What Data Looks Like for This Category
Typical subjects include professional heat treatment plants and heat treatment workshops in automotive and bearing parts factories. Daily documents cover heat treatment process cards, charging diagrams, furnace temperature curves, hardness and metallographic reports, equipment inspection forms, and abnormal disposal records. Data sources include material grades, customer technical requirements, temperature control instruments, furnace temperature uniformity testing equipment, hardness testers and metallographic equipment, and MES systems. Data is updated per heat batch or per production batch. Documents contain fields such as heating, holding, and cooling schedules, charging positions, furnace temperature curves, and hardness and metallographic test results. Some documents include quantitative units such as temperature, time, and hardness values.
What Constraints These Characteristics Impose on the Document Parsing and Chunking Stage
Heat treatment scenario documents have diverse types: text-based process cards, chart-based furnace temperature curves and metallographic images, and table-based inspection forms. Different document formats have varying adaptability requirements for parsing tools. The per-heat-batch data update frequency requires the parsing process to support real-time and batch processing, to avoid delaying response speeds for inspection and maintenance. Fields within documents are strongly linked to business operations. When chunking, context logic must be preserved. For example, charging positions and their corresponding furnace temperature curves cannot be split into different chunks, as this will reduce the accuracy of subsequent maintenance Q&A. Additionally, data from multiple sources requires a unified parsing format to ensure consistent processing of documents exported from different equipment.
How to Configure Settings
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Heat treatment documents contain continuous process parameters and equipment inspection items. This range balances context completeness and chunk granularity |
chunkSplitter | Separate by paragraph + heading | Heat treatment documents have clear process steps and equipment inspection item headings. Following this chunking rule preserves business logic associations |
customPdfParse | Enabled | Adapts to PDF documents in heat treatment scenarios such as furnace temperature curves and metallographic reports with charts, and extracts non-text data |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Parsing large furnace temperature curve PDFs or multi-document collections takes significant time. This range prevents parsing failures due to timeouts |
UPLOAD_FILE_MAX_SIZE | 500 MB | Meets the upload requirements for multi-document collections per heat batch, and adapts to batch processing scenarios |
chunkSettingMode | Manual configuration | Heat treatment documents have diverse structures. Manual adjustment allows adaptation to different file types such as process cards, inspection forms, and curve reports |
Three Common Configuration Mistakes
- After uploading image files of charging diagrams or image fragments from furnace temperature curve PDFs, the parsing result lacks corresponding equipment position or temperature data. Cause: The
customPdfParseconfiguration is not enabled. The default parsing logic cannot extract non-text data from images. - After deploying FastGPT locally and enabling enhanced PDF parsing, the local MinerU interface cannot be called. Cause: The MinerU API address and port are not correctly configured in the knowledge base settings, or the port is not open to external access.
- After parsing table data from heat treatment inspection forms, multi-vector support cannot be enabled. Cause: The multi-vector support option for table data is not checked in the knowledge base settings, or the chunk configuration does not adapt to the row and column structure of tables.
How to Verify Proper Configuration
- Upload a single heat treatment process card PDF, and check if the parsing result retains the complete paragraph content of the heating, holding, and cooling schedules.
- Upload a PDF file containing a furnace temperature curve, and check if the parsing result includes the curve's temperature data and timeline information.
- Enable enhanced PDF parsing in the knowledge base settings, and use the curl tool to test the MinerU interface, confirming that the return status code is 200.
- Adjust the
chunkSizeparameter, upload a long document, and check if the character count of each chunk falls within the set range.
The values provided are common starting points and should be evaluated against the reader's own samples.