What the data for this category looks like
Data related to photovoltaic module encapsulation equipment inspection and maintenance comes from multiple sources: battery grading records, glass adhesive film and backplane batch ledgers, workshop MES systems, string welding and lamination equipment logs, EL/IV tester reports, insulation withstand voltage records, and packaging lists. Update frequency follows one test record per module, and one batch ledger per material batch. Available document types include structured process tables, EL defect images with image annotations, parameterized test reports, and layout schematics. Fields include current gear, lamination temperature and time, crack location coordinates, power value, withstand voltage strength, and more. Corresponding units include amperes, degrees Celsius, seconds, watts, megapascals, and more.
What Constraints These Characteristics Impose on the Document Parsing and Chunking Link
This segment’s data features create multiple constraints for the document parsing and chunking workflow. Multi-source, high-frequency documents require parsing workflows to support batch processing of different formats, and avoid excessive per-document parsing time. Mixed document types, including structured process tables and unstructured EL images, require chunking rules that balance text field extraction and associative matching of image-embedded text. Exclusive test records for each module require chunking to retain complete contextual associations, ensuring full parameter matching for the corresponding module during retrieval. Differences in professional terminology across documents require retaining original field names and units during chunking to avoid retrieval ambiguity.
How to Configure Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Photovoltaic encapsulation documents include long process descriptions and single-parameter records; this range balances contextual integrity and retrieval accuracy |
chunkSplitter | By paragraph + header/footer markers | Layout diagrams and process records have fixed equipment number and batch identifier headers; splitting this way preserves business-related structure |
customPdfParse | Enabled | Defect text embedded in EL images and table fields from IV test reports must be extracted; enabling this adapts to non-standard PDF parsing |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Parsing large-size EL images and batch records takes longer; this range prevents normal parsing from triggering timeouts |
UPLOAD_FILE_MAX_SIZE | 500 MB | Packed files for single batch material records typically do not exceed this size, covering most business scenarios |
chunkSettingMode | Manual mode | Chunking rules need adjustment for process tables, EL images, and test reports separately; manual mode enables fine-grained configuration |
Three Common Mistakes
- Parsing results lack embedded text after uploading a PDF in EL image format. The cause is that the
customPdfParseconfiguration is not enabled. The default parsing rule does not extract defect annotation information embedded in images. - File parsing functionality malfunctions after upgrading to FastGPT v4.17.0. The cause is that
chunkSettingModeis not switched to manual mode. The new version’s default automatic chunking rules do not match the mixed structure of photovoltaic documents. - An upload interception prompt returns when calling the file parsing tool. The cause is that
UPLOAD_FILE_MAX_SIZEis set lower than the actual uploaded batch record package size, leading to the file being rejected for parsing.
How to Verify Proper Configuration
- Upload a standard lamination process record PDF, and check if the parsed result includes business-related field content.
- View background parsing logs to confirm no timeout errors occur, verifying that the
PARSE_FILE_TIMEOUT_SECONDSconfiguration takes effect. - Test chunked text retrieval by entering business-specific terminology, and check if chunked content from the corresponding document is matched.
- Upload a test file in EL image format, and confirm the parsed result includes annotation text embedded in the image, verifying that the
customPdfParseconfiguration takes effect.
The values provided are common starting points and should be evaluated against samples relevant to specific deployments.