Reference only; verify against the target FastGPT release and deployment documentation.
What this category of data looks like
Pharmaceutical distribution business data primarily comes from four categories: GSP compliance documents, purchase inbound and outbound orders, pharmaceutical circulation traceability reports, and internal business process manuals. Data sources include local PDF files, Excel spreadsheets, publicly shared Yuque pages, and files exported from internal enterprise systems. Update cadence varies with business nodes: purchase orders update in real time as transactions are generated, compliance documents are revised quarterly per regulatory requirements, and traceability reports are updated incrementally as pharmaceuticals move through circulation channels. Document structures include long-form compliance explanations, tabular business inventories, and process guides with embedded images. Fields include generic drug name, approval number, batch number, circulation nodes, and processing time limits. Common units include box, case, batch, and calendar day.
What constraints do these characteristics impose on document parsing and chunking
Multi-source data formats require parsing tools to support PDF, Excel, and web links. A single parsing configuration cannot cover all scenarios. Frequently updated business documents require fast synchronization to avoid parsing delays that impact the timeliness of business process guidance. Long-form compliance documents and tabular orders coexist. Chunking must avoid splitting critical business fields such as drug batch numbers and processing time limits, otherwise information breaks will occur in process guidance. Process guides with embedded images require image extraction support, otherwise visual steps for business processing cannot be fully conveyed.
How to set the configuration
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the long-text and tabular structure of pharmaceutical distribution documents, preserves the integrity of business fields |
chunkSplitter | Split by paragraph + heading | Matches the step headings and paragraph divisions of business process documents, avoids truncating process logic |
customPdfParse | Enable OCR mode | Adapts to scanned GSP compliance documents, extracts embedded images and printed text content |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Supports batch upload of large-volume circulation traceability documents, avoids parsing timeout interruptions |
chunkSettingMode | Manual configuration | Allows adjustment of chunking parameters for different document types such as compliance manuals and purchase orders |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Compatible with batch business data files exported by enterprises |
Three common mistakes
- Phenomenon: After uploading a publicly shared Yuque link, the system prompts "Unable to retrieve web content". Cause: The Yuque link was not confirmed to be fully publicly accessible. Internal business process documents for pharmaceutical distribution often set department-level permissions, which prevents parsing requests from retrieving content.
- Phenomenon: Parsed scanned GSP documents only display text, without embedded process diagrams. Cause: The OCR image extraction configuration for
customPdfParsewas not enabled. Scanned pharmaceutical circulation process guides cannot extract image information. - Phenomenon: When calling an HTTP business tool, the backend log shows successful parameter parsing, but the tool does not trigger the preset approval process. Cause: The chunked business fields were not bound to the tool's call parameters. The large model cannot obtain the correct business context to trigger the rule.
How to confirm the configuration is correct
- Upload a scanned GSP compliance document, check if the parsed result includes the pharmaceutical circulation process diagrams and printed text from the original document.
- Upload a batch purchase order Excel spreadsheet, verify that the chunked content retains complete drug batch numbers, circulation nodes, and processing time limit fields.
- Enter a publicly accessible Yuque business process link, confirm that the parsing task status is "Completed" and the content has no obvious missing parts.
- Trigger a tool call test, confirm that the backend log shows parameters were correctly passed and the tool execution records are normal.
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 2026-09-28.