What the Data for This Category Looks Like
Documents for internal office and employee services of rail transit equipment come from internal systems. These include equipment operation and maintenance manuals, maintenance records, employee training courseware, and supply chain collaboration documents. Updates occur irregularly alongside new model launches and annual maintenance specification adjustments. Single document lengths vary widely. Some operation and maintenance manuals can reach tens of thousands of characters. Documents have clear multi-level tables of contents. Most use PDF format, with some containing embedded doc attachments, engineering drawing screenshots, and professional tables. Fields include professional technical details such as equipment number, maintenance cycle, and spare parts model.
Constraints Imposed by These Characteristics on Document Parsing and Chunking
A high proportion of long text requires chunk sizes adapted to complete technical paragraphs. This avoids splitting professional terms or continuous operational steps. Multi-level table of contents structures require retaining hierarchical connections during parsing. Without this, retrieval results cannot match specific chapter content. Mixed format content requires the parsing module to support extraction of multiple attachment types. Otherwise, parameter information in drawing annotations and embedded documents cannot be retrieved. Dense professional fields require retaining field context during chunking. Splitting related parameters across different chunks reduces retrieval accuracy.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Rail transit equipment documents contain a large number of professional paragraphs and long tables. This range preserves complete technical context and avoids splitting professional terms or operational steps |
chunkSplitter | Split by title and paragraph | Documents have clear multi-level table of contents structures. Splitting by title preserves hierarchical connections and improves retrieval accuracy |
customPdfParse | Enabled | Built-in parsers have limited ability to extract tables and drawing annotations from engineering PDFs. Custom parsing supports embedded doc attachments and special format content |
UPLOAD_FILE_MAX_SIZE | 2000 MB | Single large equipment operation and maintenance manuals may exceed 1000 MB. Reserving sufficient space supports complete document uploads |
PARSE_FILE_TIMEOUT_SECONDS | 1200 seconds | Long documents take longer to parse. This setting prevents parsing from being interrupted by timeouts |
chunkSettingMode | Manual mode | Adjust chunking parameters for different document types such as maintenance records and training courseware, to meet context requirements for professional fields |
Three Common Configuration Mistakes
- Symptom: An external pdf-maker parser fails to parse documents normally. Logs show key verification failure. Cause: The corresponding service key is not filled in the FastGPT backend's custom parsing settings. External parsing services cannot be called without this key.
- Symptom: Feishu multi-level table of contents documents only parse top-level content. Embedded doc files and images cannot be retrieved. Cause: The multi-level table of contents capture switch is not enabled. Parsing adaptation for non-standard format content is not configured. This results in loss of non-top-level content.
- Symptom: Calling the relevant API to parse a file shows a successful call, but no parsing results are returned. This affects FastGPT deployments via Docker with incorrectly configured parsing module file transfer. Cause: The communication configuration between the parsing module and main service is incorrect. File parsing results are not correctly returned to the API interface.
How to Confirm Correct Configuration
- Upload a typical rail transit equipment operation and maintenance PDF document. Check if the parsed text retains professional fields such as equipment number and maintenance cycle.
- Access the System Settings > Parsing Module menu in FastGPT v4.17.0. Verify that parameters including
chunkSizeandchunkSplittermatch the preset configuration. Confirm thecustomPdfParseswitch is enabled. - Upload a test document with multi-level table of contents, embedded doc files, and engineering drawing screenshots. Verify all hierarchical content and attachments are fully parsed and retrievable.
- Trigger a large file upload under 2000 MB. Confirm the upload progress is normal and no timeout prompt appears. Check that parsing time falls within the range defined by
PARSE_FILE_TIMEOUT_SECONDS.
The values provided are common starting points and should be measured against the reader's own 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.