What this category of data looks like
Data sources for this category include the national medical insurance information platform, hospital HIS and medical insurance interfaces, medical insurance catalog coding database, and cross-regional medical treatment settlement platform. Data updates occur daily. Daily processed documents include medical insurance settlement lists, expense details, medical record homepages, prescriptions, rejection lists, and settlement statements. Data fields include diagnosis and treatment item codes, drug and consumable codes, expense amounts, settlement order numbers, insured person identity identifiers, and medical institution information. No additional custom statistical fields are included.
What constraints these characteristics impose on the "workflow orchestration" link
Decentralized data sources with daily updates require workflows to support batch data retrieval from multiple interfaces. This prevents single-request timeouts. Diverse document types and strong coding dependencies on fields require adding field validation links during orchestration. This ensures coding matching rules take effect. Audit logic differs between single and bulk settlement data. Branch nodes are needed to distinguish between single-item and bulk tasks. The reconciliation link for rejection lists and settlement statements requires strict field alignment. This mandates consistent data transfer formats between workflow nodes.
How to set configurations
| Configuration Item | Recommended Practice | Rationale |
|---|---|---|
variables | Preset settlement order number and medical insurance catalog database ID as global variables | Used to transfer fixed business parameters across nodes and avoid repeated configuration |
| Branch node trigger condition | Branch by document type (settlement/rejection/reconciliation) | Distinguish different audit logic and adapt to multi-type document processing |
| Knowledge base search node | Call the medical insurance catalog knowledge base, with a maximum of 10 results per search | Match diagnosis and treatment item and drug/consumable codes, control return result volume |
nodeId | Bind the IDs of the text splicing node and AI question answering node | Ensure correct data transfer links between nodes and avoid parameter loss |
| Single node timeout setting | Set to 600 seconds | Adapt to time consumption requirements of multi-interface retrieval and bulk data processing |
| Text splicing node | Splice settlement details and prescription content, with a splicing limit of 800–1200 characters | Adapt to the context length limit of FastGPT v4.17.0's AI question answering node and avoid content overflow |
Three common mistakes
- Symptom: The AI question answering node cannot receive the complete content passed from the text splicing node, or prompts an input format error. Cause: The output of the text splicing node is not correctly bound to the input parameters of the AI question answering node, and the global transfer link is not configured via
variables. - Symptom: The knowledge base search node cannot use the preset medical insurance catalog knowledge base ID for retrieval. Cause: The knowledge base ID is not configured as a global
variables, and the corresponding parameter is not associated in the node settings. - Symptom: The streaming output of the custom model call node does not return in segments, or only returns a single segment of results. Cause: The code running module does not follow the streaming data format requirements of FastGPT, and the output trigger logic is not correctly configured.
How to confirm the configuration is complete
Run test data to trigger the workflow. Check if input and output fields of each node match the preset variables parameters. View the return results of the knowledge base search node. Confirm that the target medical insurance catalog knowledge base is called, and the number of return results meets the configured per-search limit. Run the link including text splicing and AI question answering. Check if the AI question answering node correctly receives the complete spliced content. Test the custom model call node. Confirm that streaming output returns in segments with no content truncation or loss.
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.