Policy JSON
{
"guard": {
"pii_detection": {
"enabled": true,
"action": "REDACT",
"patterns": ["EMAIL", "PHONE", "SSN", "CREDIT_CARD", "ADDRESS"],
"redact_in_logs": true
},
"prompt_injection": {
"enabled": true,
"sensitivity": "high",
"action": "DENY",
"log_attempts": true
},
"content_moderation": {
"enabled": true,
"threshold": 0.6,
"categories": ["hate", "violence", "sexual", "self_harm"],
"action": "DENY"
}
},
"registry": {
"models": ["gpt-4o-mini", "claude-3-haiku"],
"strict": true,
"tools": ["knowledge_base_search", "ticket_create", "ticket_update", "customer_lookup"],
"deny_unlisted": true
},
"cost": {
"budget": {
"daily": 25.00,
"action_on_exceed": "DENY",
"alert_threshold": 0.8
},
"per_request_limit": 0.50,
"action_on_exceed": "DENY",
"estimate_before_execution": true
},
"memory": {
"default_scope": "session",
"allowed_scopes": ["request", "session"],
"purge_on_session_end": true,
"deny_pii_in_memory": true
},
"reliability": {
"circuit_breaker": {
"failure_threshold": 5,
"reset_timeout_ms": 15000,
"half_open_max_attempts": 1
},
"retry": {
"max_retries": 2,
"backoff": "exponential",
"max_backoff_ms": 3000,
"action_on_exhausted": "DENY"
}
},
"audit": {
"enabled": true,
"log_decisions": true,
"log_evidence": true,
"redact_pii": true,
"retention_days": 90
}
}
policy = {
"guard": {
"pii_detection": {
"enabled": True,
"action": "REDACT",
"patterns": ["EMAIL", "PHONE", "SSN", "CREDIT_CARD", "ADDRESS"],
"redact_in_logs": True,
},
"prompt_injection": {
"enabled": True,
"sensitivity": "high",
"action": "DENY",
"log_attempts": True,
},
"content_moderation": {
"enabled": True,
"threshold": 0.6,
"categories": ["hate", "violence", "sexual", "self_harm"],
"action": "DENY",
},
},
"registry": {
"models": ["gpt-4o-mini", "claude-3-haiku"],
"strict": True,
"tools": ["knowledge_base_search", "ticket_create", "ticket_update", "customer_lookup"],
"deny_unlisted": True,
},
"cost": {
"budget": {"daily": 25.00, "action_on_exceed": "DENY", "alert_threshold": 0.8},
"per_request_limit": 0.50,
"action_on_exceed": "DENY",
"estimate_before_execution": True,
},
"memory": {
"default_scope": "session",
"allowed_scopes": ["request", "session"],
"purge_on_session_end": True,
"deny_pii_in_memory": True,
},
"reliability": {
"circuit_breaker": {
"failure_threshold": 5,
"reset_timeout_ms": 15000,
"half_open_max_attempts": 1,
},
"retry": {
"max_retries": 2,
"backoff": "exponential",
"max_backoff_ms": 3000,
"action_on_exhausted": "DENY",
},
},
"audit": {
"enabled": True,
"log_decisions": True,
"log_evidence": True,
"redact_pii": True,
"retention_days": 90,
},
}
What’s Included
| Dimension | Configuration | Why |
|---|---|---|
| Security | PII redaction + injection defense + content moderation | Customers share personal data; agents must not leak it or produce harmful content |
| Registry | Cost-efficient models + support-specific tools | Support bots don’t need expensive models; limit tools to support operations |
| Cost | 25/daybudget+0.50/request limit | Prevent runaway costs from high-volume support traffic |
| Memory | Session-only + PII denial | No customer data persists beyond the conversation |
| Reliability | Circuit breaker + 2 retries | Fast recovery for real-time customer interactions |
| Audit | Full logging with PII redaction | Track interactions for quality review without storing PII |
Install
# Install the customer support bot starter
tealtiger policy install --starter customer-support-bot
# Or copy the JSON above into your tealtiger.config.json
Customize
| Setting | Default | Adjust For |
|---|---|---|
registry.tools | 4 support tools | Match your agent’s actual tool set |
registry.models | Cost-efficient models | Upgrade to gpt-4o if quality requires it |
cost.budget.daily | 25.00 | Scale with your support volume |
cost.per_request_limit | 0.50 | Raise if conversations are long |
content_moderation.threshold | 0.6 | Lower for stricter moderation |
Related
- PII Redaction — Basic — Standalone PII policy
- Budget — Daily $10 — Tighter budget option
- OWASP ASI Starter — Add compliance coverage
- TealGuard Documentation — Full API reference