This page is compatible with FastGPT v4.17.0.
What this type of data looks like
Data sources for this category include ERP orders and inventory, BOMs and process routes, equipment capacity, staff shifts, work-in-progress, and supplier deliveries. Daily processed documents include sales orders, master production schedules, material requirements plans, weekly production scheduling sheets, work orders, material shortage lists, and delivery change notices. Data updates run daily, weekly, or trigger when orders are created or exceptions occur. Document fields include order numbers, material codes, capacity units, shift durations, work-in-progress quantities, and require linked information for processes, equipment, and personnel.
What constraints do these characteristics impose on the source citation and traceability link
Multi-source, irregularly updated data requires traceability systems to accurately bind source identifiers for each business document, to avoid mixing scheduling data from different batches. Diverse document structures require traceability configurations to adapt to extraction rules for different fields. For example, extract process information from work orders, and match material delivery status from supplier delivery documents. Cross-link business processes require traceability chains to cover the full path of order decomposition, scheduling adjustments, and delivery changes, to quickly locate the root cause of anomalies in quality audits. Audit scenarios also need to balance the completeness of traceability information and display simplicity, to avoid redundant data interfering with judgments.
How to configure settings
| Config Item | Suggested Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Adapts to retaining complete content of long documents such as master production schedules and material requirements plans, avoiding truncation of critical traceability fields |
responseData | Enable data source association tags | Matches information binding across multiple business sources including ERP orders, BOMs, and supplier deliveries, meeting audit traceability requirements |
detail | Enable full-link traceability | Records the full process of work order decomposition, scheduling adjustments, and delivery changes, covering all links in production planning |
| Maximum context recall count | Top 8 matching results | Covers multiple work orders and material requirements involved in a single scheduling run, calibrated based on actual testing |
| Citation display toggle | Manually configurable on/off | Adapts to information display requirements for audit scenarios, meeting user needs to adjust citation display |
Three common configuration mistakes
- Phenomenon: Reference markers are forcibly displayed at the end of knowledge base responses and cannot be removed. Cause: The default off option for the citation display toggle is not configured, failing to adapt to non-mandatory citation requirements in audit scenarios.
- Phenomenon: Input line breaks
\nare output as original text without formatted line breaks. Cause: Text parsing line break rendering configuration is not enabled, preventing documents with line breaks such as weekly production scheduling sheets and work orders from displaying correctly. - Phenomenon: The number of citations in traceability results exceeds the expected upper limit. Cause:
quoteMaxTokenor recall count is not adjusted reasonably based on the number of production planning documents, causing excessive unnecessary historical scheduling data to be included in the traceability scope.
How to verify successful configuration
- Enter a query containing a specific work order number, check if the reply associates corresponding data source tags such as ERP orders and material requirements plans.
- Test entering a query for a weekly production scheduling sheet with line breaks, confirm that line breaks in the reply are correctly rendered as paragraph separators.
- Adjust the configuration of the citation display toggle, verify that citation markers at the end of the reply can be turned on or off normally.
- Enter a batch query involving multiple work orders, check if the number of returned traceability results falls within the set recall count range.
The values provided here are common starting points and should be evaluated against the reader's own samples.