What This Category's Data Looks Like
The data for this category primarily comes from customer surface treatment specifications, bath test results, rectifier operation data, online water quality monitoring records, and hazardous waste management ledgers. Daily documents include bath analysis records, process parameter cards, hanger usage logs, film thickness and adhesion reports, hazardous waste ledgers, and wastewater monitoring records. Data updates follow per-batch, per-shift, or special monitoring requirements. Document fields include bath composition values, process temperature and current parameters, hanger numbers, film thickness units, emission monitoring indicators, and more. There is no unified fixed format, with slight variations based on each workshop's record-keeping habits.
Constraints Imposed on Retrieval and Recall
This category's data characteristics create multiple constraints for retrieval and recall. Data sources cover offline test records, online real-time monitoring data, and paper ledgers. Parsing and accurate retrieval of multi-format content must be supported. Update rhythms are inconsistent; some data refreshes per batch or per shift. The timeliness of recall results must match business operation nodes. Document fields include numerical values, parameters, and ledger content with no unified structure. Precise matching across multiple fields must be supported. Additionally, the data involves compliance-related content. Retrieval results must accurately link to corresponding business scenarios to avoid irrelevant matches.
Configuration Settings
The following configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | This category's process parameters and test data require high matching accuracy to avoid irrelevant bath composition or emission records interfering with retrieval results |
limit | Top 6–10 entries | A single equipment inspection or maintenance involves multiple types of parameters and records. Sufficient associated information is needed to support Q&A |
usingReRank | Enabled | This category's documents contain mixed numerical and text content. Reranking can improve the ranking priority of highly relevant results |
datasetSearchUsingExtensionQuery | Enabled | This category has many business terms. Extended queries can supplement process-related synonyms and parameter association expressions |
collectionIds | Configure document collections grouped by bath, process, and hazardous waste | This category's data is stored grouped by business scenario. Grouped configuration narrows the retrieval scope and improves retrieval efficiency |
Three Common Mistakes
- Issue: When calling the recall test API,
{"code":514,"statusText":"Invalid knowledge base ID"}is returned. Cause: ThecollectionIdsparameter is configured with a non-existent knowledge base collection ID, or the document collection exclusive to this category is not specified. - Issue: The number of results returned by knowledge base retrieval does not match the set
limit. Cause: The retrieval switch for irrelevant datasets is not turned off, or document chunking settings are unreasonable, causing valid results to be truncated. - Issue: The bath adjustment parameters returned by retrieval do not match actual business requirements. Cause: The
similaritythreshold is set incorrectly, causing low-matching historical records to be incorrectly recalled, ordatasetSearchUsingExtensionQueryis not enabled, so synonyms and parameter associations related to electroplating processes are not correctly identified.
How to Confirm Configurations Are Correct
- Call the recall test API, pass a typical inspection question for this category, and verify whether the associated documents in the returned results belong to the configured
collectionIdscollection. - Adjust the
similaritythreshold, observe changes in the relevance of returned results, and confirm the threshold meets the accuracy requirements of this category. - Compare the sorting of retrieval results when
usingReRankis enabled and disabled, to verify that the reranking configuration takes effect. - Check the number range of retrieval results, and confirm it matches the set
limit.
The values provided are common starting points and should be measured against the reader's own samples.