What the Data for This Category Looks Like
This category’s data covers complete home appliance manufacturing scenarios including air conditioners, refrigerators, washing machines, and small home appliances. Daily documents include configuration BOMs, assembly work instructions, refrigerant charging records, electrical safety test reports, performance spot check reports, and repair records. Data sources cover ERP/MES systems, component batch ledgers, tightening and charging equipment logs, safety tester data, performance laboratory test results, and after-sales fault archives. Updates run per individual product, per shift production batch, or per spot check plan. Each data entry includes fields such as component model, process parameters, test results, and batch number, with some fields carrying unit identifiers.
What Constraints Do These Characteristics Impose on the Vector Models and Indexing Workflow
Individual home appliance products correspond to multi-dimensional process and test data, with high update frequency. This requires indexes to support high-frequency incremental updates, avoiding excessive computing resource usage from full reconstruction. In mixed-line assembly scenarios, BOMs and process documents for different appliance models have subtle differences. Vector models must have highly discriminative semantic encoding capabilities to avoid confusing data from similar models. Safety and performance test data include both numerical parameters and qualitative descriptions. Indexes must support vector mapping for both structured numerical data and unstructured text, ensuring retrieval can associate process and compliance results. Bidding and quotation work requires association with historical batch data. Indexes must retain association fields for batches and product models, to avoid cross-model matching errors during retrieval.
How to Configure the Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Adapt to the paragraph length of home appliance BOMs and process documents, avoiding single chunks containing cross-model data |
| Overlap Length | 100–200 characters | Cover connecting content of process steps, preventing loss of assembly process association information after splitting |
maxToken | 8192 | Adapt to the long text content of safety test reports, ensuring complete encoding of a single test document |
| Similarity Threshold | 0.75–0.85 | Distinguish process parameters of different models in the same category, avoiding mixed-line matching errors |
| Single Search Limit | Top 10 entries | Balance retrieval efficiency and coverage of historical data required for bidding and quotation |
defaultToken | 1500 | Adapt to short text encoding of repair records, reducing redundant calculations |
Three Common Mistakes
- Phenomenon: The dataset.csv exported from the generated knowledge base only includes the index field, with no content field. Cause: Source data fields of home appliance complete machine documents such as assembly work instructions and test reports were not correctly associated, and only index metadata was imported.
- Phenomenon: The custom index cannot be associated with correct home appliance document data, and retrieval results are empty. Cause: The field mapping between the FastGPT index and the corresponding MongoDB table was not correctly configured, and the text field of the source document was not bound.
- Phenomenon: Non-target model home appliance process data is mixed into vector retrieval results. Cause: The similarity threshold was set too low, making it impossible to distinguish subtle parameter differences between different models in the same category.
How to Confirm the Configuration Is Correct
- Manually import a home appliance configuration BOM document, check whether the vector generation result includes semantic encoding of core fields such as component model and process parameters.
- Submit a retrieval request containing "air conditioner refrigerant charging parameters", verify that returned results only associate air conditioner-related documents, with no non-target category data such as refrigerators or small home appliances.
- View the index logs of FastGPT v4.17.0, confirm that incremental update tasks run per individual product and per shift batch, with no abnormal triggers for full reconstruction.
- Export the dataset.csv of the knowledge base, check that both index and content fields are included, and that the content field contains complete text content of the source document.
The values given are common starting points and should be measured against the reader's own samples.