What the data for this category looks like
Process data for photovoltaic module encapsulation originates from the cell grading system, glass adhesive film backplane batch ledgers, factory MES systems, string welding and lamination production equipment, and EL/IV testers. Daily generated documents include layout drawings, material batch records, lamination process parameter records, EL defect images, IV power test reports, insulation voltage withstand records, and finished product packing lists. Data is updated in real time per module and per material batch. Each document includes fields such as batch number, process parameters, and test values, with units changing dynamically based on the test item.
What constraints these characteristics impose on the "workflow orchestration" link
Dispersed data sources and frequent updates require the workflow to support multiple data source access and real-time pulling. The data volume per component is small but has high batch association requirements, so the association between batches and components must be bound in variables. Non-structured data such as EL images and IV reports accounts for a large proportion, so the workflow needs to be configured with nodes to handle parsing of image and numerical data. The most time-consuming matching and judgment links must be split into independent nodes to avoid single-node timeouts. Errors directly affect customer claims, so the workflow needs to be configured with failure retry and exception alert nodes to ensure traceability of non-compliant processes.
How to set the configurations
The following configurations are based on FastGPT v4.17.0, adapted to the audit scenario requirements of photovoltaic module encapsulation.
| Configuration Item | Recommended Approach | Basis for This Approach |
|---|---|---|
| Data Source Binding | Access official APIs of MES, EL testers, and IV test systems | Match business data sources to enable real-time data pulling |
| Variable Association Rules | Bind material batches and test data in variables by component batch number | Ensure the correlation of full-process data for a single component and avoid matching errors |
| Node ID Configuration | Bind the nodeId of string welding, lamination, and test nodes | Ensure the execution order of workflow nodes conforms to production logic |
| Node Timeout Setting | 600 seconds | Adapt to the standard processing duration of string welding and lamination processes and avoid single-node timeout interruptions |
| Image Rendering Configuration | Enable direct URL rendering | Support online viewing of EL defect images and meet the visualization requirements of audit scenarios |
| Exception Retry Count | 3 times | Address small-probability anomalies such as temporary device disconnections and reduce manual intervention |
Three common mistakes
- Phenomenon: An image URL passed into a workflow node fails to display properly, and only plain text links are returned. Cause: The built-in image rendering configuration of FastGPT is not enabled, and only URL text is returned without triggering front-end rendering.
- Phenomenon: An error occurs when running the
localStorage.getItem('userId')code in a workflow node, with the promptlocalStorage is no. Cause: The code nodes of FastGPT workflows run in a server-side environment and do not support browser-side localStorage objects. - Phenomenon: No thinking process content is returned after an API initiates a workflow request. Cause: Fields containing the thinking process are not configured in the workflow's output node, or the corresponding return switch is not enabled.
How to confirm the configuration is complete
- Trigger a workflow test for a single batch of components, check whether the material records, EL images, and IV test data of the corresponding batch are automatically pulled.
- Pass a test image URL to confirm that the workflow node can render and display the image content normally.
- Simulate a device disconnection scenario to check whether the workflow performs retries according to the configured retry count.
- Call the API to initiate a workflow request and confirm that the returned result includes the configured output fields and thinking process.
The values provided are common starting points and should be measured against the reader's own samples.