What the Data for This Category Looks Like
Production ledgers for photovoltaic module encapsulation come from cell grading records, glass, adhesive film and backsheet batch ledgers, workshop MES systems, stringing and lamination equipment logs, EL defect tester outputs, and IV power tester outputs. Data is updated such that each module generates a single record, and each batch of materials generates a batch summary document. Daily documents include layout drawings, material batch records, lamination process records, EL defect images, IV test reports, insulation voltage withstand records, and packing lists. Fields include cell current grade, lamination temperature and time, EL crack coordinates, power value, batch number, and more, with corresponding professional units such as amperes, degrees Celsius, and watts.
Constraints on the Document Parsing and Chunking Link
This category of ledger documents has diverse types, including vector layout drawings, structured tables, and test reports, requiring adaptation to different parsing logic. Single-module records have small size but high update frequency, so parsing speed must be matched to workshop production cycles. Professional fields such as EL crack parameters and current grades must be fully retained, and must not be altered or omitted by general parsing rules. Batch-uploaded batch reports can reach tens of megabytes, so parsing stability and timeout tolerance must be balanced.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
PARSE_FILE_TIMEOUT_SECONDS | 600–900 seconds | Adapt to batch-uploaded EL/IV test reports, avoid timeout during large file parsing |
chunkSize | 800–1200 characters | Match the field density of module ledgers, avoid overly fine chunking causing retrieval redundancy or overly coarse chunking losing critical information |
chunkSplitter | Split by document type | Differentiate parsing logic for layout drawings, process records, and test reports, adapt to different document structures |
customPdfParse | Enabled | Target the specialized table structure of EL images and IV test reports, use custom parsing rules to extract hidden fields |
UPLOAD_FILE_MAX_SIZE | 2000 MB | Support batch upload of batch material records and test reports |
chunkSettingMode | Manual configuration | Adapt to differences in ledger formats across departments such as stringing and testing |
Three Common Mistakes
- When uploading EL test reports larger than 10 MB, the prompt
timeout of 360000ms exceededappears. The cause is failure to adjust thePARSE_FILE_TIMEOUT_SECONDSparameter; the default timeout duration is insufficient for large file parsing. - The cell current grade field is empty in chunking results. The cause is failure to enable
customPdfParsecustom parsing; the default parsing rules cannot recognize nested fields in specialized test tables. - Critical information is truncated after merging segments due to exceeding the ideal chunk length. The cause is failure to configure a reasonable range for
chunkSizeand enable automatically adaptive segment merging logic.
How to Confirm Proper Configuration
- Use FastGPT v4.17.0, upload a single IV test report, check the parsed chunking results, confirm that professional fields such as cell power and batch number are fully extracted.
- Adjust
PARSE_FILE_TIMEOUT_SECONDSto 720 seconds, upload a 15 MB batch test record, confirm no timeout errors occur. - Configure a custom JavaScript parsing method, verify that it can accurately match the original text responses of question-answer pairs in the knowledge base, and retain original field expressions.
- Randomly sample chunked content, confirm that individual chunk lengths fall within the 800–1200 character range, with no overly long or short invalid chunks.
The values provided are common starting points and should be measured against the reader's own samples.