What this category of data looks like
Daily documents for casting (sand casting, precision casting, die casting) include casting process cards, batching lists, heat records, mold history records, first article inspection records, material test reports, and nonconforming product review forms. Data sources include ERP material master data, melting furnaces and thermometers, spectrometers, MES heat records, mold management systems, customer drawings, and technical requirements. Data is updated per heat, per batch, and daily. Each document is linked to a unique heat number, mold number, or batch number. Fields include alloy batching ratio values, melting temperature, pouring parameters, defect types, and other details. Structures vary based on document type.
Constraints on citation sources and traceability workflows
Data for this category is updated frequently per heat and batch. Traceability requires precise association with a single heat or batch. Cited passages must include unique identifier fields to avoid cross-batch confusion. Multiple document types have distinct field structures. Source document types must be differentiated during traceability to ensure cited content matches query requirements. Frequently updated data requires cited sources to use the latest version. A real-time synchronization mechanism must be configured to prevent citing expired data. Errors may result in full heat or batch scrap, or customer claims. Citation accuracy directly impacts the reliability of process judgments. The scope and accuracy of cited content must be strictly controlled.
Configuration Settings
The following configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 1200–1800 token | Matches the typical length of a single passage from casting process documents, preventing cited content from exceeding the reply capacity limit |
responseData | Enabled | Meets the needs of process personnel to view traceability basis |
detail | Enabled | Displays unique identifiers and source types of documents, facilitating precise tracing of heat and mold numbers |
| Single recall count | Top 3–5 entries | Controls the amount of reply information, avoiding redundant content interfering with process judgment |
| Document synchronization frequency | Synchronize once per hour | Matches the frequent update rhythm of this category, ensuring cited data is the latest version |
Three Common Configuration Errors
- The console returns the
quote type errorerror. The referenced variable format does not meet system requirements, such as failing to correctly specify the unique identifier field of the citation source. - Knowledge base citation links always appear in replies and cannot be hidden. The
responseDataswitch status is not configured correctly, causing cited content to be displayed forcibly. - When the knowledge base chunk size is set to 5000 token and
quoteMaxTokenis set to 1500, sliced content exceeding the length limit is still recalled. The chunk size and citation limit do not match, and no length truncation is applied to recalled slices.
How to Verify Configurations Are Correct
- Initiate a query that includes alloy batching ratios and melting temperatures. Check if the reply includes unique identifiers such as heat numbers and mold numbers.
- Test the hidden citation feature. Confirm that knowledge base citation addresses or source information no longer appear in replies.
- Adjust the value of
quoteMaxToken. Verify that the length of returned cited content matches the set threshold. - Check system logs. Confirm that no
quote type errormessages are present.
The values provided are common starting points and should be measured against the reader's own samples.