document structures and fields require the retrieval link to support field matching across different document types. Retrieval limited only to plain text keywords cannot cover all requirements. Regional-level planning documents have large volume, so the content retrieved in a single operation must be controlled to prevent exceeding the context window. Public opinion summaries have a high proportion of unstructured text, so retrieval and recall must adapt to colloquial, fragmented opinion content to improve matching accuracy.
How to set the configuration
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision of technical terms in territorial spatial planning documents, avoiding recall of irrelevant non-planning data |
limit | Top 6–10 entries | Adapts to the volume of regional planning documents, controls the context length of a single recall, and aligns with the information density of planning document drafting |
usingReRank | Enabled | Optimizes sorting of multi-source heterogeneous documents, prioritizing recall of content directly related to planning indicators and spatial constraints |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands professional terms related to planning, such as expanding "three zones and three lines" to expressions related to ecological protection red lines, permanent basic farmland, and urban development boundaries |
collectionIds | Bound according to planning document categories | Restricts the retrieval scope to only territorial spatial planning knowledge base collections, excluding irrelevant business data |
Three common misconfigurations
- Symptom: After enabling
usingReRank, retrieval results are empty or only return a small amount of irrelevant content. Cause: Thesimilaritythreshold is set too high, resulting in an insufficient initial recall candidate set, with no valid data available for the re-ranking step. - Symptom: Retrieved document fragments do not match the requirements of planning drafting, with a large amount of non-spatial planning content appearing. Cause:
collectionIdsare not bound to a dedicated planning knowledge base collection, and the retrieval scope covers irrelevant business data. - Symptom: Newly updated annual data is not recalled after knowledge base synchronization. Cause: No trigger rule for annual incremental synchronization is configured, only a one-time full synchronization operation is performed.
How to confirm the configuration is correct
- Check the value range of
similarity, and confirm that all returned results contain spatial planning-related content by testing the retrieval of the keyword "permanent basic farmland protection". - Verify that after
usingReRankis enabled, the top-ranked results are all planning implementation evaluation or overall planning documents. - Confirm the binding status of
collectionIds, ensuring that the retrieval scope only includes knowledge base collections related to territorial spatial planning. - Test that after annual updated data synchronization, retrieving the corresponding keyword can recall new document fragments.
- Confirm that the current FastGPT version is v4.17.0 to ensure configuration items are compatible with the current system.
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.