What this category of data looks like
The data for this category mainly comes from the national land spatial basic information platform, natural resource survey and monitoring data, population and economic statistics data, and remote sensing images. It has an annual update frequency. Commonly processed documents include overall land spatial plans, detailed plans, special plans, planning implementation evaluation reports, and public opinion summaries. Document structures contain clear chapter hierarchies, spatial indicator fields, and multi-format attachments. Some documents include image screenshots with coordinate annotations or tabular data.
What constraints do these characteristics impose on the document parsing and chunking step
Batch documents updated annually create a constraint of large per-batch parsing volume. This requires adaptation to multi-source heterogeneous file formats and content structures. Different types of planning documents have distinct field requirements. The association between spatial indicators and text descriptions must be preserved. Otherwise, retrieval accuracy for subsequent review steps will be affected. Public opinion summaries are mostly scattered short texts. Excessive chunking that causes semantic breaks must be avoided. Image attachments with coordinate annotations require additional parsing adaptation. Standard text parsing cannot recognize spatial coordinate information.
How to configure settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | This category of documents includes spatial indicators and long text descriptions. This range balances semantic completeness and retrieval granularity |
chunkSplitter | Split by paragraph + heading | Planning documents have clear chapter hierarchies. Preserving the association between headings and their corresponding paragraphs improves subsequent retrieval accuracy |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Single regional planning documents may include multiple pages of remote sensing image attachments, leading to long parsing times |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Regional-level planning documents may include large-volume remote sensing image data |
customPdfParse | Enabled | Adapts to spatial tables and coordinate annotation formats in planning documents |
chunkSettingMode | Manual configuration | Adjust chunking rules for different document types to adapt to the distinct structures of overall plans and public opinion summaries |
Three common mistakes
- Uploading an image with spatial annotations results in empty parsing results. Cause: The
customPdfParseconfiguration is not enabled. The default parsing logic cannot recognize coordinate annotation content in images. - Passing an image URL when calling the document parsing API fails to complete parsing. Cause: No accessible image URL whitelist is configured, or the link does not carry correct access permission identifiers.
- Calling the parsing interface after locally deploying MinerU results in a timeout. Cause: The MinerU port is not mapped to the FastGPT container network, or the parsing address configured in FastGPT does not match the actual running port of MinerU.
How to confirm the configuration is correct
- Upload a typical detailed planning PDF document, and check if the parsed chunks retain the association between chapter headings and their corresponding content.
- Upload a remote sensing image screenshot with land spatial coordinate annotations, and confirm that the parsing result includes the annotation text and coordinate information.
- Call the document parsing API of FastGPT v4.17.0, pass the locally deployed MinerU service address, and check that the returned result has a status code of 200 and includes complete parsed content.
- View the knowledge base storage logs, and confirm that the chunked documents have been correctly written to the specified storage medium.
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.