Pipeline Presets
Three presets provide one-click pipeline configuration:
Starter (~11 steps)
Section titled “Starter (~11 steps)”For development and low-risk internal agents.
- Input validation, rate limiting
- Basic PII detection
- Prompt injection detection
- Output scanning
- Cost tracking
tappass pipelines create my-pipeline --preset starterStandard (~37 steps)
Section titled “Standard (~37 steps)”For production agents with external access.
- Everything in Starter
- Secret scanning, infrastructure detection
- Data classification
- Tool permissions and constraints
- Taint tracking
- Exfiltration detection
- Memory poison detection
- Insider threat detection
- Shell bleed detection
tappass pipelines create my-pipeline --preset standardRegulated (~43 steps)
Section titled “Regulated (~43 steps)”For financial services, healthcare, and EU AI Act high-risk applications.
- Everything in Standard
- LLM-based classification judge
- Content safety (toxicity, bias)
- Human approval gate (
require_approval) - Unicode attack detection
- Full tool call scanning with forbidden zones
- Loop guard
- Debug logging
tappass pipelines create my-pipeline --preset regulatedCustom
Section titled “Custom”Build your own pipeline from individual steps:
from tappass import TapPass
with TapPass("http://localhost:9620", admin_key="tp_admin_...") as tp: tp.create_pipeline("custom", steps={ "detect_pii": {"on_detection": "block"}, "detect_injection": {"on_detection": "block", "threshold": 0.5}, "detect_exfiltration": {"on_detection": "block"}, "detect_memory_poison": {"on_detection": "block"}, })