This document is compatible with FastGPT v4.17.0.
What This Category of Data Looks Like
Typical data related to game operation activity configuration comes from operation backends, player segmentation systems, payment data platforms, and activity log systems. Commonly produced documents include activity plans, reward configuration tables, launch checklists, and activity review reports. Most document structures mix text, tables, and a small number of embedded explanatory images. Fields include reward values, activity time ranges, player participation thresholds, and compensation rules. Units involve conventional operation metrics such as yuan, person-times, and hours. Update frequencies are mostly weekly, or temporary adjustments before holiday activities.
Constraints on Document Parsing and Chunking
The mixed structure of activity documents requires parsing tools to support both plain text paragraphs and structured tables, and avoid splitting key reward values and their associated rules. The high-frequency update feature requires the parsing process to have reasonable timeout settings, to prevent failures of batch activity document parsing due to excessive runtime. Launch checklists with embedded images require OCR capabilities to extract text from images, otherwise key check item content will be lost. Differences in data formats from multiple sources require flexible adjustment of chunking strategies, to ensure that associated fields in the same document are not split into different chunks.
How to Set Configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the field density of activity documents, avoids splitting single-line reward rules in reward configuration tables |
chunkSplitter | Paragraph + table delimiter | Adapts to the mixed text and table structure of activity documents, retains complete associated data within tables |
customPdfParse | Enable OCR mode | Handles launch checklists with embedded images, extracts text content of check items from images |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Adapts to parsing time of batch activity documents, prevents parsing timeout failures for frequently updated activity documents |
UPLOAD_FILE_MAX_SIZE | 100 MB | Matches the common package size of operation activity documents, meets batch document upload requirements |
Three Common Misconfigurations
- After uploading a PDF of an activity launch checklist with embedded images, the parsing result does not include the check item text from the images. The cause is that the OCR mode of
customPdfParseis not enabled, and only plain text content is parsed by default. - Using the default fixed chunk parameters when configuring
chunkSettingModecauses long reward configuration table paragraphs to be over-split, losing the association between reward values and issuance conditions. - Unable to select table datasets for parsing in the knowledge base creation interface. The cause is that the table parsing switch is not enabled, or the current configuration does not adapt to the field format of activity-type tables.
How to Verify Correct Configuration
- Upload a PDF of an activity launch checklist that includes images and tables, check if the parsing result contains the text from the images and the complete fields of the table.
- View the parsed chunk list, confirm that each complete row of data in the reward configuration table is not split into multiple independent chunks.
- Upload a single activity document package under 100 MB, verify that the parsing task completes within the time set by
PARSE_FILE_TIMEOUT_SECONDS. - Switch
chunkSettingModeto custom mode, adjust thechunkSizeparameter and re-parse, confirm that the chunk length matches the set value.
The values given 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.