What this category's data looks like
Data sources for this category include design documents, standard instrument and environmental monitoring data, calibration software outputs, laboratory information system records, and customer range requirements. Data update frequency is after each calibration operation for each tested piece. Daily documents include assembly and commissioning sheets, calibration procedures, original calibration records, error curves, metrology certificates, and software version records. Fields include measured parameters, calibration dates, error values, traceability numbers, environmental temperature and humidity, etc., with corresponding units such as Pa, ℃, mL/min, etc.
What constraints do these characteristics impose on the "workflow orchestration" link
The data characteristics of this category impose multiple constraints on workflow orchestration. Dispersed data sources that require integration with multiple systems mean the workflow must support multi-node pulling of content from different data sources. The data update frequency of immediate updates after a single calibration requires the workflow to bind real-time data of the current tested piece via variables to avoid mixing data across batches. Structural differences across multiple types of documents mean the retrieval node must support classification filtering by document type. Fields include specific measurement units and traceability information, so the workflow must add a verification link to ensure retrieved content matches the parameter requirements of the current tested piece. Time-consuming links such as standard instrument connection and multi-point data collection require configuring reasonable timeout thresholds.
How to set the configurations
The following configurations are written based on FastGPT v4.17.0.
| Configuration Item | Recommended Approach | Basis for This Approach |
|---|---|---|
| Knowledge Base Tag Filtering | Bind corresponding tags by document type (e.g., bind calibration records to the "calibration data" tag) | Matches the classification needs of multiple types of documents in this category |
| Workflow Branch Triggering | Branch by calibration result status in variables | Follows the requirements of audit traceability chain and error calculation links |
| Single Retrieval Count | Top 8 entries | Covers the retrieval needs of multiple types of documents including original calibration records and error curves |
| Context Window Size | 4000–6000 token | Adapts to processing long texts such as error curves and metrology certificates |
| Node Timeout Setting | 120 seconds | Addresses time-consuming links such as standard instrument connection and multi-point data collection |
| Variable Binding | Bind calibration data from the node corresponding to nodeId | Ensures all links in the workflow call real-time data of the current tested piece |
Three common mistakes
- Phenomenon: Knowledge base retrieval returns results that do not follow the configured tag filtering rules. Cause: The uploaded document tags are not bound to the retrieval node's filtering rules, or the tag names do not match the business document types.
- Phenomenon: Workflow execution does not complete all nodes and no error logs are generated. Cause: The node timeout setting is too short, and does not cover time-consuming links such as standard instrument connection and multi-point data collection, so the system automatically terminates without triggering an error.
- Phenomenon: Chat context associates more historical records than the set number. Cause: The context window filtering rules are not configured correctly, or the bound
variablesdo not limit the time range of historical records.
How to confirm the configuration is correct
- Upload an original calibration record document, add the "calibration data" tag, configure the tag filtering rules in the retrieval node, and verify that only documents with this tag are returned.
- Simulate triggering the branch condition, modify the calibration result status in
variablesto "qualified", and verify that the workflow switches to the audit approval branch process. - Simulate a calibration data collection node that takes 100 seconds, adjust the timeout setting to 120 seconds, and verify that the workflow can complete all links normally.
- Configure the context window to 4000 token, initiate a query containing multiple historical records, and verify that the returned historical record content meets the set window requirements.
The values provided are common starting points and should be measured against the reader's own samples.