Skip to main content

Admin Policy Reference

This document defines the enterprise admin policy schema used to restrict sensitive runtime behavior.

  • Template: /inquira.policy.toml.example
  • Recommended path: /etc/inquira/policy.toml (or INQUIRA_POLICY_PATH)
  • Precedence (highest first): admin policy > inquira.toml > UI/user preferences

Meta​

meta.version​

  • Type: string
  • Purpose: policy schema version.

meta.owner​

  • Type: string
  • Purpose: owning team for support/audit.

meta.change_ticket​

  • Type: string
  • Purpose: change reference for audit trail.

Enforcement​

enforcement.mode​

  • Type: string
  • Allowed: enforce, audit
  • Purpose: choose strict blocking (enforce) or report-only mode (audit).

UI​

ui.hide_managed_settings​

  • Type: bool
  • Purpose: hide centrally managed settings in the UI.

Observability​

observability.allow​

  • Type: bool
  • Purpose: master gate for tracing/telemetry providers.

observability.redact_sensitive_content​

  • Type: bool
  • Purpose: force redaction for sensitive fields in exported traces.

Backend Phoenix​

backend.phoenix.allow​

  • Type: bool
  • Purpose: allow backend process trace export to Phoenix.

backend.phoenix.locked​

  • Type: bool
  • Purpose: prevent user/UI overrides for backend Phoenix settings.

backend.phoenix.allowed_endpoints​

  • Type: array[string]
  • Purpose: endpoint allowlist for backend Phoenix exporter.

Agent Phoenix​

agent_service.phoenix.allow​

  • Type: bool
  • Purpose: allow agent process trace export to Phoenix.

agent_service.phoenix.locked​

  • Type: bool
  • Purpose: prevent user/UI overrides for agent Phoenix settings.

agent_service.phoenix.allowed_endpoints​

  • Type: array[string]
  • Purpose: endpoint allowlist for agent Phoenix exporter.

LLM Controls​

llm.allowed_providers​

  • Type: array[string]
  • Purpose: provider allowlist enforced by runtime.

llm.block_custom_base_url​

  • Type: bool
  • Purpose: block custom model base URLs when required by policy.

llm.block_user_api_keys​

  • Type: bool
  • Purpose: disallow user-provided API keys.

llm.models.<provider>​

  • Type: array[string]
  • Purpose: model allowlist by provider.

Tools​

tools.allow​

  • Type: bool
  • Purpose: master gate for interactive tools.

tools.bash.allow​

  • Type: bool
  • Purpose: enable/disable shell execution tool.

tools.bash.locked​

  • Type: bool
  • Purpose: prevent user/UI override for bash tool.

tools.bash.allowed_commands​

  • Type: array[string]
  • Purpose: strict allowlist for executable commands.

tools.pip_install.allow​

  • Type: bool
  • Purpose: allow/deny package installs from prompts.

tools.pip_install.locked​

  • Type: bool
  • Purpose: prevent user/UI override for package install tool.

Execution Limits​

execution.max_timeout_seconds / execution.max_memory_mb / execution.max_output_kb​

  • Type: int
  • Purpose: hard runtime limits for safety and cost control.

Network Controls​

network.allowed_hosts​

  • Type: array[string]
  • Purpose: restrict outbound hosts when egress enforcement is enabled.

network.require_proxy​

  • Type: bool
  • Purpose: force organization proxy usage.

Data Controls​

data.allowed_workspace_roots​

  • Type: array[string]
  • Purpose: constrain workspace roots to approved directories.

data.block_export​

  • Type: bool
  • Purpose: prevent data/artifact exports outside policy.

Retention​

retention.max_conversation_days​

  • Type: int
  • Purpose: cap retention window for conversation history.

retention.max_artifact_hours​

  • Type: int
  • Purpose: cap retention window for artifacts.

Audit​

audit.log_policy_violations​

  • Type: bool
  • Purpose: log attempts to override managed policy.