What the data for this category looks like
The data sources for this category are defect management systems, test reports, code commit records, and user feedback. The update frequency is daily or per version release cycle. Commonly processed documents include defect tickets, regression checklists, repair verification records, and defect statistical reports. Document structures include fields such as defect ID, priority, reproduction steps, associated version numbers, and verification results. Most fields are strings, enumeration values, and semantic version numbers, with no unified numerical statistical unit. Data is generated around the functional defect lifecycle, covering the full process from defect discovery to repair verification.
What constraints do these characteristics impose on the document parsing and chunking link
Documents in this category revolve around the functional defect lifecycle, and contain strongly related content such as reproduction steps and priority judgments. When chunking, context integrity must be retained to avoid splitting key logic that would make reproduction difficult. The characteristic of data being updated per version requires that chunk results be associated with corresponding version identifiers, to prevent cross-version content confusion that would affect regression verification. Differences in multi-source data formats require the parsing link to support multiple import formats such as PDF, PPT, and text, while also retaining complex layout structures such as tables and formulas. Frequently updated defect tickets and reports require controlling parsing time to avoid delaying test process progress.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Core paragraphs of defect tickets and regression checklists are mostly within hundreds of characters, avoiding splitting key context such as reproduction steps and priority descriptions |
chunkSplitter | Chunk by paragraph + heading | Defect documents have clear heading levels for priority, reproduction steps, etc. Retaining structure improves relevance for subsequent retrieval |
customPdfParse | Enabled | Adapts to PDF formats of defect statistical reports and repair verification records, retaining complex layout content such as tables and formulas |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large defect statistical reports take longer to parse, avoiding interrupting the parsing process due to timeout |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports batch import of version regression checklists and historical defect reports, adapting to upload requirements for large-scale test documents |
chunkSettingMode | Manual configuration | Adjust chunking rules for different types of defect documents, adapting to associated display of exclusive fields such as priority and version numbers |
Three Common Mistakes
- Symptom: An error occurs when executing the command
docker run --gpus all -itd -p 7231:7231during Docker deployment ofmarker_images:v0.1. Cause: The local GPU driver version does not match the image dependencies, or there is a compatibility difference between the FastGPT version and the image version. - Symptom: Clicking chunk preview on a PDF file in the FastGPT knowledge base returns a prompt that the file content cannot be read. Cause: The
customPdfParseconfiguration is not enabled, the uploaded file exceeds theUPLOAD_FILE_MAX_SIZElimit, or the document is encrypted. - Symptom: After configuring the Feishu knowledge base, PPT and PDF documents cannot be synced. Cause: The parsing permission for the corresponding document format is not enabled, or the
customPdfParseconfiguration does not cover document types synced via Feishu.
How to Confirm the Configuration Is Correct
- Upload a defect statistical report PDF containing complex formulas, and check in FastGPT v4.17.0 whether the chunk preview after parsing retains the formula structure and complete text content.
- Execute the specified Docker run command, check that the container startup log has no errors, and that port 7231 is accessible normally.
- Upload a regression checklist document exceeding the default size, confirm that the system prompt complies with the
UPLOAD_FILE_MAX_SIZErestriction rules. - Configure the Feishu knowledge base and upload a PPT-format regression checklist, confirm that the document can be synced and parsed normally.
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.