What this category’s data looks like
This category processes data sourced from the national medical insurance information platform, hospital HIS and medical insurance interfaces, medical insurance catalog and coding database, and cross-regional medical settlement platform, with daily updates. Daily processed documents include medical insurance settlement lists, expense details, medical record homepages, prescriptions, rejection lists, and settlement reconciliation statements. Core fields include diagnosis and treatment project codes, expense amounts, insured person identifiers, settlement status, etc. Coding fields must match the national unified medical insurance catalog specifications. Amounts are denominated in yuan. A single settlement data set includes tens to hundreds of detail items.
What constraints these characteristics impose on the "workflow orchestration" link
The daily data update rhythm requires the workflow to be configured with a scheduled trigger mechanism that matches the upstream data synchronization cycle. Data from multiple sources needs to connect to different interface nodes, and structured data returned by each interface must be passed via variables. The scale of batch-processed settlement data is not fixed, so the number of nodes running in a single batch must be limited to avoid resource overload. For the core links of matching diagnosis and treatment project codes, reviewing expense consistency with medical records, document fields must be mapped to variables recognizable by the workflow, and a failure retry link must be configured to handle interface fluctuations or coding matching failures. Unlike fund supervision audits, this workflow focuses on daily settlement audit before and after payment processing, and does not need to cover full fund risk investigation links.
How to Set the Configuration
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Maximum number of runs for batch execution nodes | 1000 times | Matches the platform's default batch processing limit, adapts to the scale of daily settlement data per batch |
| Number of data items loaded per single run | 100–300 items | Balances data processing efficiency and interface call frequency, adapts to the daily update batch rhythm |
variables field mapping configuration | Pass an array in the order of settlement document fields | Corresponds to the fixed field structure of daily documents, ensures accurate data transfer |
nodeId trigger node configuration | Bind by data update timestamp | Ensures the workflow triggers only after upstream data synchronization is complete |
| Workflow timeout threshold | 600 seconds | Matches the average processing duration of a single batch audit, avoids unnecessary waiting |
| Number of retries | 2 times | Addresses temporary interface fluctuations, reduces the probability of single processing failure |
Three Common Misconfigurations
- Node connections appear normal but the workflow cannot proceed to the next step. The interface shows the node status as pending execution with no ongoing progress updates. This occurs because the loop traversal logic for
variablesbatch data is not configured, and only a single settlement data item is processed, preventing subsequent nodes from triggering. - The global variable string array assignment format is incorrect, resulting in a "field not found" error in logs and null values in settlement audit results. This occurs because parameters are not passed in the standard array format
["settlement order number", "expense amount", "insured person ID"]. - Variable values output after calling the FastGPT API show content overlap. For example, the AI-generated rejection appeal reply B is spliced with the previous audit result A. This occurs because the conversation context variable is not reset in
variables, causing historical conversation content to be repeatedly brought into the workflow.
How to Verify the Configuration Is Correct
- Trigger a single batch test workflow, review the running logs of the batch nodes, and confirm that settlement data is split in accordance with the set 100–300 items per single load limit.
- Export the global variable assignment logs, and verify that the array format matches the field order of the settlement documents.
- Simulate a temporary timeout scenario for the upstream interface, and verify whether the workflow triggers retries in accordance with the set 2 retry times.
- Review the final running results of the workflow, confirm that no timeout errors are present and each audit node completes execution in sequence.
The values provided here are common starting points, and should be evaluated 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.