What data in this category looks like
This category of data comes from the national land and space planning unified map, project approval and supervision platform, land use status database, and ecological protection red line data. Update frequency is irregular. Daily documents include land pre-examination and site selection opinions, construction land planning permits, construction engineering planning permits, red line drawings, and planning conditions. Content covers project site selection control, land use scale verification, red line boundary demarcation, and similar items. Fields include project number, land scope coordinates, planning indicators, approval status, and similar items. There are no unified fixed units, and they vary based on document type.
What constraints these characteristics impose on the vector models and indexes link
This category of data has multiple sources and significant structural differences, including text approval documents and coordinate-based vector document descriptions. Chunking logic must be adjusted for different document types. The irregular update frequency requires indexes to support incremental updates to reduce reconstruction costs. Data involves planning control rules and project boundary details, so retrieval must accurately match control requirements to avoid retrieving irrelevant content. The volume of single-batch documents may be large, so index construction must adapt to batch processing logic to prevent server overload from disrupting business processes.
How to set configurations
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Adapts to the text length of approval documents and red line drawing descriptions in this category, avoiding overly fragmented chunks that lose context or exceed model limits |
| Overlap Length | 100–150 characters | Retains contextual association between chunks, ensuring cross-chunk information such as site selection and control rules is not fragmented |
| Similarity Threshold | 0.75–0.85 | Matches the precise retrieval requirements for planning control rules in this category, filtering low-relevance results |
defaultToken | 4096 | Adapts to the context window of mainstream large models for FastGPT v4.17.0, accommodating core content required for a single round of consultation |
maxToken | 8192 | Supports integrating content from multiple related documents, meeting full-process consultation needs for complex projects |
| Single Search Limit | Top 8 results | Balances retrieval recall rate and response speed, covering core control basis required for project approval |
Three Common Mistakes
- Phenomenon: After server restart, the knowledge base status shows not ready, there is no automatic indexing progress, and retrieval cannot be triggered. Cause: Incremental indexing trigger rules are not configured, or batch indexing tasks were interrupted due to server crash and no breakpoint continuation configuration was retained.
- Phenomenon: After updating the FastGPT version, the original indexed knowledge base search returns no results. Cause: The vector model of the new version does not match the vector dimension of the old version index, and the corresponding index was not regenerated.
- Phenomenon: Retrieval results include a large amount of land use data unrelated to the target project. Cause: The chunk length is set too small, splitting control rules and irrelevant plot information, leading to retrieval of unrelated content.
How to Confirm the Configuration Is Properly Set
- Upload a single typical document (such as a construction land planning permit), check the chunk preview result, and confirm that the chunk length and overlap length meet expectations.
- Enter a typical consultation question (such as "Does a certain project's site selection comply with ecological protection red line requirements"), check the relevance of the retrieval results, and adjust the similarity threshold to the required range.
- Batch upload more than 10 documents of the same type, observe server load and indexing progress, and confirm there is no overload or interruption.
- View the knowledge base's indexing status logs, and confirm that the automatic incremental indexing trigger rule is normally enabled.
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.