What the data for this category looks like
Daily documents for complete home appliance manufacturing include bill of materials (BOM), assembly work instructions, refrigerant charging records, electrical safety test reports, performance sampling inspection reports, and repair records. Data sources cover ERP/MES systems, component batch ledgers, tightening and charging equipment, safety compliance testers, performance laboratory equipment, and after-sales fault databases.
Update frequency aligns with per-unit product, per-shift operation batch, or sampling inspection schedule. Each document includes fields such as model identifier, parameter values, test results, and operator information. Some documents have unit annotations, such as kilograms for refrigerant charging volume, and volts for safety compliance test values.
What constraints these characteristics impose on document parsing and chunking
The mixed-model assembly business requires document parsing to accurately distinguish parameters for different models, to avoid confusing model identifiers during chunking. High-frequency updated per-shift operation records and after-sales repair data must be parsed quickly, otherwise production process synchronization will be disrupted.
Safety compliance test reports and performance sampling inspection reports contain a large number of numerical fields with units. Chunking must retain field association, otherwise retrieval cannot link parameters and test results.
Batch documents have large individual file sizes. If chunking parameters are not set reasonably, parsing timeout or indexing exceptions may occur. Additionally, the number of chunks for after-sales fault data must be controlled within a reasonable range to avoid exceeding the system indexing limit.
How to configure the settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapt to the length of single work steps and test data in home appliance process documents, avoid splitting critical parameters and associated information |
chunkSplitter | Split by title + paragraph | Match the hierarchical structure of BOM and assembly work instructions, retain parameter grouping logic for different product models |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Cover parsing duration for per-shift tightening records and batch repair reports, avoid parsing interruptions for high-frequency updated documents |
customPdfParse | Enabled | Retain tables, unit fields and boolean markers in safety compliance test reports, avoid loss of critical information during parsing |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Support batch upload of complete batches of performance sampling inspection reports, adapt to batch document processing requirements |
chunkSettingMode | Manual configuration | Adjust chunking parameters for different document types such as work instructions, test reports and repair records |
Three common configuration errors
- Symptom: Parsed boolean fields become null values in conditional judgment components. Cause: When
customPdfParseis not enabled, qualified/unqualified boolean markers in documents are incorrectly identified as plain text, and the original data type is not retained. - Symptom: The online environment throws the error
Cannot redefine property: toString. Cause: When uploading large files exceeding theUPLOAD_FILE_MAX_SIZElimit, the FastGPT v4.17.0 system attempts to redefine built-in object properties, resulting in an exception. - Symptom: Search results are missing after the number of chunks exceeds 3000. Cause: The
chunkSizeparameter is not adjusted, resulting in too many chunks split from a single document, exceeding the system default indexing limit. Adjust chunking parameters or split files by document type.
How to confirm the configuration is correct
- Upload a single safety compliance test report, check if parsed fields retain unit information such as voltage and current.
- Upload batch repair record files, confirm that parsing duration does not exceed the set
PARSE_FILE_TIMEOUT_SECONDSvalue. - Split a document that would generate more than 3000 chunks, verify that relevant content can be successfully retrieved during search.
- Test the conditional judgment logic for boolean fields, confirm that the field type is not incorrectly converted.
The values provided are common starting points and should be measured against the reader's own samples.