Observed since 20 Oct, 20:16 UTC
No superseded metadata states collected. Unchanged observations do not create new revisions.
Aegis policy pack: model-based content inspection — secrets, custom entities, sensitivity and prompt-attack labels.
Model-based content inspection for Aegis:
mask entities no pattern can describe (internal hostnames, project code names)
and label requests (sensitivity, intent) for the label policy to act on. The
model is a setting — GLiNER2 by default, swappable through the
aegis.content_models entry point.
# CPU-only machines: install PyTorch's CPU build first (the default wheel bundles CUDA)
pip install torch --index-url https://download.pytorch.org/whl/cpu
pip install "aegis-gateway-pack-content[model]"
guardrails:
content:
pack: aegis.content
model: fastino/gliner2-base-v1
entities:
internal hostname: hostname or address of an internal server or service
labels:
intent: [normal request, prompt injection or jailbreak attempt]
pipeline:
ingress: [content]
egress: [content]
See Content for what it's good at — measured, not claimed.
pip install aegis-gateway (add [content] for the model).Part of Aegis, a self-hosted AI gateway. Licensed under AGPL-3.0-or-later.
Download counts and popularity will appear here as the collector receives them.
| Version | Published | Direct dependencies | Open version |
|---|---|---|---|
| 2.0.0a6 | 27 Sept 2026 | 4 |
This describes our collected data; it is not a security assessment.
Observed since 20 Oct, 20:16 UTC
No superseded metadata states collected. Unchanged observations do not create new revisions.
package:fa504cd369171b34f2a55218f0aca28b0a009c23da08af711f36f229f6c960fd