What the data for this category looks like
Documents for new energy development business processing mainly originate from official process documents such as project approval, grid connection acceptance, and operation and maintenance filing under the energy and power sector of state-owned and state-controlled enterprises. These include project filing application forms, environmental impact assessment reports, grid connection acceptance forms, annual operation reports, and similar materials. Document updates are rolled out in batches according to project progress, with multiple batches of documents generated between project filing and grid connection completion. Most documents use formal multi-chapter layout, including long technical description paragraphs and engineering parameter tables. Fields include project code, installed capacity, grid connection time, approval document number, and similar items. Units mostly use industry standard units such as megawatt (MW) and ten thousand kilowatt-hours (kWh). Some documents include scanned attachments.
What constraints do these characteristics impose on the document parsing and chunking link?
The multi-chapter long document structure requires that chunking preserves contextual association between chapters and paragraphs, and avoids splitting key technical descriptions and parameter-bound content. Engineering fields with standard units require the parsing process to retain the corresponding relationship between numerical values and units, to prevent loss of parameter information. The batch update feature of documents requires that the chunking logic for the same batch of documents remains consistent, to ensure contextual coherence of business process guidance. Some documents are scanned PDFs, so OCR parsing capability must be supported to adapt to official approval documents in non-editable formats.
Configuration Settings
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
chunkSize | 1200–1500 characters | Adapts to long technical paragraphs and parameter blocks with units in new energy documents, avoids splitting key engineering information |
chunkSplitter | Recursive splitting by paragraph + title | Matches the chapter structure of new energy documents, preserves contextual association of approval processes |
customPdfParse | Enable Marker parsing mode | Supports parsing of scanned PDFs and complex-format project documents, adapts to the layout of official approval files |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to parsing time of long documents, prevents parsing interruption for large grid connection acceptance reports |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Accommodates full sets of large single project filing documents |
chunkSettingMode | Manual configuration | Adjusts chunking parameters for the special structure of new energy documents, avoids over-splitting from automatic mode |
Three Common Configuration Mistakes
- Issue: After enabling
customPdfParsein FastGPT v4.17.0, parsing some new energy PDF documents returns no valid content. Cause: The matching version of the Marker image was not used. Older versions of Marker cannot recognize engineering table formats in new energy documents. - Issue: In business process guidance conversations, the file parsing tool cannot be triggered, and the interface displays "No available parsing service found". Cause: The Marker parsing service container was not deployed correctly, or the
PARSE_FILE_TIMEOUT_SECONDSvalue was set too small, causing service startup timeout. - Issue: Key engineering parameters (such as installed capacity in MW) are split into different chunks in the parsed chunk results. Cause: The
chunkSizevalue was set too small, not adapting to the long parameter paragraph structure in the documents.
How to Verify Correct Configuration
- Upload a new energy project PDF document with more than 10 pages, and check if the parsing result retains the association between chapter titles and corresponding paragraphs.
- Access the parsing configuration page, confirm that the
customPdfParseswitch is set to enabled, and the associated Marker service version matches FastGPT v4.17.0. - Upload a test document with a single size of 800 MB, confirm that the upload and parsing process is not interrupted, and no error prompts are displayed.
- Review the chunk result list, confirm that the character count of each chunk falls within the 1200–1500 character range, and no key parameters have been split.
The values provided here are common starting points, and should be evaluated 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.