What the data for this category looks like
Daily documents for this category include grouping rules, BOM lists, welding torque records, BMS flash records, insulation and voltage withstand reports, EOL test reports, system configuration lists, and more. Data sources include cell capacity grading data, MES systems, welding equipment, torque tools, BMS software versions, and EOL test benches. Documents are updated once per produced battery pack or energy storage cabinet, and updates are also triggered by each BMS software version change. Documents contain structured fields and professional numerical units: torque in newton meters, voltage in volts, with some including embedded tables and test parameters.
Constraints for Document Parsing and Chunking
The multi-type and structured nature of category documents requires the parsing step to support multiple formats while preserving field associations. Frequently updated single-group documents require chunking logic to adapt to small-batch, multi-version submissions and avoid redundant parsing. Numerical fields with professional units must not have cross-field parameter combinations split during chunking, as this will disrupt retrieval matching. Large EOL test reports and welding torque records increase parsing time, so timeout parameters need adjustment to prevent interruptions. Documents with embedded tables require custom parsing rules, otherwise structured data will be lost.
Setting Up the Configurations
The following configurations are written based on the interface and parameter logic of FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | This category's documents include grouping rules, test records, and other content. This range avoids splitting core parameter combinations across paragraphs, while ensuring contextual integrity during retrieval |
chunkSplitter | Segment by document structure | This category's documents include structured BOMs and long-text grouping rules. Splitting by structure preserves field associations and avoids damaging parameter logic |
customPdfParse | Enabled | This category's documents include professional formats such as insulation and voltage withstand reports and EOL tests. Custom parsing preserves original formatting and embedded table content |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Parsing large EOL test reports and welding torque records requires more time. This value covers the parsing duration of most professional documents |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Large test record files for this category may exceed the default upload limit. This value adapts to the size requirements of most professional documents |
chunkSettingMode | Manual mode | This category's documents have large structural differences. Manual configuration allows adjusting chunking logic for different document types to adapt to diverse business scenarios |
Three Common Misconfigurations
- Symptom: A timeout error is prompted when uploading an insulation and voltage withstand report larger than 10 MB. Cause: The
PARSE_FILE_TIMEOUT_SECONDSparameter was not adjusted, and the default value was used, resulting in parsing timeout for large professional documents. - Symptom: Fields are empty after parsing docx-format BMS flash records. Cause:
customPdfParsewas not enabled, and the default parsing cannot recognize the embedded table fields unique to this category. - Symptom: The cross-paragraph cell grouping logic in the grouping rules is split after chunking, making it impossible to match the complete rules during retrieval. Cause:
chunkSplitterwas set to split by line breaks instead of by document structure, damaging parameter associations.
How to Verify Correct Configuration
- Upload a typical BOM list document, and check if the parsed text retains core fields such as cell model, quantity, and rated voltage.
- Adjust the
chunkSizeparameter, then upload a long-text grouping rule document, and check if chunking does not split the core grouping logic across paragraphs. - Upload a large EOL test report, and check if the parsing process does not trigger a timeout error.
- Upload a docx-format welding torque record document, and check if the parsed text retains exclusive fields such as torque values and test equipment numbers.
The values provided are common starting points and should be measured against the reader's own samples.