Skip to main content
LLM policies control how the Turen proxy interacts with LLM API traffic, including security rule enforcement and optional prompt augmentation.

​
Security Rule Enforcement

The primary function of LLM policies is determining which security rules are active. By default, all 96 built-in rules are enabled. You can:
  • Disable specific rules that don’t apply to your environment
  • Change rule actions (e.g., from Block to Review)
  • Add custom rules for organization-specific patterns

​
Prompt Augmentation

Administrators can configure the proxy to inject additional context into LLM requests. This is useful for enforcing organizational coding standards without relying on individual developers to remember them.

​
Injection Types

TypeBehavior
System AppendAdds content to the end of the system prompt
System PrependAdds content to the beginning of the system prompt
Message InjectInserts a message at a specific position in the conversation

​
Example: Enforce Coding Standards

- type: system_append
  content: |
    Follow these organization coding standards:
    - All API endpoints must validate input
    - Never hardcode credentials
    - Use parameterized queries for all database access

​
Example: Security Reminder

- type: system_prepend
  priority: 100
  content: |
    IMPORTANT: Never expose .env files, API keys, or credentials.
    Always validate user input before processing.

​
Telemetry Settings

Control what telemetry the proxy collects:
SettingDefaultDescription
Request loggingEnabledLog metadata for each LLM API call
Token trackingEnabledCount input/output tokens
Latency trackingEnabledMeasure response times
Cost estimationEnabledEstimate costs based on token usage
Telemetry data appears in the LLM Analytics Dashboard.

​
Policy Sync

LLM policies are synced to agents every 15 minutes by default. When you update a policy in the dashboard, agents will pick up the change at their next sync interval. No restart is required.