What data for this category looks like
The core documents for game operation activity configuration include four types: activity plans, reward configuration tables, launch checklists, and activity post-mortems. Data mainly comes from operation backends, player segmentation systems, paid data platforms, and activity log systems. Document update frequencies are mostly weekly or before holiday events. The reward configuration table includes clear fields for item type, distribution quantity, and effective time period. Activity plans are divided into sections covering activity objectives, rules, and compensation plans. Some documents embed structured tables and player data snippets.
What constraints do these characteristics impose on the "document parsing and chunking" link
The mixed structured and unstructured nature of activity documents requires that parsing and chunking balance the integrity of table fields and the semantic coherence of paragraphs. The high-frequency update feature means the parsing link must support fast batch processing. Data from multiple sources may lead to differences in document formats, so adaptation to different layout structures is needed. Parsing errors in core fields such as reward values and activity time will directly cause business accidents, so chunking must avoid splicing content across activities, while retaining the contextual association of key fields.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Activity plans contain both paragraph descriptions and structured reward tables. This range balances field integrity and semantic coherence |
chunkSplitter | Split by paragraph + title | Activity documents are often divided into sections for activity objectives, rules, and reward modules. Splitting by title preserves contextual association |
customPdfParse | Enabled | Reward configuration tables are mostly structured tables in PDF format. Enabling this preserves table structures and prevents field loss |
PARSE_FILE_TIMEOUT_SECONDS | 120 seconds | Activity post-mortem documents often contain multi-page activity log data, requiring sufficient time for complete parsing |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Activity post-mortem documents may include multi-page logs. This upper limit covers the size requirements of most operation documents |
chunkSettingMode | Manual configuration | Activity document formats vary widely. Manual adjustment allows adaptation to different types of operation documents |
Three common errors
- Symptom: Files uploaded via the API have inconsistent chunking compared to files uploaded directly on the platform. Cause: The
chunkSplitterparameter was not specified during API upload, and the platform's default chunking rules were used, which differ from the manually uploaded configuration. - Symptom: No parsing results after configuring a custom parsing URL. Cause: The document formats supported by the custom URL were not confirmed, or the corresponding parsing service was not bound while
customPdfParseis enabled. - Symptom: The model cannot read files after configuring the document parsing node. Cause: A reasonable duration for
PARSE_FILE_TIMEOUT_SECONDSwas not set. Large activity log documents timed out during parsing, resulting in no valid chunks being generated.
How to confirm the configuration is correct
- Upload a single activity reward configuration table, compare the parsed chunked content with the original document's table fields, and confirm that all reward type and value fields are fully retained.
- Call the file parsing API, check whether the returned chunked data includes core fields such as activity effective time and compensation rules, with no missing or misaligned fields.
- Upload an activity post-mortem document under 10 MB, wait for FastGPT v4.17.0 to complete parsing, then check the number of chunks, which should fall within the range set by the
chunkSizeconfiguration. - After enabling
customPdfParse, upload a PDF-format launch checklist, and confirm that the parsed text retains the checklist item list structure from the original document.
The values provided here 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.