What data looks like for this category
Daily data from injection molding, stamping, and die-casting mold enterprises, as well as mold centers of manufacturing enterprises. Sources include customer product models, mold flow simulation files, CAD/CAM output files, machining and measurement data, and records related to trial mold samples. Data is updated per project, with new documents generated after each trial mold or mold modification. Typical documents include mold design task sheets, DFM reports, part drawings, processing technology cards, trial mold reports, mold modification notices, and mold records. These documents contain fields such as mold number, number of cavities, material grade, hardness, and trial mold defect description. Units include millimeters, megapascals, hours, and other standard units.
What constraints these characteristics impose on document parsing and chunking
Mold design and manufacturing documents have diverse types: structured process sheets, semi-structured trial mold reports, and formatted drawings exported from CAD. General-purpose parsing logic cannot reliably extract information from all these formats. Documents are updated per project and revised across multiple trial mold cycles. Chunking must preserve contextual connections within the same project, to avoid splitting defect records across trial mold stages. Specific business fields such as mold number and number of cavities must be accurately located, and chunking must not truncate these critical parameters. Document volume varies widely, from mold modification notices of several KB to complete project packages of tens of MB. This creates varying requirements for resource consumption and processing duration during parsing and chunking.
How to set configurations
This guidance is written for FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Mold documents mix structured parameters and unstructured descriptions. This range balances the integrity of information per chunk and retrieval accuracy |
chunkSplitter | Split by paragraph + title boundary | Mold documents have clear chapter divisions. Splitting by paragraph retains semantic integrity and avoids truncating business logic |
customPdfParse | Enabled | Mold design documents are mostly formatted PDFs exported from CAD. Enabling custom parsing accurately extracts table and annotation information |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large mold project document packages have significant volume. Parsing requires a longer duration to avoid mid-process timeout interruptions |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Complete mold project document collections may include multiple CAD files and reports. Larger file uploads must be supported |
Three common configuration errors
- Symptom: After uploading mold project documents, the interface displays a total chunk count exceeding 3000, and indexing fails to complete. Cause: The
chunkSizeparameter was not adjusted to match the project granularity of the documents. Default chunking parameters produce overly small chunks, and large-volume documents were not split per project. - Symptom: After uploading a CSV-format processing technology card, parsing results show field misalignment or missing fields. Cause: The
chunkSplitterlogic was not adjusted for structured table documents. The default chunking logic does not recognize the column structure of CSV files, leading to chaotic information splitting. - Symptom: The system returns a timeout error when parsing large mold design PDF files. Cause: The
PARSE_FILE_TIMEOUT_SECONDSsetting is lower than the actual required parsing duration, and the timeout parameter was not adjusted for large-volume documents.
How to verify correct configuration
- Upload a single typical mold document, such as a trial mold report, and review the parsed chunk list to confirm the chunk count aligns with business expectations.
- Upload a structured table document, such as a processing technology card, and check the parsing results to confirm fields are not misaligned and are fully extracted.
- Upload the largest-volume project document package, and confirm that the upload and parsing processes do not trigger timeout errors.
- Adjust the
chunkSizeparameter, then compare parsing results across different chunk lengths to confirm that information integrity meets requirements.
The values provided are common starting points and should be measured against the user's own samples.