What the data for this category looks like
Production and quality data sources for pulp and paper include wood pulp and waste paper batch ledgers, DCS/QCS control system data, on-machine measurement data, laboratory test reports, and energy and water treatment system records. Data update frequencies cover real-time scenarios such as on-machine quality curves, shift-based scenarios such as beating records, and per-roll scenarios such as sheet break and grade change records. Daily documents include pulp preparation sheets, beating records, paper machine operation logs, online quality curve charts, laboratory physical and chemical test reports, and sheet break and grade change ledgers. Fields include batch number, grammage, moisture content, thickness, pulp ratio, and other relevant items. Units include tons, square meters, percentage, grams per square meter, and others. Production uses high-speed continuous web mode, so single run records and quality curves have relatively long content.
Constraints Imposed by These Characteristics on Citation Source and Traceability
Pulp and paper production data mostly consists of real-time or batch-bound structured and semi-structured content, which requires high traceability accuracy. First, differences in update frequencies across different documents require traceability systems to match data timestamps and production batches. This prevents records from non-current production cycles from being mixed into audit traceability results. Second, sheet break and grade change records from high-speed web production are strongly tied to the current production batch. Traceability must bind specific roll numbers and batch numbers to avoid errors from generalized keyword matching. Third, content such as online quality curves and operation logs has relatively long length. Traceability must retain complete original fields and fragments, and cannot only display summaries. In addition, there are large differences between laboratory report fields and DCS data fields. Precise matching of metadata from corresponding data sources is required to ensure that specific test items and measured values can be traced during audits.
How to Set Configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Documents such as pulp and paper online quality curves and paper machine operation records have long single-entry length. Sufficient tokens are required to retain core traceability fields such as batch numbers and test values |
responseData | Return only associated document metadata and content fragments | Precise display of metadata for specific sources such as pulp preparation sheets and beating records is required to support traceability verification for supplier and process quality audits |
detail | Enabled | Specific fields of original documents must be displayed, such as pulp ratio and grammage test values, to ensure that auditors can verify consistency between original data and generated conclusions |
| Similarity Threshold | 0.7–0.9 | High precision is required for pulp and paper production data fields. Filtering of historical records unrelated to the current audit is required, to only retain documents strongly tied to the current audit batch |
| Single Recall Limit | Top 6 entries | There are many batches of production documents. Too many recalls will exceed the context window, and quality audits only require core data from three categories: pulp preparation, beating, and quality testing |
Three Common Misconfigurations
- Setting the minimum relevance threshold to 1 still results in responses containing a large number of sheet break records from non-current batches. The cause is that traceability recalled documents are not bound to the current audit production batch, and recall relies only on keyword matching without filtering historical data.
- Traceability content is truncated, and complete pulp preparation sheet batch numbers and pulp ratios cannot be displayed. The cause is that
quoteMaxTokenis set too small, and does not adapt to the token usage limit of long documents. - No metadata or links from source documents appear in the response. The cause is that the
responseDataconfiguration is not enabled, and only generated audit conclusions are returned without associating original data sources.
How to Confirm Proper Configuration
- Submit a quality audit request for a specific production batch, and check whether the response includes metadata and content fragments of the pulp preparation sheet and beating record for that batch.
- Adjust the value of
quoteMaxToken, and verify whether the length of traceability content in the response changes with the configuration, to confirm that the parameter takes effect. - Test a query unrelated to the current production batch, confirm that the recall result similarity matches the preset threshold, and no unrelated content is included.
- Check whether the FastGPT v4.17.0 version identifier appears in the traceability log or at the bottom of the response, to confirm that the currently used version supports the corresponding configurations.
The values provided are common starting points and should be measured against the reader's own samples.