What this category's data looks like
Inspection and process record data for photovoltaic cell manufacturing comes from multiple sources. These include silicon wafer incoming material ledgers, MES production systems, operation logs of diffusion furnaces, coating equipment, or printing presses. They also include outputs from EL and IV testing equipment, and laboratory test data. Data updates follow three cadences: per production batch, after each wafer inspection, and daily aggregation. Daily documents include process cards, diffusion and coating parameter tables, printing and sintering records, EL inspection images, IV grading data, and yield reports. Some documents contain structured parameter fields and mixed text and image content.
What constraints do these characteristics impose on the document parsing and chunking link
This category’s documents combine structured parameter tables, mixed text-image inspection reports, and pure image content, and have high update frequency. Different documents have widely varying fields and formats. For example, IV grading data uses a structured table format, while EL inspection images are images with defect annotations. Tailor parsing logic to each document type. Optimize chunking for fast retrieval to support frequently updated production data. Implement unified parsing across multiple document types to avoid chunking chaos or information loss from format differences. Some documents rely on native output formats from industrial equipment. Support non-standard file encodings and layouts to accommodate these documents.
The following configurations are based on FastGPT v4.17.0.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Matches the effective information length of single blocks for photovoltaic cell process parameter tables and inspection reports, avoiding overly fragmented chunks or splitting across critical parameters |
chunkSplitter | Split by paragraph + table boundaries | Preserves the integrity of structured tables and text paragraphs, preventing splitting that breaks the associative relationship of process parameters |
customPdfParse | Enabled | Adapts to non-standard PDF layouts common in documents from industrial equipment and MES systems for this category, accurately extracting parameter fields and image content |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Meets the requirements for uploading large files such as single-batch EL inspection image collections and summary reports |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to the parsing time required for large EL image collections and multi-page summary reports, preventing parsing failures from timeouts |
chunkSettingMode | Custom mode | Flexibly adjusts chunking rules to suit different document types, addressing the wide format differences across documents in this category |
Three common mistakes
- Phenomenon: After uploading an image URL or inserting an image link, the parsing result has no image content or returns a parsing failure. Cause: Custom parsing rules are not configured to adapt to the URL format of industrial inspection images, or access permissions for the corresponding storage path are not enabled.
- Phenomenon: When calling the workflow API to upload images, the interface returns parameter errors or fails to recognize file fields. Cause: Image parameters are not passed according to FastGPT interface specifications, or request headers and parameter formats for file uploads are not correctly set.
- Phenomenon: When batch parsing multiple batches of yield reports, parsing timeouts occur or some documents fail to complete chunking. Cause: The
PARSE_FILE_TIMEOUT_SECONDSparameter is not adjusted to a value suitable for large-batch documents, or the number of files parsed per batch is not limited.
How to confirm the configuration is properly set
- Upload a typical EL inspection image PDF, and check if the parsing result extracts the image content and attached parameter annotations.
- Call the file parsing interface, pass a test IV grading data table, and confirm that the chunking result retains complete table rows and fields.
- View the system configuration page, and confirm that the values of
UPLOAD_FILE_MAX_SIZEandPARSE_FILE_TIMEOUT_SECONDSmatch the preset requirements. - Upload a test file of the maximum single size, and verify that the parsing process does not produce timeout errors.
The values provided are common starting points and should be validated against local sample data.