What this category's data 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 test reports, and repair records. Data sources cover ERP/MES systems, component batch ledgers, tightening and charging equipment, safety compliance testers, performance laboratories, and after-sales fault databases. Update cadence follows these rules: corresponding records are updated when a single device is produced, assembly work files are updated per shift, and test reports are generated per sampling plan. Document fields include material code, station number, test value, batch number, and other items. Some documents have clear units: refrigerant charging volume uses kilograms, and test voltage uses volts.
How these characteristics impose constraints on document parsing and chunking
The need to parse mixed document types requires chunking logic to adapt to different formats such as structured BOMs, free-text work instructions, and standardized test reports. The high-frequency, small-batch update cadence requires the parsing process to have stable response speeds to avoid disrupting production takt times. The mixed-model assembly business scenario requires chunking to accurately associate with the corresponding product model to prevent mixing of document content from different models. Differences in document formats from multiple device sources require parsing tools to support non-standard exported files and ensure the integrity of key fields. The sensitivity of safety compliance and performance data requires the chunking process to retain the association of original data, avoiding situations where the corresponding test batch cannot be linked after splitting.
How to configure settings
The following configurations are based on FastGPT v4.17.0, and are tailored to the document characteristics of complete home appliance manufacturing:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of single-process or test content in assembly work instructions and safety compliance reports, avoiding splitting of critical steps |
chunkSplitter | Split by paragraph + heading | Matches the hierarchical structure of documents, retains the material grouping logic of BOMs and project classification logic of test reports |
customPdfParse | Enabled | Supports non-standard PDF formats exported by equipment, adapting to test report files from safety compliance testers and performance laboratories |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Accommodates large-volume documents such as batch records and monthly sampling test reports |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Processes large-volume safety compliance test reports and repair records, avoiding parsing interruptions |
chunkSettingMode | Manual configuration | Adjusts chunking parameters for different document types, adapting to the differentiated needs of structured BOMs and free-text work instructions |
Three common mistakes
- Incorrect return parameters and empty corresponding fields when calling the image parsing interface. Cause: The image file was not uploaded in multipart/form-data format, and the image address was mistakenly passed as a regular string parameter.
- Cross-model content concatenated chunking results after document parsing. Cause:
chunkSplitterwas not set to split by heading, causing content from assembly work instructions of different home appliance models to be merged into the same chunk. - Parsing timeout and 504 status code returned after locally deploying MinerU. Cause: The
PARSE_FILE_TIMEOUT_SECONDSparameter was not adjusted, and the default timeout duration was insufficient to process large-volume safety compliance test reports.
How to confirm correct configuration
- Upload a single air conditioner assembly work instruction PDF, view the parsed chunk list, and confirm that each chunk corresponds to one assembly station or process step.
- Call the file upload interface, upload batch performance sampling test reports within 1000 MB, and confirm that the interface returns a successful status.
- View the parsing log, confirm that MinerU service call requests return normally, with no timeout or format error reports.
- Compare the original BOM document and the chunking results, confirm that key fields such as material code and applicable model are not lost during splitting.
The values provided are common starting points and should be measured against the reader's own samples.