Skip to content
TruthVouch Docs

Governance API

The Governance API lets you define and enforce policies that control how LLM calls are processed through the Firewall.

Overview

Use the Governance API to:

  • Define policies for PII masking, fact-checking, and policy enforcement
  • Manage policy versions and rollouts
  • Evaluate if a request violates policies
  • Audit policy enforcement actions

Key Endpoints

Create Policy

POST /api/v1/governance/policies

Terminal window
curl -X POST https://api.truthvouch.com/api/v1/governance/policies \
  -H "Authorization: Bearer tv_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Production Safety Policy",
    "description": "Enforce PII masking on all production calls",
    "code": "package policies.pii\n\ndeny[msg] {\n  input.prompt contains_ssn\n  msg := \"SSN detected\"\n}\n\ncontains_ssn {\n  re_match(\"\\\\d{3}-\\\\d{2}-\\\\d{4}\", input.prompt)\n}",
    "active": true
  }'

Get Policy

GET /api/v1/governance/policies/{policyId}

Terminal window
curl https://api.truthvouch.com/api/v1/governance/policies/policy_abc123 \
  -H "Authorization: Bearer tv_live_..."

List Policies

GET /api/v1/governance/policies

Terminal window
curl https://api.truthvouch.com/api/v1/governance/policies?status=active \
  -H "Authorization: Bearer tv_live_..."

Update Policy

PATCH /api/v1/governance/policies/{policyId}

Terminal window
curl -X PATCH https://api.truthvouch.com/api/v1/governance/policies/policy_abc123 \
  -H "Authorization: Bearer tv_live_..." \
  -d '{
    "code": "package policies.pii\n...",
    "active": true
  }'

Evaluate Policy

Check if a request violates a policy:

POST /api/v1/governance/evaluations/run

Terminal window
curl -X POST https://api.truthvouch.com/api/v1/governance/evaluations/run \
  -H "Authorization: Bearer tv_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "policyId": "policy_abc123",
    "prompt": "What is 2+2?",
    "response": "2+2 equals 4"
  }'

Response:

{
  "data": {
    "verdict": "allowed",
    "violations": [],
    "alerts": []
  }
}

Policy Language

Policies are written in Rego (OPA — Open Policy Agent). Rego is a declarative language designed for policy enforcement.

Example Policies

PII Detection (Block SSN):

package policies.pii


deny[msg] {
  re_match("\\d{3}-\\d{2}-\\d{4}", input.prompt)
  msg := "SSN detected in prompt; please remove PII"
}

Content Safety (Block Violent Content):

package policies.safety


deny[msg] {
  input.response.category == "violent"
  input.response.score > 0.7
  msg := "Response contains violent content; not sending to user"
}

Model Allowlist:

package policies.model_control


deny[msg] {
  not is_approved_model(input.model)
  msg := sprintf("Model %v is not approved", [input.model])
}


is_approved_model(model) {
  approved = ["gpt-4", "gpt-4-turbo", "claude-3-sonnet"]
  model in approved
}

Token Limits:

package policies.tokens


deny[msg] {
  input.tokens_used > 10000
  msg := "Token limit exceeded for this request"
}

Policy Evaluation

Policies are evaluated using OPA (Open Policy Agent). Each policy returns:

  • allow — Request is allowed to proceed
  • deny — Request is blocked (with reason)
  • warn — Request is allowed but warning is logged

Policy Versioning

Policies have multiple versions for safe rollouts:

GET /api/v1/policies/{policyId}/versions

Terminal window
curl https://api.truthvouch.com/api/v1/policies/policy_abc123/versions \
  -H "Authorization: Bearer tv_live_..."

Each version tracks:

  • Rules and changes
  • Activation date
  • Rollback capability

Next Steps