Alerts
Alerts are events GPTfake raises when a monitored model’s behavior changes materially — a policy shift or a censorship spike. Use the alerts endpoints to list recent events and configure thresholds. To receive alerts in real time instead of polling, subscribe via Webhooks.
List alerts
GET /alertsQuery parameters:
| Param | Values | Default |
|---|---|---|
model | any monitored model id | all |
type | policy_change, censorship_spike | all |
since | ISO 8601 date | last 30 days |
curl "https://api.gptfake.com/v1/alerts?model=chatgpt&type=censorship_spike" \
-H "Authorization: Bearer YOUR_API_KEY"Response:
{
"alerts": [
{
"id": "alrt_8f21",
"model": "chatgpt",
"type": "censorship_spike",
"category": "political",
"delta": "+5.8pp",
"detected_at": "2024-11-24T06:00:00Z",
"methodology": "https://gptfake.com/monitoring/methodology"
}
]
}Alert types
type | Triggered when |
|---|---|
policy_change | A model’s stated content policy or observed rules change |
censorship_spike | A refusal rate rises sharply versus its recent baseline |
Each alert links to the methodology used to detect it and to the model’s metrics so you can verify the change. Detected changes are also summarized on the reports feed.
Get a single alert
GET /alerts/{alert_id}curl https://api.gptfake.com/v1/alerts/alrt_8f21 \
-H "Authorization: Bearer YOUR_API_KEY"Next steps
- Subscribe to alerts via webhooks for push delivery.
- Inspect the underlying metrics that triggered an alert.
GPTfake