What Data for This Category Looks Like
Normally, daily data for level content creation teams comes from level editors, art resource libraries, player test data, and defect tracking systems. Daily documents include level design drafts, resource requirement sheets, test feedback forms, and version submission records. Level design drafts contain fields such as level ID, path node coordinates, and interaction rules. Resource requirement sheets list resource file names and storage usage units. Test feedback forms record defect numbers and reproduction steps. Version submission records are updated weekly or per version.
What Constraints These Characteristics Impose on Document Parsing and Chunking
Documents for level content creation include both structured fields and free-form text descriptions. The length of documents varies widely across types. Short test feedback forms are only a few dozen characters long. Long version submission records can reach several thousand characters. Most data comes from standardized formats exported by editors. A small number of documents have custom headers and footers. Document update frequency is high. Some test feedback forms are added in real time during player testing. Chunking must balance accurate extraction of structured data and contextual completeness of free-form text. If key coordinate or resource information is lost during chunking, it will directly affect subsequent level debugging and launch progress.
How to Configure Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Aligns with the paragraph length of level design drafts. It can cover complete path rule descriptions without causing contextual association failure due to overly large chunks |
chunkSplitter | Split by title + paragraph | Level documents mostly divide different level modules by chapter titles. Splitting by title preserves complete context for each module |
customPdfParse | Enabled | Filters fixed headers and footers in PDFs exported from level editors to avoid including irrelevant content in chunks |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports large level project packages containing multiple design drafts and resource lists |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Allows sufficient parsing time for large volume level documents to avoid parsing failure due to timeout |
chunkSettingMode | Manual mode | Supports adjusting chunking parameters for different document types such as design drafts and test forms to adapt to diverse business scenarios |
Three Common Misconfigurations
- Symptom: After uploading level documents, the chunk ID generated during parsing cannot be copied. Cause: The default document parsing module in FastGPT v4.17.0 does not enable chunk ID copy interaction, and the corresponding configuration must be manually enabled.
- Symptom: After custom chunking of documents, duplicate content is automatically deleted, resulting in disordered custom index order. Cause: The knowledge base has global deduplication enabled by default. Failing to turn off this switch will cause custom chunks to be filtered out.
- Symptom: When parsing test feedback forms, structured data cannot be returned in table form. Cause: Custom parsing rules adapted to the table format of test feedback forms are not enabled, resulting in structured fields being split into plain text.
How to Verify Correct Configuration
- Upload a single level design draft, view the parsed chunk list, and confirm that each chunk contains complete level path or resource descriptions.
- Enter the parsing settings page of the knowledge base, confirm that
chunkSettingModeis set to manual mode andcustomPdfParseis enabled. - Try to copy the ID of any chunk, confirm that the copy function works properly.
- Upload a test feedback form, check that the parsed result retains the structured display effect of the original table.
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 2026-10-19.