What This Category of Data Looks Like
Daily documents for the general mechanical equipment assembly category include assembly process cards, BOMs, torque records, gap adjustment records, commissioning reports, factory inspection sheets, and packing lists. Data sources include PLM/ERP systems, part inspection data, assembly tool collection records, test bench data, MES systems, and customer technical agreements. Updates occur once per device or per order. Most document structures include fields such as device model, order number, specific process parameters, and inspection results. Some documents have multi-level nested directory structures.
Constraints for Document Parsing and Chunking
Multi-level nested directory structures require parsing tools to fully traverse subdirectory content, without missing lower-level documents. A large number of precise process parameters with units in documents require retaining the binding relationship between parameters and corresponding process descriptions during chunking. This prevents retrieval failure caused by parameters being separated from their context after splitting. High-frequency updates at the single order or single device level require the parsing process to adapt to processing efficiency for batch small-volume documents. Some documents embed doc format files and images, so parsing and text extraction for non-PDF formats must be supported.
Configuration Settings
These configuration items apply to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the combined length of process parameters and descriptions in documents, preventing parameters from being separated from context after splitting |
chunkSplitter | Split by paragraph + heading | Matches the hierarchical structure of assembly process cards and inspection sheets, retaining the association between process steps and parameters |
customPdfParse | Enabled | Adapts to composite document parsing with embedded doc files and images, covering multiple document types in this category |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Compatible with the size of assembly document packages bundled in batches for single orders, preventing upload truncation |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Addresses the full parsing time required for multi-level directory documents, preventing mid-parsing timeout interruptions |
chunkSettingMode | Manual mode | Allows adjusting chunking rules for different document types such as BOMs and process cards, adapting to the diversity of category documents |
Three Common Configuration Mistakes
- A 413 error is prompted when uploading an assembly document package. This occurs when the
UPLOAD_FILE_MAX_SIZEconfiguration value is smaller than the total size of the uploaded document package, and the bulk bundled volume of multiple assembly files is not accounted for. - The external pdf-maker parser key is not configured. This prevents documents with embedded doc files and images from being parsed, and corresponding content cannot be extracted.
- Calling the API to upload parsed content shows success but no returned results. This happens when a reasonable
PARSE_FILE_TIMEOUT_SECONDSduration is not set, or the parsed result is not correctly bound to the callback link, leading to failure to return the result normally.
How to Verify Correct Configuration
- Upload an assembly process package that includes multi-level directories, embedded doc files and images. Check if all subdirectory documents and embedded content are fully included after parsing.
- Check the
UPLOAD_FILE_MAX_SIZEconfiguration value. Upload a test file of the corresponding size, and confirm that no 413 error is prompted. - Initiate a batch parsing task for single-order documents. Check if the parsing time is within the range set by
PARSE_FILE_TIMEOUT_SECONDS, and there are no timeout interruptions. - Adjust
chunkSizeto different ranges. Search for process parameter-related questions, and confirm that parameters and corresponding process descriptions are always bound to the same chunk.
The values provided here are common starting points and should be measured against the reader's own samples.