Three notebooks that run the full GuardEx pipeline in-process - no API keys, no server, no mock data.
| # | Notebook | What you'll do | Time |
|---|---|---|---|
| 01 | Quickstart | Install, first screen, block unsafe content and injection, mask PII, guard an LLM call | 10 min |
| 02 | PII Detection | 31 entity types, mask vs block, thresholds, deny/allow lists, the reversible PII Vault | 15 min |
| 03 | Content Safety | S1-S14 taxonomy, the classification cascade, observe-only rollout, injection defense, refusals | 15 min |
| Notebook | Badge |
|---|---|
| 01 Quickstart |  |
| 02 PII Detection | |
| 03 Content Safety | |
Opens notebook 01; use the JupyterLab file browser for the others.
pip install 'guardex-ai[local]' jupyter
jupyter lab docs/notebooks/The first Guard() construction downloads about 250 MB of models to ~/.cache/; every run after that is warm.