What this category’s data looks like
Household appliance complete machine manufacturing data for workmanship and standards includes documents such as assembly work instructions, refrigerant charging records, electrical safety test reports, performance spot check reports, and rework records. Data sources cover ERP/MES systems, component batch ledgers, tightening and charging equipment logs, safety tester data, performance laboratory reports, and after-sales fault records. Data updates run per individual product, per shift production batch, or per spot check plan. Each document includes fields such as model identifiers, process parameters, test thresholds, and batch numbers. Units include pressure values, torque values, energy efficiency grades, and more.
What Constraints These Characteristics Impose on Knowledge Base Retrieval and Recall
The mixed-model assembly production mode requires retrieval to accurately distinguish process parameters for different models in the same category, to avoid risk of incorrect assembly due to mixed lines. High-frequency data updates require the knowledge base to support real-time or near-real-time incremental synchronization, to adapt to document changes per shift or per individual product. Multi-source heterogeneous data formats, covering device logs, test reports, after-sales records and other types, require retrieval to support both structured fields and unstructured text. In scenarios with tight production line takt times, retrieval response must match production rhythm to avoid delays affecting process flow. At the same time, strictly control the confidence of recall results to prevent incorrect process parameters from triggering production risks such as electric leakage, refrigerant leakage, or fire.
How to Set Configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Household appliance process parameters have high precision requirements. Filter low-match irrelevant documents to avoid mixed-line assembly errors |
limit | Top 8–12 results | A single product has many process steps. Cover sufficient process parameters while controlling context length to fit model constraints |
searchMode | mixedRecall | Balance semantic matching and exact matching of structured fields such as model and batch number, to adapt to multi-source heterogeneous data |
usingReRank | Enabled | Optimize the ranking of recall results, prioritize safety and assembly parameters most relevant to the current workmanship issue, and reduce mismatching probability |
datasetSearchUsingExtensionQuery | Enabled | Automatically expand synonyms related to workmanship to cover expression habits of different positions, improving recall breadth |
collectionIds | Bind knowledge base collections corresponding to target household appliance models | Limit retrieval scope to avoid interference from process parameters of other models, improving retrieval accuracy |
Three Common Mistakes
- Symptom: The re-ranking model fails to run normally, and the interface shows model initialization failure. Cause: Deployment parameters for the re-ranking model were not configured correctly, causing an error during model loading.
- Symptom: Knowledge base retrieval results are incomplete, failing to cover all necessary parameters for the current workmanship issue. Cause: The similarity threshold was set too high, or the question expansion configuration was not enabled, causing some relevant documents to not be recalled.
- Symptom: Content returned after knowledge base search in the workflow does not strictly match the preset workmanship question. Cause: The retrieval scope of the knowledge base collection was not limited, and process documents for other household appliance models were included.
How to Confirm Correct Configuration
- Log in to the FastGPT v4.17.0 knowledge base management page, and check whether the values of configuration items such as
similarityandlimitmatch the preset plan. - Initiate a test query for a typical workmanship issue, and check whether the number and ranking of recall results meet the configuration requirements.
- Verify that the re-ranking model loads normally, and confirm that the ranking logic of retrieval results meets expectations.
- Initiate a test query for cross-model household appliances, and confirm that retrieval results do not include process documents for non-target models.
The values provided are common starting points and should be measured against the reader's own samples.