This configuration is formulated based on the rules of FastGPT v4.17.0.
What the data for this category looks like
The core data for this category comes from orders and size charts, CAD pattern files, fabric and accessory inspection records, hanging system data, needle detector logs, and customer sample sealing materials from garment OEM/ODM and brand supply chains. Daily generated documents include process sheets, layout plans, cutting lists, process quality tables, packaging and size matching lists, etc. Each document is bound to the corresponding style and order batch, with update frequency adjusted per order style and production batch. Document fields include style number, size range, fabric gram weight, cutting quantity, process man-hours, etc. Some layout charts exported from CAD come with pixel-level piece coordinate information, with units mostly centimeters, grams, and pieces.
What constraints do these characteristics impose on the "document parsing and chunking" link
Documents in this category have the characteristics of multiple styles and strong binding of single-style data. Each order style corresponds to independent process and cutting parameters, and cross-style data confusion must be avoided during parsing. Documents include structured fields and unstructured CAD drawings and layout charts. Chunking must retain field association relationships, and avoid splitting that destroys key information such as piece coordinates and size matching ratios. Frequently updated batch data requires the parsing process to adapt to rapidly iterated document versions, while also accurately matching order batches with corresponding document content to prevent confusion of fabric color batches and cutting quantity information across different batches.
How to set the configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | This category's documents include structured fields and unstructured drawing descriptions. This range can retain the integrity of core information such as processes and sizes |
chunkSplitter | By paragraph + semantic boundary | Process sheets and process tables have clear paragraph divisions. Splitting by semantic boundaries can avoid separating process requirements from corresponding piece information |
customPdfParse | Enable structured parsing mode | Some documents come from CAD-exported PDFs. Enabling this mode can extract table fields and drawing metadata, avoiding loss of key information such as cutting quantity and fabric specifications |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports uploading large layout chart CAD files, adapting to the storage and parsing needs of batch piece drawings for this category |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large CAD drawing parsing takes a long time. This duration can cover the parsing process of most batch files |
chunkSettingMode | Manual configuration | Documents in this category have large structural differences. Manual configuration allows targeted adjustment of chunking rules to adapt to different types of process sheets and layout charts |
Three common mistakes
- Attempting to upload and parse audio or video files results in a failure, with an error prompt indicating that the file format is not supported. The parsing process for this category is adapted to document formats such as process sheets and layout charts by default, and no audio and video parsing adaptation rules are configured.
- Calling the workflow API to upload image files returns a parameter error. The image file field is not passed as required, or the interface parameter mapping rules for image parsing are not correctly configured.
- Chunking results lose key drawing information such as piece coordinates. The structured parsing mode of
customPdfParseis not enabled, causing CAD-exported PDF drawings to only be recognized as plain text without metadata extraction.
How to confirm the configuration is correct
- Upload a process sheet PDF of this category, and check if the parsed text blocks retain core fields such as processes and sizes, with no obvious content fragmentation.
- Call the file upload interface, upload a CAD layout chart file within 1000 MB, and confirm that the upload progress is normal with no errors.
- Check the parsing task logs, and confirm that the configured duration of
PARSE_FILE_TIMEOUT_SECONDSdoes not trigger a timeout interruption. - Manually adjust the
chunkSizeparameter, upload the same layout chart document, and compare the chunking results under different configurations to confirm that the expected information integrity is met.
The values provided are common starting points and should be measured against the reader's own samples.