What the data for this category looks like
Data related to construction machinery service consultation and response primarily comes from official equipment operation and maintenance manuals, parts specifications, construction work condition records, and on-site maintenance guides. Updates are released irregularly, alongside equipment model iterations and new construction scenario additions. Document structures mostly consist of chaptered technical parameters, operation steps, and troubleshooting workflows. Fields include equipment model, rated load, working radius, rotational speed, and more. Units are mostly common engineering units such as kN, m, r/min. Some documents include nested tables and sequential step lists.
What constraints these characteristics impose on document parsing and chunking
The characteristics of construction machinery documents create clear constraints for parsing and chunking:
- Technical parameter fields are bound to their corresponding units. Do not split parameters and their units during chunking.
- Troubleshooting workflows follow sequential logic. Preserve step continuity when chunking, and avoid truncating complete troubleshooting paths.
- Some documents include nested tables. Extract full table content during parsing, and do not split cells or table structures.
- Construction work condition records are ordered by time. Avoid truncating single complete records during chunking.
How to set configurations
| Configuration Item | Recommended Approach | Basis for This Approach |
|---|---|---|
chunkSize | 800–1200 characters | Matches the average length of technical parameter sections and troubleshooting step sections in construction machinery documents, avoids splitting complete parameter groups or steps |
chunkSplitter | Split by paragraph + heading | Aligns with the document's chapter structure, preserves the integrity of technical modules |
customPdfParse | Enabled | Preserves nested tables and sequential step list formats in PDFs, prevents separation of parameters and their corresponding units |
PARSE_FILE_TIMEOUT_SECONDS | 120 seconds | Accommodates the parsing duration of large maintenance manuals, prevents timeout interruptions of parsing tasks |
UPLOAD_FILE_MAX_SIZE | 500 MB | Accommodates bulk uploads of collections of equipment operation and maintenance manuals |
chunkSettingMode | Custom mode | Allows separate adjustment of chunking rules for parameter sections and step sections |
Three Common Mistakes
- After uploading a PDF maintenance manual, search returns no matching results, and the interface displays a parsing failure error. The cause is that
customPdfParseis not enabled, leading to incorrect splitting of nested parts specification tables, and valid parameter fields cannot be extracted. - After updating FastGPT, the advanced configuration entry for document parsing cannot be found. The cause is that
chunkSettingModeis not switched to custom mode. The default quick mode only displays basic chunking parameters. - After setting
chunkSizeto 500 characters, technical parameters and their corresponding units are split into different chunks. The cause is that the chunk length is too small, truncating complete paragraphs that include both parameters and units.
How to Confirm Proper Configuration
- Upload a single equipment specification document, view the parsed chunk content, and confirm that each chunk contains complete parameters and their corresponding units.
- Initiate a knowledge base search test, enter typical parameter keywords such as "rated load", and confirm that complete chunks containing this parameter are retrieved, with no truncated or missing content.
- Navigate to the knowledge base settings page in FastGPT v4.17.0, confirm that
customPdfParseis enabled, andchunkSettingModeis set to custom mode. - Upload a single maintenance manual larger than 200 MB, and confirm that the parsing task completes normally, with no timeout errors.
The values listed here are common starting points, and should be evaluated against your own specific samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-28.