What the data for this category looks like
Daily documents for landing page conversion optimization include landing page plans, event tracking requirement lists, AB test reports, and form field configuration tables. Data sources include landing page access logs, heatmap data, form submission records, and ad source parameters. Update frequency is adjusted weekly or per single campaign cycle. Document structure mixes structured field lists and unstructured analysis descriptions. Fields include page dwell time, form submission count, ad channel identifiers, and other items. No fixed unified statistical value format exists.
What constraints these characteristics impose on the document parsing and chunking link
Documents for landing page conversion optimization mix structured fields and unstructured analysis content. Chunking must retain the contextual association between fields and their corresponding analysis. Individual fields from event tracking requirement lists or experimental group data from AB test reports must not be split. Documents updated weekly or per campaign cycle require chunking rules adapted to frequently iterated content. This avoids splitting updated content due to fixed chunk lengths. Mixed document formats from multiple sources require parsing tools compatible with the structural features of different document types. General one-size-fits-all character-based chunking logic cannot be used.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the single-chapter length of landing page plans and event tracking requirement lists, retaining contextual association between fields and analysis content |
chunkSplitter | Split by document structure | Adapts to the structured format of AB test reports and form field configuration lists, avoiding splitting associated information from individual event tracking fields |
customPdfParse | Enabled | Resolves the issue where embedded images cannot be viewed after PDF upload, retaining heatmap and experimental result chart elements within documents |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Adapts to the upload requirements for accumulated landing page logs and report documents across a campaign cycle |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Prevents parsing timeout for large AB test reports, ensuring complete parsing of all experimental data and analysis content |
chunkSettingMode | Manual configuration | Tailored to the exclusive document type for landing page conversion optimization, with custom chunking rules adapted to mixed-structure documents |
Three common errors
- Low knowledge base answer accuracy after uploading docx or excel documents. The symptom is missing fields or broken associations after parsing. The cause is failing to configure chunking rules according to document structure, splitting structured event tracking requirement lists into unrelated fragmented segments.
- Embedded images cannot be viewed after PDF upload. The symptom is loss of heatmap and chart elements within the parsed document. The cause is not enabling the
customPdfParseconfiguration item. - Large campaign report parsing times out. The symptom is timeout error returns or parsing task failure. The cause is not adjusting
PARSE_FILE_TIMEOUT_SECONDSto a value adapted to large documents.
How to confirm configurations are set correctly
- Users may upload a test landing page plan document and verify that parsed chunks retain the association between event tracking fields and corresponding plan descriptions.
- Users may upload a PDF document with embedded charts and confirm all image elements are viewable normally after parsing.
- Users may upload a large AB test report under 1000 MB and confirm no timeout errors occur during the parsing process.
- Users may access the configuration page of FastGPT v4.17.0 and confirm
chunkSettingModeis set to manual configuration, withchunkSizefalling within the 800–1200 character range.
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.