What this category of data looks like
Data for this category originates from community report systems, user-submitted content details, user historical interaction records, and official review rules. Common processed documents include report tickets, violation content handling records, user appeal forms, and standardized user notification templates. Data updates run daily or in real time. Each individual data entry includes fields such as report ID, involved content link, report reason, processing progress, and more. There is no unified fixed unit. The size of a single data entry varies based on the type and quantity of reported content.
What constraints do these characteristics impose on the "document parsing and chunking" link
Since this category processes reported user content on a per-case basis, each report corresponds to an independent document set. Document parsing must strictly bind to the individual report ID to avoid mixing parsing results from different reports. Common processed documents use multiple formats, including text tickets, scanned documents, and chat record screenshots. Parsing must support these varied input types. The real-time business rhythm requires the parsing process to have low latency. It must also support batch parsing of multiple attachments within a single report, ensuring that each attachment’s chunking results bind to the corresponding report entry.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Matches the information density of single paragraphs in texts such as report tickets and appeal forms, avoiding overly fragmented chunks that cause semantic breaks or overly long chunks that impair retrieval |
chunkSplitter | Split by paragraph + title | Adapts to the structured layout of report tickets, preserving the semantic integrity of sections such as report reasons and processing records |
customPdfParse | Enable OCR mode | Adapts to scanned report attachments and image-format violation evidence, extracting text content from images |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Covers the parsing time of batch multi-attachment reports, preventing parsing failures due to timeouts |
UPLOAD_FILE_MAX_SIZE | 100 MB | Supports large-volume report attachments such as long video screenshots and multi-page scanned documents uploaded by users |
trainingType | Configure according to knowledge base data source | Adapts to the import of report data from external data sources such as third-party multidimensional tables, meeting multi-channel data parsing requirements |
Three common mistakes
- Phenomenon: After uploading multiple report attachments in batch, the parsing result only displays the content of a single file. Cause: Independent parsing binding per file was not configured, and multiple attachments were merged into the same document chunk, leading to mixing of attachments from different reports.
- Phenomenon: Knowledge base search tests fail to locate violation keywords within report tickets. Cause:
chunkSizeis set too large, causing the paragraph containing the target keyword to be truncated outside the chunk, or the chunk overlap ratio is too low to cover key information. - Phenomenon: No text content is returned in the parsed result of a scanned report PDF. Cause: OCR mode in
customPdfParsewas not enabled, and only native PDF text was attempted to be parsed, which cannot extract content from image-scanned documents.
How to confirm the configuration is correct
- Upload multiple attachments for a single report, confirm that the file names in the parsing list match those used during upload, verifying correct binding.
- Enter specific keywords from a report ticket, validate in the FastGPT v4.17.0 knowledge base search test that the corresponding chunked content is returned, and adjust chunking parameters until the expected result is achieved.
- Upload a scanned PDF attachment, check whether the parsed result includes text content extracted from the image, confirming that the OCR mode is active.
- Upload an attachment larger than 100 MB, verify that the upload and parsing processes do not trigger errors, confirming that the file size limit setting is appropriate.
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.