What the data for this scenario looks like
Data for channel attribution and lead distribution scenarios comes from advertising platforms, CRM systems, attribution SDKs, and form systems. Commonly processed documents include attribution rule documents, lead distribution spreadsheets, channel traffic reports, and abnormal lead checklists. Most documents use structured tables or text with fixed fields. These documents contain fields such as channel identifier, lead source identifier, conversion time, and assigned sales ID. Update frequencies are typically daily or weekly.
Constraints for document parsing and chunking
Retain field integrity when chunking structured table documents. Do not split content that spans table rows or columns. Splitting this content breaks key information such as channel parameters and attribution relationships. Split text documents with fixed fields along field boundaries. Do not mix attribution information from different leads. High-frequency daily or weekly reports require fast parsing speeds. Overly long report files may exceed parsing time limits. For attribution rule documents with nested logic, retain rule context during chunking. Do not break the association between conditions and execution actions.
Configuration Settings
This configuration applies to FastGPT v4.17.0. Use the following parameter settings:
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Fits the single-table block size of channel reports and lead distribution spreadsheets, retaining field association integrity |
chunkSplitter | Split by table delimiters | For structured documents, avoids splitting cross-row/column table content and preserves field associations |
customPdfParse | Enabled | Adapts to nested logic and fixed layout in attribution rule PDFs, improving parsing accuracy |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Supports uploading ~10MB report files, preventing parsing failures due to file size limits |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Fits the parsing duration of large channel reports, preventing timeout errors |
chunkSettingMode | Manual configuration | Adjust chunking parameters for different types of attribution documents, adapting to the fixed structure of business documents |
Three Common Configuration Mistakes
- Uploading ~10MB PDF files triggers a timeout error. This occurs when the
PARSE_FILE_TIMEOUT_SECONDSparameter is not adjusted. The default parsing duration is insufficient for processing large report files. - Some DOCX documents fail to parse correctly. This happens when the
customPdfParseconfiguration is not enabled. The configuration cannot handle special layouts or nested table structures in DOCX files. - Tables in knowledge base documents are not chunked correctly. This occurs when
chunkSplitteris not set to split by table delimiters. This splits cross-row/column table content and loses field association relationships.
How to Verify Proper Configuration
- Upload a single ~10MB channel report file. Confirm the parsing task completes within 600 seconds with no timeout errors.
- Upload an attribution rule document with nested tables. Confirm parsed chunked content retains full table structure with no broken fields.
- Import a preset Q&A knowledge base. Send an API Q&A request. Confirm returned results match the preset original answers from the knowledge base.
- View document parsing logs. Confirm chunked content aligns with the original document’s field correspondence with no information loss.
The values provided are common starting points and should be evaluated against the user’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.