What the data for this category looks like
Core data for ad creative review comes from ad delivery platforms, landing page resources, product qualification documents, and advertising law compliance clauses. Daily processed documents include pending ad creative files, landing page screenshots, review feedback receipts, and modification records. Data updates occur in real time or via daily batch updates. Core fields include creative unique identifier, delivery channel, violation type tag, review status, and modification timestamp. The data structure covers text and image formats, and cross-source verification of content consistency is required.
What constraints these characteristics impose on scheduled tasks and automation
The real-time or daily data update rhythm requires scheduled task trigger frequencies to match the data update pace, to avoid delayed processing or repeated execution. Multi-source, multi-format data requires targeted preprocessing logic, to ensure review rules for text and image creatives adapt to their respective verification standards. The core one-by-one review process requires controlling single submission volume during batch processing, to prevent node resource overload. Error consequences involve advertising delivery compliance risks, so failure retry and alert mechanisms must be configured. Complete review logs must also be retained for fault retrospective purposes.
How to configure settings
The following configurations are based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Task trigger frequency | Trigger in real time according to creative update events, or once every 5 minutes | Matches the daily or real-time data update rhythm |
| Batch processing count | 20-50 items per single processing | Adapts to the average time required for single creative review and node resources |
| Timeout threshold | 300 seconds | Covers the maximum processing duration for a single creative review |
| Failure retry count | 2 times | Avoids resource overload caused by repeated triggers |
| Idempotency verification key | Bind to creative unique ID | Prevents scheduled tasks from repeatedly generating review records |
| Log retention period | 7 days | Meets log retrieval requirements for fault troubleshooting |
Three common misconfigurations
- Phenomenon: When submitting ad creatives in batches, the first 20 entries are reviewed normally, but subsequent tasks return no results or status code 504. Cause: No maximum batch processing limit is set, and the single submission volume exceeds the memory limit of FastGPT nodes.
- Phenomenon: Repeated review feedback entries appear in scheduled generated review reports. Cause: No creative unique ID is bound as the idempotency verification key for the scheduled task, leading to repeated record generation when the task runs multiple times.
- Phenomenon: No alert notification is sent after a review failure. Cause: The alert trigger condition is set to single failure, and no logic for cumulative trigger based on failure count is configured, leading to redundant alerts from frequent temporary exceptions.
How to confirm configurations are correct
- Manually submit a single pending creative, check if a corresponding entry is generated in the scheduled task queue, confirm that the trigger logic matches the data update rhythm.
- Simulate batch submission of multiple pending creatives, check the FastGPT node resource monitoring panel, confirm that the single processing volume does not exceed the node's carrying limit, verify the rationality of the batch configuration.
- Manually trigger a simulated failed review process, wait for the preset number of retries, check if the preset alert channel is triggered, confirm that the alert logic works correctly.
- Check the log storage path of review tasks, confirm that logs are being retained normally according to the configured retention period, verify that the log configuration is active.
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.