What the data for this product category looks like
Typical entities in this category are air conditioner, refrigerator, washing machine, and small household appliance manufacturers. Daily documents include configuration BOMs, assembly work instructions, refrigerant charging records, electrical safety test reports, performance spot check 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 cycles follow per-unit product, shift production, or spot check schedules. Document fields include material codes, charging volumes, pressure test values, fault codes, and similar items, with corresponding units such as kilograms, volts, and hours.
What constraints do these characteristics impose on the document parsing and chunking link?
Document version differences caused by mixed-line production of multiple product categories require the parsing link to accurately match the document structure of the corresponding model, to avoid cross-model field confusion. High-frequency updated per-unit or shift data requires chunk granularity to adapt to small-batch documents generated in short cycles, to avoid redundant chunks occupying storage space. Compliance-related documents such as safety and performance documents contain mandatory compliance fields, so field integrity must be retained, and key parameters must not be truncated during chunking. Heterogeneous data formats from multiple devices require the parsing link to support documents that mix structured tables and unstructured text, while adapting to the extraction of fields with units such as batch numbers and test values.
How to configure the settings
The following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of single-segment process descriptions and test data in documents of this category, avoiding truncation of key parameters such as charging volumes and pressure test values |
chunkSplitter | Chunk by paragraph + semantic meaning | Adapts to documents with natural processes such as assembly work instructions and repair records, retaining the integrity of process steps |
customPdfParse | Enable structured parsing mode | Adapts to documents with tables such as BOMs and test reports, accurately extracting field and unit information |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Meets the upload requirements for multiple large test reports per batch and complete production line BOM packages |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to the parsing time of large safety compliance test reports, avoiding parsing failures caused by timeouts |
chunkSettingMode | Manual configuration | Adjust chunk parameters for different document types such as work instructions and test reports to adapt to heterogeneous document scenarios |
Three common configuration mistakes
- Slow vectorization after uploading a PDF, with empty search test results. The cause is that structured parsing for
customPdfParseis not enabled, resulting in redundant unstructured text chunks, andchunkSizeis not adjusted to adapt to document length. - Empty data processing after local deployment, with no parsing completion prompt after uploading a file. The cause is that
PARSE_FILE_TIMEOUT_SECONDSis not configured correctly, orUPLOAD_FILE_MAX_SIZEis set lower than the actual uploaded file size, resulting in interrupted parsing processes. - Document parsing process is triggered by query text content. The cause is that association rules are not set correctly, causing document parsing logic to be triggered for non-knowledge base queries, confusing training and conversation scenarios.
How to confirm the configuration is effective
- Upload a typical refrigerant charging record PDF, check whether the parsed text retains fields with units such as charging volume and equipment number, with no obvious truncation.
- Check that there are no timeout errors related to
PARSE_FILE_TIMEOUT_SECONDSin the system logs, confirming that large documents can be parsed normally. - Upload a safety compliance test report package exceeding 500 MB, confirm that the upload progress is normal and there is no file size limit prompt.
- Initiate a conversation unrelated to the knowledge base, confirm that the document parsing process is not triggered, and verify that the configuration takes effect.
The values provided are common starting points and should be measured against the reader's own samples.