What Data for This Category Looks Like
Daily documents for this category include mold design task sheets, DFM reports, part drawings, machining process cards, trial mold reports, mold modification notices, mold records, and more. Data sources cover customer product models, mold flow or forming simulation results, CAD/CAM exported files, machining and measurement data, and trial mold part inspection data. Data updates are triggered per individual mold project, each trial mold, or each mold modification. Individual documents contain fields such as mold number, part material, machining parameters, defect type, repair records, and others. Units mostly use millimeters, megapascals, hours, and other commonly used engineering measurement standards.
How These Characteristics Impose Constraints on Citation Sources and Traceability
Data for this category is tied to a clear project cycle and a unique mold ID. Citation traceability must strictly match the current project context to avoid process document confusion caused by cross-project citations. Document formats exported by various industrial tools vary widely, so mixed traceability of structured fields and unstructured text must be supported. Data update frequency changes frequently with trial molds and mold modifications, so expired non-current project documents must be filtered in real time. A traceability error may cause process document citation failure, leading to mold delivery delays or workpiece defects.
Configuration Settings
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 8000–12000 token | Single documents related to molds are usually several thousand characters long. This range covers the full content of process cards and trial mold reports, avoiding truncation of critical citation information |
responseData | Recall only documents associated with the current project | Data for this category is managed by project. This configuration eliminates irrelevant mold data from cross-project sources and ensures accurate traceability |
detail | Recall the top 8 matching sources | Mold process documents are linked to multiple machining, trial mold, and mold modification datasets. A sufficient number of recalls covers all relevant citation sources |
| Citation Source Filtering Rules | Filter by document creation time within the project cycle | Matches the project-based update characteristic of the data, filtering out expired non-current project documents |
| HTTP Data Source Binding | Mandatorily carry project ID metadata | When referencing trial mold or machining data returned via HTTP, this allows association with the current mold project and avoids traceability confusion |
Three Common Configuration Errors
- Issue: When setting
quoteMaxToken, values between 100 and 900 cannot be selected; only 100 or values above 900 are available. Cause: The system defaults to a fixed interval for the optional step size of this parameter, and custom step size configuration is not enabled. - Issue: An error is triggered when attempting to reference machining data returned via HTTP in an AI response, and citation traceability identifiers cannot be generated normally. Cause: Project ID metadata binding for the HTTP data source is not configured, so the system cannot identify the owning project of the citation source.
- Issue: After setting the minimum relevance score to 1, the response still outputs a large number of unrelated citation contents. Cause: The configuration for filtering citation sources by project ID is not enabled, resulting in the recall of irrelevant mold documents from cross-project sources.
How to Verify Correct Configuration
- Upload the trial mold report for the current project, initiate a query that references the defect types mentioned in the report, and check if only documents associated with this project are recalled in the response.
- View the citation traceability module of the response, confirm that the source fields (such as mold number, project ID) of each cited content match the current task.
- Adjust the
quoteMaxTokenparameter to a lower value, verify that the total token count of cited content in the response does not exceed the set value. - Submit a query that includes machining data returned via HTTP, confirm that the response correctly labels the citation identifier and owning project for this HTTP source.
The values provided are common starting points and should be measured against the reader's own samples.