What the data for this category looks like
Daily documents for this category include art requirement sheets, asset specification tables, outsourcing feedback records, and acceptance documents. Data sources include art asset libraries, project management systems, engine resource directories, and outsourcing delivery packages. Most updates occur daily or weekly. Most documents have structured fields such as resource ID, polygon count, resolution, and version number. Common field units are pixels, triangle face count, and file format. Some documents include combined text and image reference descriptions. Some outsourcing delivery packages bundle multiple associated files.
Constraints on Document Parsing and Chunking
Documents with a high proportion of structured tables require chunking to retain complete field associations. Do not split across specification items or requirement entries. Documents with combined text and images require retaining the correspondence between original graphics and text, to prevent losing visual reference information. For batch delivery scenarios with multiple files, parsing configurations must adapt to maximum single file size and batch upload limits. For frequently updated documents, chunking logic must match fast retrieval requirements. Avoid using chunking that is too coarse or too fine, which breaks associated information. Outsourcing delivery packages that bundle multiple files require parsing configurations that support batch file processing to avoid parsing failures due to single-file limits.
How to Configure Settings
These configurations are based on FastGPT v4.17.0. The details are as follows:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the field density of asset specification tables, and avoids splitting across specification items |
chunkSplitter | Split by heading + paragraph | Matches the hierarchical structure of art requirement sheets, and retains the association between requirements and reference images |
customPdfParse | Enabled | Preserves embedded tables and vector graphics in outsourcing delivery package PDFs, and avoids forced general OCR that discards visual information |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Adapts to the maximum batch file size for outsourcing delivery packages |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Addresses parsing time requirements for large asset specification documents |
chunkSettingMode | Manual configuration | Adjusts chunking rules for the special structure of art asset documents |
Three Common Misconfigurations
- Symptom: Importing a PPT or DOC document and enabling PDF enhancement results in no parsing effect. The task status shows parsing completed, but no valid text content is present. Cause: PDF enhancement only supports PDF format documents. Corresponding parsing rules for Office files are not enabled, leading to incorrect document recognition.
- Symptom: Excel parsing results do not include sheet names, and only the default sheet content is displayed. Cause: The default chunking configuration does not extract sheet tab information, and multi-sheet recognition logic is not enabled. Sheet names are not included in the parsing scope.
- Symptom: Parsing a combined text and image manual results in only original graphics being lost, and only text extracted via OCR. Cause:
customPdfParseis not enabled, so general OCR mode is used instead of preserving original vector graphics.
How to Verify Correct Configuration
- Upload a single art requirement sheet PDF with heading hierarchies. Check if parsing results retain the association between headings and corresponding paragraphs, and no splitting across headings occurs.
- Upload an Excel asset specification table with multiple sheets. Confirm that parsing results include each sheet's name and corresponding data, with no missing tab content.
- Upload an outsourcing delivery package PDF with embedded vector graphics. Check if original graphics are retained instead of only extracting text via OCR.
- Upload an outsourcing delivery package within 100 MB. Confirm that the parsing task completes within 600 seconds with no timeout errors.
The configuration values listed here are common starting points, and should be adjusted based on 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 2026-10-19.