What this category of data looks like
Data for the campaign setup scenario originates primarily from advertising platform backends, budget scheduling sheets, attribution tools, and user profile documents. Daily processed documents include campaign outlines, target audience settings, daily budget amounts, conversion tracking configurations, and other content. Some documents are structured tables, while others are unstructured explanatory text. Data updates occur daily or when a single campaign launches. Fields include target package ID, budget unit, conversion event type, and more. There is no unified fixed format, so support for multiple source file types is required.
What constraints these characteristics impose on the document parsing and chunking stage
Documents in the campaign setup scenario mix structured and unstructured content. The parsing process must retain both table fields and associated text information, avoiding separation of linked configurations such as target settings and budget allocation. Data updates happen at a high frequency, so parsing efficiency must align with business timelines to prevent delays to campaign launch progress from parsing lags. Multiple source file types require parsing tools to support Excel, PDF, plain text, and other formats. The tool must also handle field differences across documents, while controlling parsing timeouts and file size limits to avoid campaign workflow interruptions caused by oversized single files or overly long parsing times.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Aligns with the field length of campaign setup documents, avoiding splitting linked content such as target settings and budget schedules |
chunkSplitter | Auto-split by document structure | Adapts to campaign setup documents that mix structured tables and unstructured explanatory text, preserving content associations |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Matches the parsing time of large budget scheduling sheets, preventing parsing interruptions due to timeouts |
customPdfParse | Enabled | Retains table structures and field information in campaign setup PDFs, avoiding loss of targeting parameters |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports bulk import of multiple annual campaign setup documents, adapting to bulk configuration needs |
Three common mistakes
- Symptom: After uploading an Excel budget sheet, the parsed result only retains the header row and no specific budget values. Cause: Adaptation configuration for structured tables was not enabled, causing table content in mixed-format documents to be skipped.
- Symptom: Parsing a large campaign setup PDF triggers a timeout error. Cause:
PARSE_FILE_TIMEOUT_SECONDSwas set to 300 seconds, which does not match the actual parsing time of large files. - Symptom: Chunked content retrieved by the knowledge base lacks complete target audience configurations. Cause:
chunkSizewas set to 500 characters, splitting complete target setting configurations into multiple independent chunks, causing linked information to break.
How to confirm configurations are set correctly
- Upload a standard campaign setup Excel file, and verify whether the parsed chunked content includes complete budget amounts and target audience parameters
- Upload a campaign setup PDF under 100 MB, and confirm that parsing completes within the duration set by
PARSE_FILE_TIMEOUT_SECONDS - Upload a file larger than
UPLOAD_FILE_MAX_SIZEfor testing, and confirm the system displays a compliant file size limit exceeded prompt - Access the knowledge base parsing logs for FastGPT v4.17.0, and confirm that the
chunkSplittergenerates the expected number of chunks based on document structure
The values provided are common starting points and should be validated against in-house sample data.
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.