What Data for This Category Looks Like
Data sources for this category include cell capacity grading data, factory MES systems, welding equipment logs, torque tool records, BMS software version files, and EOL test bench data. Data is generated once per battery pack or energy storage cabinet, and updated synchronously each time the BMS software version changes. Daily documents include grouping rules, BOM lists, welding and torque operation records, BMS flashing logs, insulation and voltage withstand reports, EOL test reports, and system configuration lists. Data fields cover cell capacity, torque values, test voltage, software version numbers, and other metrics, with no unified cross-device standardized units.
Constraints on Multi-turn Dialogue and Prompt Engineering
The data characteristics of this category impose multiple constraints on multi-turn dialogue and prompt engineering workflows. First, each individual device has exclusive data. Multi-turn dialogue must bind the full historical records of a single battery pack or energy storage cabinet using chatId to prevent cross-device data mixing. Second, document types are diverse and fields are dispersed. Prompts must explicitly define the range of document types to retrieve, such as only calling grouping rules or BMS flashing logs for the target device. Third, data updates are frequent and sources are scattered. Prompts must restrict retrieval to the latest version of data matching the current device, while multi-turn dialogue must retain context like the device’s BMS version and test parameters to ensure Q&A relies on the device’s current configuration. Fourth, equipment failure can have severe consequences. Multi-turn dialogue must validate input parameter compliance in each round, such as confirming torque values meet grouping requirements for the target model.
How to Configure Settings
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
chatId | Bind to the unique device code of a single device | Ensure multi-turn dialogue is linked to the exclusive historical data of the corresponding device, avoiding confusion across devices |
messages | Retain 3 rounds of historical dialogue and the latest input for the current device | Control context length to fit the limited historical data volume of a single device |
variables | Pass in the current device’s BMS version, cell model, and test date | Allow prompt engineering to accurately match exclusive configuration documents for the corresponding device, and clarify the scope of template calls |
| Question Optimization | Enable and set to optimize only for equipment maintenance scenarios | Adapt to the professional maintenance Q&A needs of this category |
| Maximum Context | 8000 token | Cover core documents and historical dialogue content for a single device, calibrated via actual testing |
| Single Retrieval Limit | Top 6 matching documents | Match the retrieval needs of multiple types of documents in this category, avoiding interference from redundant data |
This configuration is based on FastGPT v4.17.0.
Three Common Misconfigurations
- Specifying a knowledge base tag when calling the dialogue interface fails to take effect. The symptom is that returned results include irrelevant documents. The root cause is not passing the corresponding tag parameter in
variables. - Calling the interface after configuring prompts returns a 400 status code. The symptom is a parameter verification failure prompt from the interface. The root cause is failing to correctly bind device-exclusive fields corresponding to
variablesin the prompt template, such as omitting the BMS version parameter reference. - Context is lost during multi-turn dialogue. The symptom is that subsequent dialogue cannot access previously retrieved device test data. The root cause is failure to correctly pass
chatId, resulting in failed session binding.
How to Verify Proper Configuration
- Initiate a test dialogue, enter the model of a single device, and check if returned results include exclusive document content for that device.
- Review the
messagesfield returned by the interface to confirm it contains the latest 3 rounds of historical dialogue records. - Pass preset
variablesparameters, and verify that the prompt correctly calls configuration information for the target device. - Initiate 3 or more consecutive dialogue sessions for the same device, and confirm that session data bound to
chatIdis not lost.
The values provided are common starting points and should be measured against the reader's own samples.