This document is compatible with FastGPT v4.17.0.
What the data for this category looks like
Typical entities include motor manufacturers and power equipment enterprises’ winding, insulation, final assembly, and testing departments. Daily documents include winding process cards, insulation treatment records, dynamic balance reports, voltage withstand insulation test records, and type and factory inspection test reports. Data sources include electromagnetic design data, copper wire and insulation material batch information, winding equipment operation data, balancing machine detection data, electrical test bench test results, and ERP and MES system records. Update frequency is once per unit or batch, with synchronized updates when product type changes. Documents contain professional fields such as winding turns, paint dipping process parameters, test voltage, and temperature rise values. Supporting units include turns, megapascals, degrees Celsius, and volts.
What constraints these characteristics impose on the "document parsing and chunking" link
Document types for electric motor and generator manufacturing are diverse. They include both procedural text paragraphs and structured test tables, which demand high adaptability of chunking logic. Data updates are strongly bound to product batches. Chunking must retain batch-related metadata to avoid splitting parameter combinations across batches. Many professional fields and dedicated units are used. The parsing process must fully retain field precision and unit information to prevent parameter ambiguity during retrieval. The volume of batch-uploaded documents varies widely, ranging from single test reports to multi-batch collections. Parsing systems must support requirements for different volumes.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Motor documents contain professional content such as winding turns and test data. This length preserves the integrity of a single set of parameters and avoids splitting critical information |
chunkSplitter | Split by paragraph + table | Daily documents include both procedural text paragraphs and test data tables. Mixed splitting preserves the structural integrity of a single set of data |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Large type test reports have large data volumes. This duration reserves sufficient parsing time and avoids timeout interruptions |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Batch-uploaded multi-batch test report collections have large volumes. This upper limit supports batch upload requirements |
chunkSettingMode | Custom mode | Separate chunking rule adjustments are required for table documents to adapt to the mixed structural characteristics of motor documents |
Three common mistakes
- Symptom: Excel files uploaded via the dialog box always fail to parse, and no error is displayed in the background. Cause: Custom parsing rules for table documents are not configured. The default chunking logic does not adapt to the multi-column test data structure in motor documents.
- Symptom: Document parsing succeeds after local deployment, but the model does not answer based on document content. Cause: The
chunkSizevalue is too small, splitting critical winding turns and wiring parameter groups, resulting in failure to recall complete context during retrieval. - Symptom: Batch document uploads are interrupted mid-process, and parsing cannot resume after restarting Docker. Cause: The
PARSE_FILE_TIMEOUT_SECONDSsetting duration is insufficient. A single parsing timeout triggers task interruption, and no parsing breakpoint is retained.
How to confirm the configuration is correct
- Upload a single winding process card, check if the parsed chunks retain complete parameter paragraphs such as turns and wire diameter, with no split breaks.
- Upload multiple documents of different types, verify that the uploaded file size does not exceed the
UPLOAD_FILE_MAX_SIZElimit, and no upload failure prompts appear. - Start a batch parsing task, wait longer than the
PARSE_FILE_TIMEOUT_SECONDSsetting duration, and confirm that no forced interruption error is displayed. - Search for professional terms in the test document, confirm that the recalled chunk content includes associated test data and process parameters.
The values provided are common starting points and should be measured against the reader's own samples.