What data for this category looks like
This category covers complete manufacturing scenarios for air conditioners, refrigerators, washing machines and small home appliances. Daily documentation includes configured BOMs, assembly work instructions, refrigerant charging records, electrical safety test reports, performance spot check reports and repair records. Data sources include ERP/MES systems, component batch ledgers, tightening and charging equipment logs, safety compliance tester data, performance laboratory test results and after-sales fault records. Data updates are performed per individual product, per shift production batch or per spot check plan. Each individual document contains fields such as component model, process parameters, test values, batch number, etc. Some documents require matching specific units: for example, the unit for refrigerant charging volume is grams, and the unit for tightening torque is newton meters.
What constraints do these characteristics impose on the citation sources and traceability process
A single home appliance product corresponds to multiple types of associated documents such as configured BOMs, assembly work instructions and process records. Citation traceability requires precise binding of single-batch components with records from corresponding production and testing links, to prevent mixed-line misassembly of multiple models. The high-frequency data update rhythm requires the traceability system to support real-time association with latest data sources such as ERP/MES and equipment logs, to ensure that cited content matches the current production status. The diverse document types include highly constrained process parameters, which requires accurate matching of fields and units during traceability to avoid confusion between parameters such as refrigerant charging volume and tightening torque. The product attributes targeted at end consumers require that traceability results can be traced back to specific production shifts and batches, to meet after-sales verification and compliance requirements.
How to configure it
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Adapts to the token processing logic of FastGPT v4.17.0. The single-item length of core home appliance documentation falls within this range, which can fully cover key fragments of assembly work instructions and test reports |
responseData | Enabled | Must return traceable raw data for the technical document cross-review process, to support verification of content accuracy |
detail | Display source path and field extraction results | Must clearly mark core traceability information such as component batches and process parameters, to avoid review confusion between data of different product models |
| Maximum number of citation sources | Top 3–5 | The number of production and testing documents associated with a single home appliance is limited. Excessive citations will interfere with the focus of the review logic |
| Citation data cache duration | Calibrated based on actual measurements | Production data updates are performed per individual product, shift or spot check plan, which needs to adapt to real-time requirements |
Three common configuration mistakes
- Phenomenon: An empty
contentfield is returned after calling the FastGPT API, with no citation source related data. Cause: TheresponseDataconfiguration item is not enabled correctly, or the request does not carry correct context-related parameters. - Phenomenon: The original document link in the knowledge base cannot be accessed normally, and download fails. Cause: The nginx proxy path for knowledge base documents is not configured correctly, resulting in failure to read the original document content during traceability.
- Phenomenon: The cited document parameters do not match the actual production data, and confusion occurs between refrigerant charging volume and torque values. Cause: The field extraction rules for the
detailparameter are not configured, and the component batch and production link information corresponding to the document is not accurately matched.
How to confirm the configuration is correct
- Initiate a test request, check whether the
responseDatafield is included in the returned result, and confirm that the citation data switch has taken effect. - Access the returned original document link, verify that the corresponding document can be opened normally, and confirm that the nginx proxy configuration is correct.
- View the field content of the cited fragment, confirm that core traceability information such as component batches and process parameters has been extracted, and match the configuration rules of the
detailparameter. - Simulate the review scenario of multiple product models, check whether the citation sources are accurately bound to the documents of the corresponding products, to avoid confusion of mixed-line data.
The values provided here are common starting points and should be measured against the reader's own samples.