What this category's data looks like
Business data for SMT mounting and electronic assembly mainly comes from customer-provided BOM, Gerber files, mounting coordinates, as well as factory MES systems, mounting machines, SPI/AOI inspection equipment, reflow oven temperature records, and similar sources. Daily documents include BOM and coordinate files, changeover sheets, stencil and furnace temperature records, first article confirmation reports, SPI/AOI inspection reports, repair records, and traceability labels. Data is updated in real time per work order and each changeover. Fields include part number, reference designator, component lot number, furnace temperature parameters, and more. Units are mostly millimeters, degrees Celsius, and similar units.
What constraints do these characteristics impose on workflow orchestration
This category’s data has strong real-time performance and frequent updates. Workflows must support dynamic calls to the latest work order data and equipment records to avoid using expired process parameters. Multiple types of business documents require workflows to adapt to knowledge base imports and retrievals of different formats, ensuring coverage of specific documents such as changeover sheets and furnace temperature records. Frequent changeover operations require workflows to quickly switch the knowledge base for the corresponding work order, reducing configuration redundancy. At the same time, errors such as component misplacement and cold solder joints have serious consequences. Workflows must add parameter verification links to ensure that retrieval results match the part numbers and reference designators of the current work order.
How to set configurations
This document is written based on FastGPT v4.17.0.
| Configuration Item | Recommended Practice | Rationale |
|---|---|---|
| Dynamic Knowledge Base Specification | Bind the work order ID variable in variables | Matches the real-time knowledge base update requirement per work order, and adapts to scenarios where different work orders correspond to different process documents |
| Context Retention Duration | 600 seconds | Adapts to the conversation cycle of multi-round process Q&A, preventing Q&A interruptions caused by expired context |
| Single Search Limit | Top 5 results | Balances retrieval efficiency and result coverage, meeting the demand for small-volume precise matching of SMT process parameters |
| Maximum Context Length | 800–1200 characters | Adapts to the single-paragraph content length of process documents, avoiding redundant information interfering with large model inference |
| Node Output Control | Turn off the session output switch | Ensures that the execution results of AI conversation nodes are only written to workflow variables, and do not mix into user-visible conversation streams |
| Variable Scope | Global variables | Ensures that work order parameters, component lot numbers and other information can be passed between nodes across multi-round conversations |
| Knowledge Base Import Limit | 1000 MB | Adapts to the size of a single SMT process document package, preventing import failures |
Three common mistakes
- Symptom: The variable list is empty when dynamically specifying a knowledge base, and the knowledge base for the corresponding work order cannot be selected. Cause: The knowledge base variable associated with the work order ID is not pre-defined in the workflow's
variablesconfiguration, causing the node to fail to recognize optional parameters. - Symptom: The execution results of AI conversation nodes are forcibly output to the session return content. Cause: The "session output" switch of the node is not turned off, and the node execution results are mixed into the user-visible conversation stream by default.
- Symptom: Process parameters are misaligned after multi-round conversations and cannot match the current work order. Cause: The context retention duration is not set, causing variables from old work orders to be retained and interfering with the current Q&A logic.
How to confirm the configuration is complete
- Upload test SMT changeover sheets and furnace temperature records, bind the test work order ID, trigger the workflow, and check if the dynamic knowledge base automatically associates the corresponding documents.
- Initiate multi-round process Q&A, check if the work order parameters defined in
variablesare normally passed between each node, with no loss or misalignment. - Run the AI conversation node, confirm that the execution results are only written to workflow variables and do not appear in the final session return content.
- Simulate two changeover operations, check if the workflow can quickly switch the corresponding knowledge base, with no configuration conflicts or retrieval errors.
The values given are common starting points and should be measured against the reader's own samples.