What Data Looks Like for This Category
This category covers manufacturers of air conditioners, refrigerators, washing machines, and small home appliances. Daily documents include bill of materials (BOM), assembly work instructions, refrigerant charging records, electrical safety test reports, performance spot check reports, and rework 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 feedback data. Data updates follow per-unit product, shift-by-shift operations, or spot check schedules. Each data entry includes fields such as product model, batch number, process parameters, and test results. Some process data includes physical units.
Constraints on Knowledge Base Retrieval and Recall
This category has high data update frequency and fine-grained data granularity. Each product unit corresponds to multiple real-time production and test data entries. Retrieval and recall must quickly match the latest batch process and compliance information. In mixed-model assembly scenarios across multiple product lines, it is necessary to accurately distinguish exclusive BOM and process requirements for different home appliance models. This prevents cross-model confusion in retrieval results and avoids misassembly risks. Documents for safety compliance and performance spot checks have strict compliance requirements. Retrieval must prioritize high-confidence compliance data to ensure audits meet standards. After-sales rework records have high relevance to current production data. Cross-data source associated retrieval must be supported, while controlling the number of returned results to avoid interfering with the audit process.
Configuration Settings
The following configurations apply to FastGPT v4.17.0.
Choose a retrieval mode through representative-query testing; configure category, batch, or entity filters separately because the mode itself does not enforce those boundaries.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches precise fields such as home appliance model and process parameters, filters low-similarity mixed-model documents, and prevents mixed-line misassembly |
limit | Top 8 entries | Controls the number of documents returned per batch audit, adapts to rapid audit scenario requirements, and avoids information overload |
usingReRank | Enabled | Rearranges initial retrieval results by relevance, distinguishes process details of different models, and improves visibility of key information |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands associated data such as component batches and rework records based on audit questions, covering full-process audit dimensions |
collectionIds | Exclusive IDs corresponding to the supplier audit and process quality library for complete home appliance manufacturing | Limits retrieval scope, prevents mixing of documents from other categories or unrelated knowledge bases |
Three Common Configuration Errors
- Retrieval results return a large number of unrelated assembly work instructions for other home appliance models. The cause is incorrect configuration of
collectionIds, and failure to limit the retrieval scope to the knowledge base set for the current production model. This leads to retrieval of cross-model data. - Historical call logs for individual documents cannot be viewed. The cause is failure to enable knowledge base log collection configuration, or failure to associate the log query entry with the audit process. This makes it impossible to trace document usage.
- Duplicate refrigerant charging records appear in retrieval results. The cause is failure to enable document deduplication logic, or the presence of duplicate uploaded batch equipment log files in the data source, with no prior cleaning of duplicate data.
How to Confirm Configuration Is Correct
A mixed-line assembly audit question for a specific home appliance model is submitted. Retrieval results are reviewed to confirm they only include relevant documents for that model, with no unrelated content from other models. The knowledge base management page is accessed. Uploaded documents such as BOM and safety compliance reports are confirmed to be correctly associated with the specified collectionIds set. An audit process is triggered. Initial retrieval results are reviewed to confirm they have been rearranged, and the top 2 results are the core documents most relevant to the current question. The detail page of an individual document is accessed. Historical call records for that document are confirmed to be visible, verifying that log collection is enabled correctly.
The values provided here are common starting points, and should be evaluated against samples appropriate to the intended use case.