Data for This Category
The core data for this category comes from resin batch property sheets, injection molding machine parameters, mold temperature controllers, MES systems, and dimensional and appearance inspection results. Daily documents include injection molding process cards, raw material drying records, mold histories, first-piece inspection forms, patrol inspection logs, and material change and machine cleaning records.
Document structures vary widely by type:
- Process cards contain pressure and temperature parameters for the filling, packing, and cooling stages
- Drying records note drying temperature and duration
- Mold histories track mold change counts and maintenance cycles
Data update frequency aligns with production pace. New documents are created for each raw material batch, each mold or material change, and after each production shift. Most fields include clear units.
Constraints on Document Parsing and Chunking
Frequently updated, multi-type documents require parsing and chunking workflows that adapt to different structures, and avoid splitting critical process parameters.
Injection molding documents include many numerical fields with attached units. Chunking must preserve the link between parameters and their units, otherwise cross-review cannot match raw material properties to process requirements.
Large mold history files and multi-batch drying records have large file sizes and long parsing times. These workflows must support longer timeout thresholds.
Structured table documents require full field extraction, to avoid losing associated data via default parsing logic. Otherwise, risk points such as mixed materials or mold damage will be missed.
Configuration Settings
The following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the single-section parameter length of injection molding process cards, and avoids splitting core process information such as filling, packing, and cooling windows |
chunkSplitter | Split by paragraph + semantic split | Adapts to structural differences across multiple document types, and preserves complete business logic for drying records and mold histories |
customPdfParse | Enable table recognition mode | Adapts to structured table data in patrol inspection logs and material change and machine cleaning records, and avoids losing numerical values and units when extracting fields |
PARSE_FILE_TIMEOUT_SECONDS | 180 seconds | Adapts to the parsing time of large mold histories or multi-batch drying records, and avoids triggering timeout errors |
UPLOAD_FILE_MAX_SIZE | 500 MB | Adapts to bulk uploads of resin property sheet collections, and meets import requirements for multi-batch production data |
chunkSettingMode | Manual configuration | Adjusts chunk boundaries for specific fields in injection molding documents, and ensures critical parameters are not split |
Three Common Configuration Errors
- Scenario: After uploading multi-batch injection molding process cards or large mold history files, the interface displays a timeout error with status code 504. Cause: The
PARSE_FILE_TIMEOUT_SECONDSvalue is not adjusted, and the default duration is insufficient for parsing large files. - Scenario: After uploading structured patrol inspection logs or material change and machine cleaning records, the parsed results show empty detection values and corresponding unit fields. Cause: The table recognition mode of
customPdfParseis not enabled, and the default parsing logic cannot extract structured data from tables. - Scenario: After exporting a knowledge base and importing it to a new server, critical configurations such as return material ratio and raw material drying parameters are missing. Cause: The link between parameters and their context is not preserved during chunking, and paragraphs containing complete configurations are split.
How to Verify Configurations Are Set Correctly
- A single injection molding process card is uploaded, and parsed chunks are checked to confirm they retain complete filling, packing, and cooling window parameters and their corresponding units.
- A structured patrol inspection log is uploaded, and parsed results are checked to confirm all detected values and corresponding fields are included.
- A collection of mold histories under 500 MB is uploaded, and the parsing process is checked to confirm no timeout errors are triggered.
- A current knowledge base configuration is exported, imported to a test environment, and critical parameters such as return material ratio are checked to confirm they are not lost.
The values provided here are common starting points, and should be evaluated against samples relevant to the specific use case.