What the data for this category looks like
Routine documentation for photovoltaic module encapsulation includes layout diagrams, material batch records, lamination process records, EL images, IV test reports, insulation voltage withstand records, and packing lists. Data sources include battery cell grading, glass adhesive film and backsheet batches, MES systems, string welding and lamination equipment, and EL/IV testers. The update frequency is once per module and per batch of materials. Document fields associated with a single module include battery cell current gear, lamination temperature and time, EL crack coordinates, IV test peak power, insulation voltage withstand value, and others. Corresponding units include ampere, degrees Celsius, second, watt, kilovolt, and additional relevant units.
What constraints these characteristics impose on the "tool calling and plugins" workflow
High-frequency, small-batch data updates for each module require tool calling to support lightweight batch processing, to avoid excessively long single-call execution times. The need to handle mixed document types requires tools to support both structured text extraction and multimodal image parsing. The requirement for standardized dedicated fields and units means that tool calling parameter formats must strictly match the document fields used in photovoltaic module encapsulation. Failure to do so prevents accurate extraction of valid information. The need to connect across multiple data sources requires plugins to adapt to the interface protocols of different devices and systems, to ensure stable data acquisition.
How to configure the setup
This configuration applies to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Workflow/API tool selection | Specified as ["extractField", "validateCompliance"] | Accurately matches the document review requirements for photovoltaic module encapsulation, and calls two types of tools: field extraction and compliance verification |
| Workflow/API function-call configuration | Force return results in JSON format, with fields including dedicated parameters such as battery cell current and lamination temperature | Standardizes output results to facilitate subsequent field comparison and verification for cross-review |
| Timeout period | 60 seconds | Processing time for documents associated with a single module is relatively short, avoiding prolonged occupation of calling resources |
| Tool call retry count | 2 times | Interface fluctuations of device data sources occur occasionally; retries reduce the probability of call failures |
| Multimodal tool enablement | Enabled | Required to process unstructured images and mixed document content such as EL images and IV test reports |
Three Common Mistakes
- When calling the text extraction tool, the prompt "Field extraction failed" appears. The cause is that the JSON Schema for
functionCallhas not been configured for the dedicated document fields of photovoltaic module encapsulation, so the tool cannot recognize specific parameters such as battery cell current and lamination temperature. - When using a model connected via OneAPI, empty results are returned after entering the API key. The cause is that the model's calling address and request headers have not been correctly configured in FastGPT, and the token verification rules of OneAPI have not been adapted.
- After publishing a workflow interface with global variables, external calls cannot pass global variables. The cause is that the global variable pass-through switch has not been enabled in the interface configuration, and the parameter format for variable reception has not been specified.
How to Confirm the Configuration is Correct
- Upload the documents associated with a single module, and call the tool to verify whether the preset dedicated fields can be accurately extracted.
- After connecting the model API key via OneAPI, initiate a small-batch test call to confirm that the returned results are not empty and meet expectations.
- Publish a workflow interface with global variables, use an external calling tool to pass test variables, and verify that the variables are read normally.
- Enable the multimodal tool, upload an EL test image, and verify that the tool can recognize and parse the image-based document content.
The values provided here are common starting points and should be measured against the reader's own samples.