What the data for this category looks like
The data for this category comes from cell capacity grading systems, MES platforms, welding equipment, torque tools, BMS software, and EOL test benches. Updates occur each time a single battery pack or single energy storage cabinet is produced, or each time the BMS software version is changed. Daily documents include grouping rules, BOM lists, welding and torque records, BMS flashing logs, insulation and voltage withstand reports, EOL test reports, and system configuration lists. Fields include cell parameters, torque values, test voltages, software version numbers, and more. Formats include structured tables, non-standard PDFs exported by equipment, and log files.
Constraints on the document parsing and chunking workflow
Documents for this category have diverse formats. They include structured Excel tables, as well as non-standard PDFs and log files exported by industrial equipment. This requires adapting to multiple types of parsing solutions. Data updates happen at high frequency, and each individual pack corresponds to separate documents. This requires precise chunking of single files to avoid mixing information across packs. Fields include specialized parameters such as torque values and insulation resistance. Original units and numerical precision must be preserved. Chunking must avoid truncating critical parameter combinations. Some documents have format shifts or garbled text issues. Targeted parsing must be enabled to adapt to industrial scene formats.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the combined length of specialized parameters and text in category documents, avoids truncating critical information such as grouping rules and BOM lists |
chunkSplitter | Split by paragraph + title | Matches the characteristic that category documents are mostly structured chapters, preserves the integrity of paragraphs such as grouping rules and test processes |
customPdfParse | Enabled | Adapts to non-standard PDF formats exported by industrial equipment, resolves OCR recognition deviations and garbled text issues |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Addresses parsing time for large-volume BOMs or multi-page EOL test reports, avoids timeout interruptions |
UPLOAD_FILE_MAX_SIZE | 500 MB | Accommodates batch upload requirements for multiple documents supporting a single energy storage cabinet |
chunkSettingMode | Manual configuration | Addresses the diversity of category documents, allows adjusting chunking rules to adapt to different formats of logs and reports |
Three Common Misconfigurations
- An OCR Error alert appears after enabling
customPdfParse. The cause is failure to adapt to embedded fonts or encrypted formats of industrial PDFs, which prevents the high-precision parsing module from recognizing text. - Excel files fail to parse when uploaded in simple mode. The cause is failure to configure parsing trigger rules for the corresponding file type, or failure to specify that the model supports structured file parsing.
- The
PARSE_FILE_TIMEOUT_SECONDSconfiguration is set too low in private deployments. Parsing succeeds but FastGPT reports a timeout. The cause is that parsing service time exceeds the set threshold, and adjustments are not synchronized with FastGPT's global timeout parameters.
How to Verify Proper Configuration
- Upload a standard BOM Excel file under FastGPT v4.17.0. Check that the parsed text includes complete cell models, quantities, and grouping parameters, with no missing fields or garbled text.
- Upload a multi-page industrial PDF-format EOL test report. Confirm that parsed chunks are split by chapter, with no cross-chapter concatenation.
- Upload a single document package with a volume not exceeding 500 MB. Confirm that the upload and parsing process has no timeout errors, and parsing time meets expectations.
- Adjust
chunkSizeto 1000 characters. Verify that the length of chunked text matches the set value, with no truncation of critical parameters.
The values provided are common starting points, and should be evaluated against the reader's own samples.