What this category of data looks like
Data for new energy development declaration materials and file review mainly comes from official process documents such as project approval, environmental assessment acceptance, and grid connection application. Update rhythm adjusts irregularly with project progress, and multiple versions are generated from project initiation to grid connection archiving. Documents include fixed-format approval forms, free-form technical descriptions, financial calculation Excel files, and scanned PDF drawings. Fields include project installed capacity (unit: MW), total investment (unit: ten thousand yuan), and land area (unit: hectare). Some documents include structured equipment lists and progress ledgers.
What constraints do these characteristics impose on the "document parsing and chunking" link
Scanned PDF files require OCR recognition, which adds extra requirements for the parsing service's OCR capabilities. Structured Excel lists contain numerical fields with units. Chunking must retain the association between fields and units to avoid data separation. Long documents such as environmental assessment reports that can be dozens of pages long will increase parsing time, so timeout configurations need to be adjusted. Different types of declaration materials such as project approval applications and grid connection acceptance have different chapter structures. Flexible chunking rules must be adapted to ensure complete contextual logic.
How to set the configurations
These configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
chunkSize | 1000–1500 characters | New energy declaration materials contain long technical descriptions and financial data. Excessive length will cause contextual separation, while insufficient length will lose field associations |
chunkSplitter | Split by paragraph + title | Declaration materials have clear chapter titles and paragraph structures. Splitting this way preserves the logical hierarchy of the document |
customPdfParse | Enable OCR mode | Some files are scanned PDF documents, and OCR is required to extract text content |
PARSE_FILE_TIMEOUT_SECONDS | 120–180 seconds | A single declaration material PDF may exceed 50 pages, resulting in long parsing time. This avoids timeout errors |
UPLOAD_FILE_MAX_SIZE | 500 MB | Some project declaration materials include multiple high-definition drawings and attachments, so the upload limit needs to be expanded |
chunkSettingMode | Manual configuration | Chunking parameters must be adjusted for different types of declaration materials such as environmental assessment and grid connection acceptance to adapt to specific document structures |
Three common mistakes
- Some uploaded files are not parsed in simple mode. The cause is that
chunkSettingModeis not configured to automatically trigger parsing, so the large language model does not recognize the document types that need parsing. - The interface shows that the file was parsed successfully, but the model does not reference the document content. The cause is that
customPdfParseis not enabled for scanned PDF processing, or the chunking parameters are set too short, resulting in the loss of core field associations. - The parsing service call in the workflow returns a timeout error. The cause is that
PARSE_FILE_TIMEOUT_SECONDSis set too low, which cannot cover the parsing time of long documents such as environmental assessment reports with more than 50 pages.
How to confirm the configurations are correct
- Upload a scanned environmental assessment PDF, check whether the parsed text includes the drawing description text extracted by OCR, to confirm that the
customPdfParseconfiguration takes effect. - Upload an Excel list containing the installed capacity field, check whether the chunked content retains the association between the value and the unit "XX MW".
- Upload a 100-page project declaration PDF, check whether there is no timeout error in the parsing log, to confirm that the value of
PARSE_FILE_TIMEOUT_SECONDSis reasonable. - Call the document parsing node in the workflow, verify that the returned chunked content is divided by chapter titles, to confirm that the
chunkSplitterconfiguration is correct.
The values listed here are common starting points, and should be evaluated against the user'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.