What the data for this category looks like
Data for this category comes from defect management systems, test reports, code commit records, and user feedback. Update frequency is daily or per release cycle. Commonly processed documents include defect tickets, regression test checklists, repair verification records, and defect statistics reports. Most document structures include fields such as defect ID, priority tags, reproduction steps, associated version numbers, and verification status. Some reports include statistical values classified by severity. Overall formats are a mix of structured tables and plain text descriptions.
Constraints imposed on document parsing and chunking
Data comes from multiple system outputs and manually filled documents. Parsing must support mixed structured table and free text formats. Avoid breaking field associations between defect IDs, reproduction steps, and associated version numbers during chunking. High-frequency daily or per-release updates require parsing workflows to support batch processing of uploaded documents. The workflow must also adapt to differences in document lengths. Overlong regression test checklists need reasonable chunking to preserve context integrity. Some defect statistics reports have dense numerical values. Chunking must retain the binding between numerical values and their corresponding tags, to avoid field misalignment after parsing.
Configuration Settings
This configuration applies to FastGPT v4.17.0.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the average context association range of documents such as defect tickets and regression checklists, avoids breaking field binding relationships |
chunkSplitter | Split by paragraph + heading | Matches the heading hierarchy and paragraph-based defect descriptions of documents, preserves semantic integrity |
customPdfParse | Enabled | Retains table and typesetting structure in PDF documents, prevents structured fields from being split and misaligned |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Adapts to the parsing time requirements of large defect statistics reports, prevents parsing failures due to timeouts |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports batch uploading multiple large regression test documents and defect reports |
Three Common Configuration Mistakes
- Phenomenon: After uploading large compliance-related PDF documents, question answering results show field misalignment or missing key information. Cause: Chunking parameters were not adjusted for documents mixing structured and unstructured content, leading to associated fields being split.
- Phenomenon: After batch uploading multiple defect ticket PDFs, only some documents are successfully parsed. Cause: Reasonable
UPLOAD_FILE_MAX_SIZEandPARSE_FILE_TIMEOUT_SECONDSwere not set, leading to timeouts or file truncation. - Phenomenon: After uploading an Excel-format defect statistics report, the parsing result only extracts plain text and loses table numerical values. Cause: The parsing switch for the corresponding format was not enabled, leading to structured table content being flattened.
How to Verify Proper Configuration
- Upload a standard defect ticket PDF, check if the parsed text blocks retain the binding relationship between defect ID and reproduction steps.
- Batch upload 3-5 test documents in different formats, verify that all parsing statuses show success.
- Upload a large regression test checklist, confirm that parsing time does not exceed the preset
PARSE_FILE_TIMEOUT_SECONDSthreshold. - Enter the knowledge base configuration page, confirm that the selected chunking parameters match the preset configuration.
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 2026-10-19.