What the data for this category looks like
Daily documents for this category include creative scripts, material test sheets, delivery data reports, and iteration lists. Data sources are advertising platforms, material libraries, user interaction data, and conversion data. Most documents update daily or weekly. Document fields carry industry-specific identifiers. For example, material test sheets include material ID, material type, click-through rate, and conversion cost. Creative scripts include screen duration, copy version, and delivery targeting. Iteration lists include test variables and adjustment records.
What constraints these characteristics impose on the document parsing and chunking link
Diverse document types require different chunking logic. These range from structured material test sheets to unstructured creative scripts. Most data sources are reports exported from advertising platforms. These reports may have non-standard structures, such as misaligned headers and merged cells. Parsing accommodates these issues. High-update-frequency documents need fast parsing. Delays are avoided to prevent impacts on test iterations. Parsing accurately identifies specific business fields. Chunking retains field association relationships. This prevents loss of business context after splitting.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the content length of creative scripts and delivery reports, avoids including too much irrelevant information in a single chunk, while ensuring semantic completeness |
chunkSplitter | Split by paragraph + title | Matches the hierarchical structure of creative scripts and iteration lists, retains the relevance of business logic |
customPdfParse | Enable table-enhanced parsing | Compatible with non-standard formats such as merged cells and misaligned headers in material test sheets, accurately extracts structured fields |
UPLOAD_FILE_MAX_SIZE | 100 MB | Adapts to batch-imported delivery data reports and material library material packages, prevents large file upload failures |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Addresses parsing time for multi-page delivery data reports, prevents parsing failures due to timeout |
chunkSettingMode | Auto-adapt by document type | Differentiates between structured table documents and unstructured script documents, automatically adjusts chunking rules |
Three Common Mistakes
- Issue: After uploading a PDF creative script, the original file cannot be opened in the knowledge base results. Cause: The file retention configuration for
customPdfParseis not enabled, and only text content is extracted during parsing while the original file association is discarded. - Issue: After uploading a material test sheet, table fields are empty in the parsing results. Cause: Table-enhanced parsing is not enabled, merged cells cannot be identified, leading to loss of structured data.
- Issue: A request error is triggered when batch uploading multiple delivery data reports. Cause:
UPLOAD_FILE_MAX_SIZEis not adjusted to fit the size of batch packages, and requests are blocked after exceeding the limit.
How to Confirm Configurations Are Set Correctly
- Upload a single creative script PDF, check that the parsed text chunks are split by paragraph, with no excessively long or short chunks.
- Upload a material test sheet with merged cells, confirm that the parsing results include complete fields such as material ID and click-through rate.
- Upload a batch material package under 100 MB, confirm that the upload and parsing processes have no timeout errors.
- Check the background configuration panel of FastGPT v4.17.0, confirm that
PARSE_FILE_TIMEOUT_SECONDSis set to 600 seconds.
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.