What the data for this category looks like
Daily documents for this category include purchase orders, arrival acceptance slips, serial number lists, and invoice materials. Data sources cover ERP systems, supplier shipping documents, logistics tracking information, and equipment manufacturer backend management systems. Most documents use structured table formats, with fields including purchase order numbers, equipment models, serial numbers, arrival dates, quantities, and others. Updates occur per order or weekly. Individual document groups typically have small file sizes.
What constraints do these characteristics impose on the document parsing and chunking link
Structured table documents for this category require retention of field relationships. Document parsing and chunking must avoid splitting table cell content, to prevent loss of key identifiers such as equipment serial numbers and purchase order numbers. Heterogeneous document formats from multiple sources require parsing logic with adaptability, to support compatibility with document layout differences exported from different systems. The high-frequency update rhythm (per order or weekly) requires lightweight parsing processes, to adapt to small-batch, high-frequency processing scenarios. Consequences of errors such as equipment mismatches and project delays require that parsing and chunking accuracy meets basic requirements for business compliance review.
How to set the configuration
The following configurations are based on official logic for FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Most documents for this category are short forms with tables. This range covers a single piece of equipment information plus associated document fields, and avoids broken relationships after splitting |
chunkSplitter | markdown chunker (table recognition mode enabled) | This category contains a large number of structured tables. Enabling table recognition mode retains complete content within cells, and prevents damage to field relationships |
customPdfParse | Enabled | Invoices and arrival acceptance slips for this category mostly use PDF formats. Enabling custom PDF parsing retains original layout and table structure, and prevents text disorder |
PARSE_FILE_TIMEOUT_SECONDS | 60–90 seconds | Individual documents for this category have small file sizes. This range covers parsing time for documents from multiple sources, and avoids timeout failures |
UPLOAD_FILE_MAX_SIZE | 100 MB | Individual document groups uploaded daily for this category typically do not exceed 100 MB, which aligns with standard business upload limits |
chunkSettingMode | Manual configuration | Document formats for this category are affected by differences between suppliers and systems. Manual configuration allows adaptation to document structures from different sources |
Three Common Mistakes
- Symptom: Missing equipment serial number or purchase order number fields in parsed chunks. Cause: Table-first chunking logic is not enabled, and cell content from structured tables is split into independent chunks, damaging field relationships.
- Symptom: Empty results returned when calling the API to get chunk indexes. Cause:
chunkSettingModeis not correctly configured to manual mode, orchunkSizeis set beyond the actual character length of individual documents, so no valid chunks are generated. - Symptom: Parsing fails when uploading multiple packaged documents, with a timeout prompt. Cause:
PARSE_FILE_TIMEOUT_SECONDSis set too small, failing to cover parsing time for batch processing of multiple documents, orcustomPdfParseis not enabled, leading to timeout when parsing complex PDF formats.
How to Confirm the Configuration is Correct
- Upload a typical arrival acceptance slip PDF, check the parsed chunk list, and confirm that each chunk contains complete equipment model, serial number, and associated order number fields.
- Call the document parsing API, check that the returned results contain valid chunk indexes and corresponding content, with no null values or missing fields.
- Upload the largest individual document, confirm that the parsing process does not trigger a timeout error, and that
PARSE_FILE_TIMEOUT_SECONDSaligns with actual parsing time. - Adjust
chunkSizeto different ranges, compare the completeness of parsed chunks, and confirm that the current value does not split a single complete piece of equipment information.
The values provided are common starting points, and should be evaluated against the samples relevant to the deployment.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-10-19.