What data for this category looks like
Image annotation project data primarily comes from image samples produced by annotation platforms, annotation specification documents, bounding box example files, category dictionaries, as well as quality inspection reports and rework lists generated from random inspections. Data update frequency is adjusted daily or per batch annotation task cycle. Daily output documents include detailed annotation rules, target bounding box reference examples, single-batch quality inspection result summaries, and rework sample details. Data fields include unique image identifiers, annotation categories, bounding box coordinates, confidence scores, quality inspection results, rework reasons, etc. Units are mostly pixels, items, batches, and similar.
What constraints these characteristics impose on workflow orchestration
Image annotation project data includes pixel-level coordinates and batch-level quality inspection results, which impose clear requirements on node trigger logic for workflows. Data updated daily or per batch must be aligned with task cycle orchestrated nodes to avoid mixing cross-batch data. Multi-document type inputs require format validation between nodes to ensure annotation specifications and bounding box examples can be read correctly. For the transfer of structured data such as bounding box coordinates and confidence scores, clear field mapping rules must be defined to prevent data loss or misalignment. Associated processing of rework lists must bind the unique identifier of the corresponding annotation task to ensure rework workflows are traceable. Additionally, the decentralized nature of data sources requires workflows to include multi-source data aggregation nodes to unify processing of information from different entry points such as annotation platforms and inspection results.
How to set configurations
This configuration applies to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
variables | Bind annotation batch ID, image storage path, category dictionary ID | Ensure workflow nodes accurately associate annotation data for the corresponding batch, avoid cross-batch data mixing |
| Trigger Timing | Trigger on a daily schedule or after a single batch annotation is completed | Match the data update rhythm of image annotation projects (daily or per batch) to avoid unnecessary execution |
| Format Validation Rules | Validate JSON format for annotation specification documents, validate image paths and coordinate formats for bounding box examples | Adapt to input requirements for multiple document types including annotation specifications, bounding box examples, and quality inspection reports |
nodeId | Assign unique identifiers sequentially along the data transfer link | Clarify workflow node execution order, ensure data such as bounding box coordinates and quality inspection results follows the correct path during transfer |
| Node Timeout Threshold | 300-600 seconds | Adapt to processing duration for image annotation data, prevent node execution interruptions caused by large data volumes |
Three Common Mistakes
- Symptom: An error
localStorage is not definedappears in the console when the workflow code node executeslocalStorage.getItem('userId'). Cause: The code execution environment for FastGPT workflows is server-side, which does not support browser-exclusive localStorage APIs, and cannot directly read local storage content. - Symptom: After calling a workflow via API, the returned results do not include the thinking process logs for AI nodes. Cause: The thinking process output option was not enabled in the AI node configuration of the workflow, so the node did not generate corresponding log content.
- Symptom: After calling the code node that retrieves the application URL in the workflow, the returned results are empty. Cause: A pre-node for application sharing permission configuration was not included in the workflow, or the unique identifier parameter of the application was not bound correctly.
How to Confirm Configuration is Complete
- Navigate to the
variablesconfiguration page for the workflow, confirm that necessary associated parameters such as annotation batch ID and image storage path have been bound. The parameter loading status can be viewed via the node debugging panel. - Trigger a single test workflow, check the node execution logs to confirm that the node execution order matches the preset
nodeIdlink. - Upload a batch of test annotation documents to verify whether the format validation node can validate different types of input files according to the configured rules.
- Simulate workflow execution with batch data to confirm that the node timeout threshold configuration meets the actual duration requirements of current data processing.
The values provided are common starting points and should be measured 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.