What data looks like for this category
Tire, seal, hose and rubber product enterprises are typical entities in this category. Daily documents include formula sheets, weighing records, mixing curves, rubber material rapid test reports, vulcanization process cards, mold records, finished product performance reports, and other materials. Data sources cover formula systems, raw material batch ledgers, internal mixers, Mooney viscometers, rheometers, vulcanizers, and laboratory testing equipment. Data updates occur per raw material batch, per mixed rubber batch, and per vulcanization operation. Document fields include formula ratio, feeding sequence, mixing temperature, vulcanization duration, and other items, with exclusive units such as degrees Celsius, seconds, and kilograms.
What constraints these characteristics impose on document parsing and chunking
Documents in this category have diverse types and frequent updates. They include editable formula sheet spreadsheets, scanned vulcanization process cards, and documents with time-series data such as mixing curves. The mixed format of multi-source data requires parsing components to support recognition of both structured tables and unstructured text. Frequently updated operation data means the parsing process must adapt to the maximum volume limit for single documents, to avoid parsing interruptions caused by overly large documents. Exclusive fields and units require parsing results to retain original data format and unit information, to prevent parameter ambiguity in bid documents. In addition, key information such as feeding sequence and vulcanization parameters must not be split across chunks, otherwise it will affect the accuracy of subsequent quotation and process matching.
This configuration is formulated based on FastGPT v4.17.0.
How to set the configuration
| Configuration Item | Recommended Value | Rationale for this Setting |
|---|---|---|
chunkSize | 800–1200 characters | Adapt to the field density of mixed documents such as mixing curves and formula sheets, avoid splitting key process parameters |
customPdfParse | Enable OCR mode | Process non-editable PDF documents such as scanned vulcanization process cards and mold records |
PARSE_FILE_TIMEOUT_SECONDS | 300 seconds | Adapt to the parsing duration of large-volume documents such as single mixing curves |
chunkSplitter | Split by paragraph + table | Retain the table structure of formula sheets and weighing records, avoid truncating feeding sequence information across chunks |
UPLOAD_FILE_MAX_SIZE | 500 MB | Accommodate archived documents with multiple batches of mixed rubber data |
chunkSettingMode | Manual configuration | Adjust chunking rules for different document types to meet the structured requirements of bid documents |
Three common mistakes
- Parsing results show empty fields after uploading a scanned vulcanization process card. The cause is that OCR recognition mode for
customPdfParseis not enabled, making it impossible to extract process parameters from the image. - A timeout error occurs when parsing a large-volume mixing curve document. The cause is that the set value of
PARSE_FILE_TIMEOUT_SECONDSis less than the actual parsing time of the document. - After uploading a document containing a complete formula table, the chunking result splits the table into multiple scattered pieces. The cause is that
chunkSizeis set too small, truncating the complete row and column structure of the table.
How to confirm the configuration is correct
- Upload a scanned vulcanization process card, check if the parsing result includes core fields such as mold number and vulcanization temperature.
- Upload a single mixing curve document, check if the parsing duration is within the preset range of
PARSE_FILE_TIMEOUT_SECONDS. - Upload a document containing a complete formula table, check if the chunking result retains the complete row and column structure of the table.
- Adjust
chunkSize, then upload a weighing record document, confirm that feeding sequence information is not split across chunks.
The values provided are common starting points and should be measured against the reader's own samples.