What the data for this category looks like
Forging and die forging business data originates from raw material quality certificates, heating furnace systems, press collection data, MES systems, non-destructive testing instruments, and customer forging drawings. Update frequency is per piece, per batch, or per furnace run. Daily documents include forging process specifications, cutting cards, heating furnace records, forging process cards, mold life records, heat treatment work orders, and ultrasonic testing reports. Fields include furnace batch number, forging dimensions, forging ratio, hardness values, testing time, and more. Units cover professional industrial parameters such as millimeters, megapascals, and hours.
What constraints these characteristics impose on the document parsing and chunking link
Mixed multi-source data formats require chunking rules to adapt to structured tables, process step text, and professional parameter blocks, to avoid splitting associated information across fields. Frequently updated batch data requires rapid parsing and chunking to avoid delays that reduce traceability efficiency. Structural differences across document types — such as tabular data in testing reports and linear steps in process specifications — require flexible chunking boundaries. Key associated fields such as furnace batch number and forging batch must be retained to ensure accurate data matching during subsequent quality traceability.
How to set the configurations
These configurations are formulated based on the rules of FastGPT v4.17.0:
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the field length of process steps and testing data in forging documents, avoiding splitting key parameters and associated context |
chunkSplitter | By paragraph + fixed character boundary | Adapts to mixed-format process cards and testing reports, retaining complete associated information for table rows and process steps |
PARSE_FILE_TIMEOUT_SECONDS | 120–180 seconds | Adapts to the parsing time required for large ultrasonic testing reports and multi-furnace batch records, avoiding termination of the parsing process due to timeout |
customPdfParse | Enabled | Adapts to heat treatment work orders and forging drawing documents with tables and annotations, retaining structured data and handwritten annotation content |
UPLOAD_FILE_MAX_SIZE | 500 MB | Adapts to the upload requirements for batch furnace records and multi-batch testing reports, supporting parsing scenarios for large documents |
Three common mistakes
- When uploading an ultrasonic testing report with more than 100 pages, FastGPT returns a 504 timeout error. The cause is that the
PARSE_FILE_TIMEOUT_SECONDSparameter is not adjusted, and the default timeout duration is insufficient to complete full parsing of the multi-page testing report. - When reading Excel cutting cards in a workflow, some furnace batch numbers and forging specification fields are empty. The cause is that the
chunkSplitterrule adapted to tabular documents is not configured, and the default chunking logic splits the associated data of table rows and columns. - The large language model fails to answer questions related to deformation steps based on the uploaded forging process specifications. The cause is that the
chunkSizeis set too small, splitting the continuous context of deformation steps and heating zones, preventing the model from associating complete process requirements.
How to confirm the configuration is correct
- Upload a single ultrasonic testing report with more than 100 pages, check whether FastGPT completes parsing within 180 seconds with no error logs.
- Upload an Excel cutting card file with tables, verify that the parsed data retains complete fields such as furnace batch number, forging specification, and cutting dimensions.
- Adjust
chunkSizeto 1000 characters, upload the forging process specification document, and check whether the chunking results include continuous deformation steps and heating zone content. - After enabling
customPdfParse, upload a heat treatment work order with annotations, verify that the parsed results include the heat treatment requirement content from the annotations.
The values provided are common starting points and should be measured against the reader's own samples.