Try Shield Live
Type any prompt or pick an example below. Watch the 5-stage detection pipeline analyze it in real time with actual ML scores.
Powered by a real oubliette-shield instance with an embedded ML classifier (F1=0.98). No LLM needed for the demo.
shield.analyze()
0 / 2000
Pipeline Execution
Sanitize
Pre-Filter
ML Classifier
LLM Judge
Session
Analyzing...
Analysis failed.
Check your connection or try again.
Verdict
Threat:
ML Confidence
-- / 1.00
SAFE MALICIOUS
Sanitizations Applied
Threat Mapping
Session State
Interactions: 0
Attacks detected: 0
Escalated: No
Reason:
Total time: --ms
Select an example prompt or type your own to get started.
Run Locally with Full Pipeline
This demo uses the ML classifier only (LLM stage is skipped). Install locally for the complete 5-stage pipeline with your own LLM backend:
terminal
# Install from PyPI
$ pip install oubliette-shield[flask,ollama]
# Quick test in Python
$ python -c "from oubliette_shield import Shield; s = Shield(); print(s.analyze('ignore all rules').verdict)"