What the data for this category looks like
Core data for text annotation projects comes from conversation texts provided by clients, custom vocabularies, and quality inspection sampling results produced by annotation platforms. Data is updated daily or per batch sample package cycle. Daily produced documents include annotation specifications, sample packages to be annotated, quality inspection reports, and problem feedback forms. Data fields include text content, entity start and end positions, intent tags, quality inspection markers, and more. Units include characters, tag IDs, sampling batch numbers, and others. The combination of fields for different annotation tasks is adjusted based on specific requirements.
What constraints these characteristics impose on the "workflow orchestration" link
Multi-source data requires workflows to support cross-platform input adaptation, and variables must be used to uniformly manage field formats from different sources to avoid data parsing errors. Daily or batch update rhythms require workflows to support scheduled triggering and batch task scheduling, reducing repeated manual triggering operations. Diverse document types require workflows to embed document reading nodes to quickly call reference content such as annotation specifications and quality inspection reports, adapting to the need for tag boundary explanations. Personalized field requirements require workflows to support branch judgment, routing processing links based on different tag types and quality inspection results, adapting to time-consuming links such as consistent sampling and difficult sample arbitration.
How to set configurations
| Configuration Item | Recommended Approach | Basis for This Approach |
|---|---|---|
| Variable Settings | Define three core variables: text content, tag type, and quality inspection result via variables | Matches the core data fields of text annotation projects, covering three core business scenarios: samples, tags, and quality inspection |
| Node ID | Assign a unique nodeId to each annotation link node | Facilitates workflow debugging and abnormal node location, in line with FastGPT workflow node management rules |
| Batch Trigger Threshold | 100–500 items/time | Adapts to the sample package update frequency of text annotation projects, avoiding overly high load from single triggering |
| Exception Retry Count | 2–3 times | Addresses temporary fluctuations in annotation platform interfaces, reducing process interruptions caused by single call failures |
| Branch Judgment Condition | Divert based on quality inspection marker fields | Matches the consistent sampling requirements of text annotation, assigning qualified/unqualified samples to different processing links |
| Global Variable Reference | Reference global parameters of annotation specification documents | Unifies tag boundary rules, avoiding tag definition deviations across different links |
Three common misconfigurations
- Attempting to bind custom variables when configuring HTTP service input parameters results in a prompt that the field does not exist. The cause is failing to declare the field name and data type of the input parameter in
variablesin advance. - Unable to export edited workflow files. The cause is failing to save the workflow to an authorized project space, or failing to enable the export permission configuration for the workflow.
- Default values of form input nodes cannot reference global variables or other form fields. The cause is failing to register the target variable in the workflow's global
variableslist.
How to confirm the configuration is correct
- Check whether the
variableslist includes core data fields such as text content, tag type, and quality inspection markers, ensuring coverage of input and output requirements for all annotation links. - Run a test workflow to verify whether branch nodes can correctly divert qualified and unqualified samples based on quality inspection marker fields, matching the quality inspection requirements of the annotation process.
- Trigger a batch test task to confirm that the workflow can process the corresponding number of samples according to the set threshold, adapting to the batch scheduling rules of FastGPT v4.17.0.
- Attempt to export the workflow file to confirm that the export function works normally, verifying that permission configurations and storage path settings are correct.
The values provided here are common starting points and should be evaluated against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-10-19.