What Data Looks Like for This Category
This category covers daily documents of air conditioner, refrigerator, washing machine and small home appliance manufacturers. Documents include configured BOMs, assembly work instructions, refrigerant charging records, electrical safety test reports, performance sampling test reports and repair records. Data sources include ERP/MES systems, component batch ledgers, tightening and charging equipment, safety compliance testers, performance laboratory data and after-sales fault records. Updates are performed per single product, per shift operation or sampling plan. Documents contain fields such as process parameters, test results and batch information. Units vary by document type, and there is no unified fixed format.
Constraints Imposed by These Characteristics on Source Citation and Traceability
High-frequency updated data for single products requires traceability to be accurately bound to corresponding batches and processes, to avoid calling outdated process information across products. Mixed-line assembly production scenarios require strict distinction between process documents for different home appliance models, to prevent mismatched assembly rules and testing standards. Multi-source, heterogeneous data sources require traceability links to cover field mapping across different systems, to avoid confusion between direct equipment-collected data and laboratory reports. Safety and performance testing have strong compliance requirements, so cited content must fully include test items, results and test subject information to ensure traceability is verifiable. The update frequency per shift and per single product requires rapid location of the latest process and exception records during retrieval, to avoid using outdated data that does not meet current production requirements.
How to Configure the Setup
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Matches the core content length of a single process document or exception report, to avoid truncating key parameters and test results |
responseData | Return only associated document metadata and original text fragments | Adapts to the need for tracing batches, equipment numbers and test times in home appliance manufacturing scenarios, reducing redundant information |
| Similarity Threshold | 0.75–0.85 | Filters similar process documents for different home appliance models in mixed-line production, to avoid incorrect use of assembly rules |
| Single Search Limit | Top 6 results | Covers data from multiple links including assembly, testing and repair involved in a single product, to avoid excessive results interfering with judgment |
detail | Enabled | Recalls full fields of documents, including key information such as component batches, tightening torque, refrigerant charging volume |
Three Common Configuration Errors
- Phenomenon: Search results include irrelevant documents not associated with the target knowledge base, showing unexpected citation sources. Cause: The configuration to only retrieve the specified knowledge base is not enabled, leading to cross-knowledge-base recall of documents from non-current production batches.
- Phenomenon: The number of citations returned by retrieval is too high or too low, or documents with overly low relevance are included in the results. Cause: The single search limit and similarity threshold are not reasonably set, leading to recall of redundant data or omission of key process information.
- Phenomenon: Parameter parsing errors occur when calling knowledge base-related variables in the application, while directly writing the knowledge base ID or reference variables shows no abnormalities. Cause: The knowledge base association is not configured according to the variable binding format required by the platform, leading to incorrect parameter parsing.
How to Verify Successful Configuration
- Upload the assembly work instruction and corresponding repair record for a single home appliance, initiate a retrieval, and check whether the returned citations only include metadata and original text fragments of the target document.
- Adjust the similarity threshold, verify the changes in the number and relevance of recalled documents under different thresholds, and confirm that it meets the differentiation needs of current mixed-line production.
- After configuring the
quoteMaxTokenparameter, initiate a retrieval, and check whether the returned citation content has not truncated key process parameters and test results. - Call the application variable bound to the target knowledge base, initiate a test, and confirm that there are no parameter parsing errors.
The values provided here are common starting points and should be measured against the reader's own samples.