Data Profile
This category of data comes from advertising placement platforms, landing page links, product qualification documents, and general advertising law compliance clauses. Update frequency is either daily or real-time. Daily processed documents include ad creative copy, landing page screenshots, rectification opinions from reviewers, and modification records submitted by advertisers. Data fields include creative unique identifier, placement channel type, violation content type, review status, and last modification time. Data formats differ across sources. Some landing page screenshots are unstructured image files, while ad creative copy uses plain text format.
Constraints for Scheduled Tasks and Automation
Since data sources include multi-platform APIs, static files, and rule texts, scheduled tasks must support cross-data-source pulling to avoid missing cross-channel creatives. Data update frequencies range from real-time to daily. Task trigger timing must match the update rhythm of corresponding creatives. Real-time creatives can be configured for event triggering, while daily-updated creatives can use scheduled triggering. Documents include unstructured images and plain text, so task workflows must support parsing and validation for both formats. The one-by-one review business characteristic requires that the number of creatives processed in a single task be controlled within a reasonable range to avoid task timeouts or resource overload, ensuring stability of the review process. Violation outcomes directly relate to advertising account risk, so a fast alert and retry mechanism must trigger after task failure.
Configuration Settings
These configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Task Trigger Timing | Every 15 minutes (real-time creatives), 2:00 AM daily (daily-updated creatives) | Matches the real-time/daily update frequency of creatives, avoids task redundancy or delay |
| Number of Creatives Processed Per Run | First 200 entries | Adapts to the resource load limit of a single task, avoids timeouts or interruptions |
| Timeout Threshold | 600 seconds | Covers the average time required to review a single creative, reserves buffer time for temporary network fluctuations |
| Number of Retries on Failure | 2 times | Covers scenarios of temporary API fluctuations or format parsing errors, reduces invalid retries |
| Alert Trigger Condition | 3 consecutive task failures | Notifies relevant personnel in a timely manner to handle exceptions, avoids accumulation of violation risks |
| Document Format Adaptation | Automatically identify image/text formats | Adapts to mixed processing requirements for landing page screenshots and ad creative copy |
Common Misconfigurations
- Symptom: Only creatives from a single channel appear in review results after task execution. Cause: Cross-ad platform data source pulling scope was not configured, only creatives from one channel were pulled.
- Symptom: After a scheduled task generates a report, the system prompts that automatic copying is not possible, requiring manual operation. Cause: Automatic copying permission was not enabled in system settings, or the browser blocked the page copy request.
- Symptom: Ad creative copy imported into the knowledge base has missing blocks. Cause: The segmentation parameter settings did not match the actual length of the creative copy, and insufficient text buffer space was reserved.
How to Verify Correct Configuration
- Manually trigger a test task, confirm that the number of pulled creatives matches the configured per-run processing limit.
- View task execution logs, confirm there are no format parsing errors, timeout errors, or data source connection failure records.
- Simulate a task failure scenario, verify that the alert rule triggers notifications normally.
- Trigger a report generation operation, confirm that the automatic copying function completes normally without requiring manual intervention.
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.