What data for this category looks like
Typical users of this category include IDC operations engineers, network engineers, and equipment implementation personnel. Daily processed documents include rack installation applications, cabinet cabling diagrams, port label sheets, and construction records. Data sources include DCIM systems, asset systems, network topologies, and customer work orders. Update frequency is per order or daily. Documents contain structured fields such as rack unit number (unit: U), port number, cable length (unit: meters), and construction date, plus unstructured application descriptions and acceptance records. Supported formats are PDF, CSV, and text files.
What constraints do these characteristics impose on the "document parsing and chunking" workflow
Documents have mixed types, with both structured and unstructured content. Parsing and chunking must retain both text paragraphs and table structures, and avoid splitting core compliance fields such as port correspondence. Data updates occur frequently, so single-parsing time must be controlled to support real-time work order processing. Document formats exported from different source systems are inconsistent, so custom parsing rules must be supported to adapt to port label sheets and cabling diagrams of varying formats. Compliance review requirements mandate that chunked content does not lose key associated information such as rack units and ports. Chunk breaks can cause logical errors in the review link, leading to incorrect port connections, impaired heat dissipation, or delayed customer service activation.
How to configure settings
The following configurations are based on the functionality of FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of single records in port label sheets for cabinet documents, avoids splitting associated information of the same port, and ensures semantic integrity of chunked content |
chunkSplitter | Split by paragraph + table | Supports retaining both text application descriptions and structured cabling diagrams and port sheet content, and avoids damaging the association relationships required for compliance review |
customPdfParse | Enabled | Accurately extracts vector graphic text and tables from PDF cabinet cabling diagrams, and restores structured information from the original document |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Adapts to parsing time for large PDF cabinet cabling diagrams, and avoids work order processing interruptions caused by timeouts |
chunkSettingMode | Manual mode | Allows adjusting chunking rules for different document types such as rack installation applications and port label sheets, to meet the field integrity requirements of compliance review |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Adapts to upload requirements for large-capacity PDF cabinet cabling diagrams, and avoids failed submission of parsing due to oversized files |
Three common misconfigurations
- Symptom: Custom-chunked document blocks are automatically deleted, causing the original index order to differ from expectations. Cause: The default document block deduplication mechanism is enabled, and deduplication rules are not adjusted for repeated port labels and rack unit information in cabinet documents.
- Symptom: Parsed port label sheets cannot be displayed as tables, with content split into scattered text. Cause: The table extraction switch for
customPdfParseis not enabled, or thechunkSplitterconfiguration does not retain structured table fields. - Symptom: No real-time parsing status feedback is available after creating a knowledge base via the API. Cause: Status return parameters are not configured in the API request, or the correct query interface for polling parsing progress is not used.
How to verify proper configuration
- Upload a cabinet port label sheet in CSV or PDF format, and check if the parsed result retains complete port numbers and the table structure of corresponding rack units.
- Submit a test document with repeated port labels, and confirm that the knowledge base does not automatically delete duplicate compliance-related blocks, and retains the original chunk order.
- Call the API to create a knowledge base, then poll the query interface, and confirm that status returns for parsing in progress, ready, and failed are available.
- Upload a single PDF cabinet cabling diagram under 1000 MB, and confirm that the parsing task completes normally without triggering timeout errors.
The values provided here are common starting points, and should be evaluated against samples relevant to the specific deployment scenario.
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.