What Data Looks Like for This Category
This category covers daily documents from process and quality departments of foundries, die casting workshops, and equipment component suppliers. Documents include casting process cards, batching lists, heat records, mold histories, first-piece inspection records, material reports, and nonconforming product review forms. Data sources include ERP material master data, melting furnace thermometers, spectrometers, MES heat records, mold management systems, and customer drawing technical specifications. Updates run per heat, per batch, or daily. Document fields include alloy composition, pouring temperature, mold cycle number, heat number, and other parameters with clear units.
What Constraints These Characteristics Impose on the "Source Citation and Traceability" Link
High document update frequency, triggered per heat or batch, requires traceability systems to quickly bind all data for a single heat or batch. Coexisting multi-source data requires systems to accurately match associations between melting parameters, mold histories, and quality inspection records. Fields include process parameters with units. Retaining original units and record numbers during citation avoids confusion between batching ratios of different heats. Exception reports must trace back to specific mold cycles and heats. Traceability links must associate cross-document information and clarify the corresponding relationship of each data source. Nonconforming product review forms must be linked to customer technical specifications. Cited customer drawing parameters must be accurate.
How to Set Up Configurations
The following configuration scheme applies to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Matches the single-section parameter length of heat records and process cards in casting documents |
| Similarity Threshold | 0.75–0.85 | Filters low-relevance melting parameter and mold data, avoids confusing records of the same model with different heats |
| Single Search Limit | Top 8 entries | Covers multiple documents linked to a single heat, including batching lists, furnace temperature records, and inspection reports |
quoteMaxToken | 1500–2000 token | Retains complete heat parameters and nonconforming product review details |
responseData | Enabled | Directly links original document fields, ensures traceability allows viewing original batching lists and spectral reports |
detail | Enabled | Displays full document fields, including key traceability information such as heat number and mold cycle number |
Three Common Mistakes
- Phenomenon: The answer to the second question in a sequence of consecutive queries is off-topic. Cited heat data is unrelated to the current question. Cause: Context association logic is not configured, or the
quoteMaxTokenvalue is too small to retain heat number parameters from prior questions. - Phenomenon: Retrieval returns no results. The corresponding heat batching list cannot be found. Cause: The similarity threshold is set too high. This filters documents that meet matching requirements but have misaligned fields.
- Phenomenon: Cited content is truncated. Complete alloy composition ratio data is missing. Cause: The
quoteMaxTokenvalue is smaller than the token count of a single heat record in the document, leading to content truncation.
How to Confirm Proper Configuration
- Import a single heat record document. Submit a query that includes the heat number. Verify that the citation result includes the complete batching list and temperature data for that heat.
- Submit consecutive queries. Inquire about parameters for different heats in sequence. Confirm that subsequent queries still correctly associate traceability data for the corresponding heat.
- Check that the citation module displays original document field information, including key parameters such as heat number, mold cycle number, and alloy composition.
- Adjust the similarity threshold. Verify that the matching accuracy of retrieval results meets business requirements.
The values provided here are common starting points. They should be evaluated against the reader's own samples.