Skip to main content

Documentation Index

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

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

Concept: Applications | Policy Engine
Returns the configuration in effect for the application your API key is bound to: which application and environment the key belongs to, the governance policy applied to its evaluations, and whether that policy is actively enforced or running in monitor mode. This is read-only. Policies are edited in the dashboard; the API key cannot change them.

​
Request

curl https://api.responsibleailabs.ai/railscore/v1/config \
  -H "Authorization: Bearer YOUR_RAIL_API_KEY"

​
Response

{
  "application": {
    "id": "app_1a2b3c",
    "environment": "production",
    "organization": "Acme",
    "plan": "pro"
  },
  "policy": {
    "enforcement": "regenerate",
    "evalMode": "deep",
    "overallThreshold": 7.5,
    "domain": "healthcare",
    "dimensionThresholds": { "Safety": 8 },
    "dimensionWeights": { "Safety": 20 },
    "compliance": ["India DPDP"],
    "safeRegenerate": { "enabled": true, "maxAttempts": 2, "fallback": "human_review" },
    "locked": false
  },
  "enforcement": { "active": false, "mode": "monitor" }
}
​
application
object
The application this key is scoped to: id, environment, organization, and plan.
​
policy
object
The governance policy applied to this application’s evaluations.
  • enforcement — log_only (record outcomes only), block (reject below threshold), or regenerate (attempt a safe rewrite, then apply the fallback).
  • evalMode — default evaluation depth: basic or deep.
  • overallThreshold — minimum RAIL score (0–10) to pass.
  • domain — domain context that tunes dimension sensitivity.
  • dimensionThresholds — per-dimension minimum scores; a dimension below its threshold fails the policy.
  • dimensionWeights — how much each dimension contributes to the overall score.
  • compliance — frameworks tracked for this application.
  • safeRegenerate — rewrite settings: enabled, maxAttempts, and fallback (block, log, or human_review).
  • locked — when true, this policy is authoritative and per-request overrides are ignored.
​
enforcement
object
Whether the policy is currently shaping responses. active: true (mode: enforce) means the policy can block or regenerate; active: false (mode: monitor) means outcomes are recorded but responses are not altered.
Use enforcement.mode to know how your application is governed right now. In monitor mode every evaluation still reports a policy_outcome (see Evaluation), so you can observe what would be blocked before enforcement is turned on.