Data Profile for This Category
Data sources for this category include customer surface treatment specifications, bath chemistry test results, rectifier operation data, online water quality monitoring records, and hazardous waste management ledgers. Updates are performed per bath cycle, per shift, or at fixed monitoring intervals. Daily documents include bath analysis records, process parameter cards, fixture usage records, film thickness and adhesion reports, hazardous waste ledgers, and wastewater monitoring records. Fields cover process parameters, test values, and ledger record items. Units include g/L, ℃, A, μm, mg/L, and others.
Constraints for Knowledge Base Retrieval and Recall
The high-frequency update nature of this category requires the knowledge base to support incremental synchronization per bath or shift, to avoid delays caused by full re-imports. Mixed storage of multiple document types requires that during chunking, both the field integrity of structured parameter cards and the contextual coherence of unstructured ledgers are prioritized. Test data with specific units requires retrieval to match combinations of parameters and units, to reduce irrelevant results. Additionally, errors in inspection and audit processes can lead to batch rework or production shutdowns. Retrieved results must have high confidence levels to support rapid decision-making.
Configuration Settings
This configuration is based on the knowledge base feature of FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches unit-bearing process parameters and test data, filters low-correlation non-target content |
limit | Top 8–12 entries | Balances coverage of relevant documents per bath cycle and control of context length |
searchMode | embedding | Prioritizes matching semantically related process parameters and historical test records |
usingReRank | Enabled | Performs secondary ranking of retrieved results by relevance, elevates display priority of high-confidence results |
datasetSearchUsingExtensionQuery | Enabled | Expands relevant process terminology and unit expressions based on user questions, adapts to professional vocabulary for this category |
collectionIds | Specify document collection ID for the current workshop | Restricts retrieval scope to documents from the current production workshop, avoids interference from cross-workshop data |
Three Common Misconfigurations
- Phenomenon: Retrieval results do not include chapter title information from documents. Cause: The chapter title extraction configuration during chunking is not enabled, resulting in loss of contextual association information for structured documents, making accurate matching of the process chapter where parameters are located impossible.
- Phenomenon: Reply content includes reference links to knowledge base documents. Cause: The citation output option is not disabled in the retrieval configuration, causing unnecessary link information to be mixed into results.
- Phenomenon: Retrieval only matches core parameters in the main text, and does not associate auxiliary note data from bath analysis records. Cause: The
datasetSearchUsingExtensionQueryconfiguration is not enabled, resulting in failure to expand matching of semantically related auxiliary content, leading to missed associated information.
How to Verify Correct Configuration
- Upload a bath analysis record and process parameter card from the current workshop, initiate a retrieval request including "current bath temperature", and confirm returned results only include documents from the current workshop.
- Initiate a retrieval request, and confirm returned results do not include reference links to knowledge base documents, to verify output switch configuration is correct.
- Upload a fixture record in doc format, initiate a retrieval request including "fixture location", and confirm returned results include chapter title information from the document.
- Call the create document collection API and pass the
metadataparameter, initiate a retrieval request, and confirm returned results include the passed metadata content.
The configuration values listed are common starting points, and should be validated against applicable internal samples.