What this category of data looks like
This category of data primarily comes from the national land and space planning unified map, project approval and supervision platform, current land use database, and ecological protection redline data, with irregular update frequency. Daily processed documents include land pre-examination and site selection opinions, construction land planning permits, construction engineering planning permits, redline drawings, and planning conditions. Fields cover project site selection scope, land scale, planning control requirements, pre-approval conditions, and more. Units mostly use standard land and space measurement units such as square meters and hectares. The data structure varies slightly based on project type.
What constraints these characteristics impose on the knowledge base retrieval and recall link
Dispersed data sources and irregular update cycles require knowledge base retrieval to support associated recall of multi-source data. It prevents missing the latest ecological protection redline or planning control adjustment information. Diverse document types and specialized fields require setting exclusive retrieval matching rules for different documents. It ensures text content from drawing-based documents such as redline drawings and planning conditions can be accurately identified. Strongly correlated data linked to projects requires prioritizing recall of entries directly related to the current project's site selection and land scale during retrieval. It prevents irrelevant data from interfering with the accuracy of ledger summary and report generation.
How to set the configuration
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Semantic similarity threshold for matching professional land and space planning documents. It avoids recalling irrelevant non-land-related data |
limit | Top 8–12 entries | Covers core data such as pre-approval conditions and redline requirements required for a single project approval. It prevents excessive context from affecting generation efficiency |
searchMode | Vector + keyword hybrid retrieval | Balances semantic matching for professional terms and precise keyword recall for documents such as redline drawings and planning conditions |
usingReRank | Enabled | Performs secondary sorting on recalled entries. It prioritizes screening results directly related to the current project's site selection and land scale |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands retrieval keywords. It covers pre-approval related information such as development and reform, housing and urban-rural development that may be involved in project approval |
collectionIds | Assign exclusive collection IDs by business domain | Isolates land pre-examination and planning permit data from other natural resource category data. It prevents data confusion |
Three common errors
- Phenomenon: An API call returns the error "No knowledge base specified" with status code 400. Cause:
collectionIdsis not properly configured in the request parameters, or the passed collection ID does not belong to the exclusive knowledge base of the current business domain. - Phenomenon: Two locally deployed Docker instances cannot communicate normally. Timeouts occur when calling external APIs to connect to the knowledge base. Cause: Cross-container network access permissions are not properly configured. The second instance cannot access the exclusive knowledge base service. Adjust Docker network settings or specify the correct knowledge base service address.
- Phenomenon: Imported web links that require login cannot be retrieved. Retrieval results are empty. Cause: Login state restrictions are not handled. FastGPT cannot normally crawl web content that requires login. Configure open link access permissions or a proxy in advance.
How to confirm the configuration is complete
- Manually upload a sample document of land pre-examination and site selection opinions. Perform a retrieval test, and confirm that relevant content from the document can be recalled.
- View the knowledge base retrieval logs. Confirm that the
similarityscore of the returned results falls within the set 0.75–0.85 range. - Check the API request parameters. Confirm that
collectionIdshas specified the exclusive collection ID of the current business domain, and no cross-category data is recalled. - Simulate a ledger summary scenario. Input a query containing the project's site selection and land scale. Confirm that the recalled results include core document content such as redline drawings and planning conditions.
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.