Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.neuraltrust.ai/llms.txt

Use this file to discover all available pages before exploring further.

Policies define the action chains the Actions API will execute. Each policy contains one or more actions and optional observability (telemetry) and TrustLens settings. You reference a policy by its id when calling POST /v1/actions.

​
What is a policy?

A policy is a declarative object that describes:
  • The list of actions to run, with order (priority) and parallelization hints
  • Optional telemetry exporters and settings
  • Optional TrustLens configuration for advanced insights
Policies enable repeatable, versionable orchestration of actions with clear access control through allowed policies on API keys.

​
Create a policy

Use the admin endpoint POST /v1/policies with at least one action: 
curl -X POST \
  https://<host>/v1/policies \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <ADMIN_API_KEY>" \
  -d '{
    "name": "policy-textops",
    "actions": [
      {"name": "sanitize", "priority": 1, "parallel": false, "settings": {"level": "standard"}},
      {"name": "classify", "priority": 2, "parallel": false, "settings": {"labels": ["spam", "ham"]}}
    ]
  }'
Response: 
{ "id": "<uuid>" }

​
Use a policy

Call POST /v1/actions with the policy_id you received and a payload to process: 
curl -X POST \
  https://<host>/v1/actions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <API_KEY>" \
  -d '{
    "policy_id": "<uuid>",
    "payload": {"input_text": "Hello world", "user_id": "u-123"}
  }'

​
Best practices

  • Keep actions small and composable; use multiple actions rather than one complex action
  • Prefer sequential execution unless parallel execution is safe and idempotent
  • Name policies clearly and use tags/metadata in your platform to track versions
  • Restrict execution using allowed policies on API keys