Guardrails
Blocks, redactions, and warnings — last 24h.
Total events
7
last 24h
Blocked
2
GuardrailError raised
Redacted
3
PII / secrets removed
Warnings
2
flagged, not blocked
Rules triggered · last 24h
Guardrail enforced
Blocks forbidden terms
Redacts personal data
Blocks low-quality responses
Flags conflicting recalled memories
Recent events · newest first
Response blocked: hallucination score 0.210 is below threshold 0.500
PIIRedact: [EMAIL, PHONE] removed from [input]
HallucinationBlock: passed (score=0.820, threshold=0.500)
MemoryContradiction: recall returned a memory and its correction together — "The production database is hosted at cloud.acme.…" vs…
Configure guardrails
Rules run automatically on every LLM span.
Strips email, phone, SSN, credit card
peekr.guard.PIIRedact()Blocks secrets, API keys, forbidden terms
peekr.guard.Blocklist(
patterns=Blocklist.COMMON_SECRETS
)Blocks responses below faithfulness threshold
peekr.guard.HallucinationBlock(
threshold=0.5
)HIPAA, FDCPA, GDPR, UAE PDPL + 13 more
peekr.instrument(
compliance=["HIPAA"]
)Regulatory compliance packs
HIPAA, FDCPA, FINRA, GDPR, UAE PDPL — toggle from dashboard, enforced on next request.