What the Data for This Category Looks Like
Data for this category draws from customer Gerber files, drill files, BOM lists, factory CAM system data, production equipment parameters, bath solution test results, AOI inspection reports, electrical testing data, and laboratory slicing reports. Data updates align with production batch progress, with corresponding documents generated either per shift or via continuous collection. A single production batch document includes engineering production instructions, lamination parameters, drilling offset records, chemical solution concentration data, and more. Fields include layer number, aperture size, defect coordinates, on-resistance, and more. Units cover imperial mil, metric mm, and ohms.
What Constraints These Characteristics Impose on the "Multi-turn Dialogue and Prompt Engineering" Link
PCB manufacturing inspection and process record audits involve multiple data types: vector drawings, numerical parameters, and text reports. Multi-turn dialogue must associate corresponding context to the current production batch. Data updates generate in real time or per shift as production advances, so dialogue must limit context scope to match the current batch’s documents to avoid cross-batch data interference. Field units across different documents vary, including imperial mil, metric mm, and ohms. Prompt engineering must clearly define unified unit conversion rules to prevent parameter confusion. Multi-turn dialogue uses chatId to link historical interactions for the same production batch, ensuring audit process continuity. It also limits the context length of messages to stop redundant historical data from disrupting logical judgment.
How to Set Up the Configuration
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
chatId | Bind to the unique identifier of the current PCB production batch, formatted as PCB+date+serial number | Ensure context association for multi-turn dialogue in the same batch inspection, avoiding cross-batch data confusion |
messages | Retain the most recent 3 rounds of interactions and core document metadata for the current production batch, with total length not exceeding 1200 characters | Control context length, avoid redundant data interfering with audit logic, and adapt to the context processing limit of FastGPT v4.17.0 |
variables | Inject the Gerber file ID, AOI report number, and bath solution test time for the current batch | Let the AI quickly locate exclusive data for current production, improving audit accuracy |
| User Question | Fixed prefix of "Please conduct inspection record audit for PCB production batch {batch number}, combined with the following documents", followed by specific audit questions | Clarify the dialogue scenario, guiding the AI to focus on inspection and process records for the current batch |
| File Links | Upload AOI reports, electrical testing data, and engineering production instructions for the current batch, set the association path to /PCB/production/{batch number}/ | Let the AI directly retrieve exclusive documents for the corresponding batch without cross-batch searching |
Three Common Misconfigurations
- Phenomenon: Form input fields set in the workflow do not appear in the dialogue interface, and interaction content is empty. Cause: Form variables are not bound to the
variablesparameter, causing preset input items to fail to load in the dialogue context. - Phenomenon: The audit results returned by the AI contain unit confusion, such as misjudging a mil value as a mm value. Cause: The prompt engineering does not clearly specify conversion rules for field units, and does not limit the unit standard for data reading.
- Phenomenon: Cross-batch historical data interferes with multi-turn dialogue, leading to deviations in audit conclusions. Cause: The unique identifier of the current production batch is not bound via
chatId, and the context associates document data from non-current batches.
How to Confirm the Configuration is Properly Set Up
- Check the AI dialogue module in the workflow, confirm that form input fields are synchronously displayed on the dialogue launch interface with no missing items.
- Test initiating a dialogue, enter the current batch identifier, and verify that the AI can correctly retrieve the associated documents for the corresponding batch.
- Initiate multi-round interaction, check that the content scope of the dialogue context meets the preset configuration requirements, with no irrelevant historical data.
- Review the prompt content, confirm that the audit requirements for the current specific scenario are clearly marked, with no ambiguous statements.
The values provided here are common starting points, and should be evaluated against the reader's own samples.