What the data for this category looks like
The data for query result reports comes from structured data sources including internal state-owned enterprise and public sector system compilations, network segment record sheets, Feishu online spreadsheets, and other similar sources. Update cycles follow on-demand manual upload or scheduled synchronization. Single uploads are mostly multi-column Excel files with headers or exported files from online spreadsheets. Document structure is a two-dimensional table containing fields such as system number, effective date, clause content, network segment address, and remarks. Most fields use text or short text formats, and some fields have format constraints—for example, network segments must comply with IP segment specifications.
What constraints these characteristics impose on the document parsing and chunking step
The multi-column structure of structured tables means parsing tools must accurately identify the correspondence between headers and data rows, to avoid semantic breaks caused by cross-column misalignment. The short text nature of multiple fields means chunking processes must retain the complete field association of a single record, and cannot sever the binding relationship between system clauses and their numbers, or between network segment addresses and their remarks. The update cycle of on-demand upload or scheduled synchronization means chunking configurations must adapt to both single batch upload and incremental update scenarios. Fields with fixed formats such as IP segments and system document numbers require parsing to retain the original format, to avoid damaging field semantics. Some tables contain merged cells, so parsing logic must be compatible with non-standard table structures to prevent data loss.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the field length of single records in multi-column tables, retaining complete semantic associations |
chunkSplitter | Split by table rows | Matches the row-level data structure of structured tables, avoiding splitting single records across rows |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to the parsing time required for large-volume table files during batch uploads |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Meets the upload requirements for batch system tables and network segment record sheets in state-owned enterprises and public sectors |
chunkSettingMode | Manual configuration | Leverages the custom chunking capability of FastGPT v4.17.0 to define rules tailored to the special structure of structured tables |
Three common misconfigurations
- Phenomenon: After uploading an Excel or Feishu-exported spreadsheet file, only the first two columns of data are identified, and subsequent columns are lost. Cause: No chunking and splitting rules adapted for multi-column tables were configured. The default parsing logic only recognizes basic text columns and does not match multi-header structures.
- Phenomenon: After the knowledge base is chunked and stored, searching for network segment records fails to match the complete associated information of network segments and remarks. Cause: Chunking was performed based on character length, severing the binding relationship of multiple fields in a single table row.
- Phenomenon: Parsing tasks time out and fail when uploading large-volume table files in batches. Cause: The
PARSE_FILE_TIMEOUT_SECONDSparameter was not adjusted, and the default timeout duration is insufficient to complete full parsing of multi-column tables.
How to confirm the configuration is correct
- Upload a test structured table file, check that the number of parsed fields matches the original file, and confirm that all multi-column data has been fully identified.
- Run a retrieval test to verify that single records containing multiple fields can be fully recalled, confirming that chunking did not sever field associations.
- Upload a large-volume test file, confirm that the parsing task does not time out, and verify that the timeout parameter configuration matches the actual time required for uploading the file.
- After adjusting chunking parameters, compare chunking results across different configurations, confirming that the chunking rules adapt to the table structure.
The values provided here are common starting points and should be evaluated 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 2026-09-28.