This content is configured based on FastGPT v4.17.0.
What the Data for This Category Looks Like
Daily documents for this category include dyeing formulas, chemical weighing records, dye vat curves, sample confirmation reports, color difference reports, setting process sheets, sewage ledgers, and more. Data sources cover customer color samples, formula libraries, dye chemical batches, dye vat control equipment, colorimeters, setting machines, and environmental monitoring systems.
Update frequency is per vat, per batch, or per shift. Document structures include structured tables and text descriptions. Fields include specific units, such as degrees Celsius for dye vat temperature, kilograms or liters for dye chemical dosage. Batch numbers use a combination of letters and numbers.
What Constraints These Characteristics Bring to the Document Parsing and Chunking Link
Document types in this category are diverse, with significant structural differences. Some are plain text process descriptions, while others contain tabular data with multiple rows and columns. Parsing must balance extraction accuracy across different formats.
High-frequency updates per vat or batch require chunking logic to retain the integrity of single-batch data. It must avoid splicing parameters across vat runs or batches. Specific units attached to fields must be accurately identified and preserved. Otherwise, parameter accuracy in subsequent quotations and bid documents will be compromised.
Long dye vat curve documents may require extended parsing time. A reasonable timeout threshold must be configured.
How to Set the Configuration
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to documents with both text and tables such as dyeing formulas and dye vat curves. Avoids splitting that disrupts the correlation of single-vat process parameters. |
chunkSplitter | Split by paragraph + header | Matches the structured characteristics of daily documents. Accurately splits tabular documents with headers such as dye vat curves and chemical weighing records. |
customPdfParse | Enable table recognition mode | Targets documents with structured tables such as chemical weighing records and setting process sheets. Fully preserves field and unit information within cells. |
PARSE_FILE_TIMEOUT_SECONDS | 120 seconds | Covers the parsing time of long dye vat curve documents. Prevents parsing failure due to insufficient default duration. |
chunkSettingMode | Manual configuration | Targets document types with large differences such as formula libraries and sewage ledgers. Flexibly adjusts chunking parameters to adapt to different formats. |
Three Common Mistakes
- Phenomenon: A large number of empty fields appear in the parsing result after uploading a dyeing formula PDF. Cause: The table recognition mode of
customPdfParseis not enabled. This causes the parameter tables in the formula to not be extracted correctly. - Phenomenon: A timeout error is triggered when parsing a long dye vat curve document. Cause: The
PARSE_FILE_TIMEOUT_SECONDSparameter is not adjusted. The default parsing duration is insufficient to process complete curve data. - Phenomenon: Chunked content contains spliced dye vat parameters from different vat runs. Cause:
chunkSplitteris not configured to split by batch number. This causes cross-vat-run data to be merged into the same chunk.
How to Confirm the Configuration Is Correct
- Upload a standard dyeing formula PDF. Check if the parsing result includes complete fields such as dye chemical names, dosages, and units.
- Upload a long dye vat curve document. Confirm that no timeout error is triggered during parsing, and that the content fully covers the time and temperature data of the curve.
- Randomly select multiple documents of different types. Check that chunked content does not have parameter splicing across batches or documents.
- Adjust the
chunkSizeparameter. Verify that the length of each chunked text falls within the preset range.
The values provided here are common starting points. They should be evaluated against the reader's own samples.