> ## Documentation Index
> Fetch the complete documentation index at: https://developers.callaro.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Environments

> Use a strict sandbox-to-production promotion model for keys, data, webhooks, and operational controls.

# Keep sandbox and production contracts operationally isolated.

Callaro environments should be treated as separate trust zones, not just separate URLs.

## Environment matrix

| Area             | Sandbox                                | Production                                        |
| ---------------- | -------------------------------------- | ------------------------------------------------- |
| Data             | Synthetic or masked data only          | Real customer and operational data                |
| Keys             | Dedicated sandbox keys                 | Dedicated production keys with tighter ownership  |
| Webhooks         | Test endpoints/request bins            | Hardened receiver with SLO and alerting           |
| CRM integrations | Staging/sandbox CRM org                | Live CRM org with validated mappings              |
| Numbers          | Test inventory and non-critical routes | Live dialing inventory and number-health controls |
| Incident impact  | Low, controlled                        | Revenue and customer experience impact            |

## Hard boundaries

* Never reuse API keys across environments.
* Never point sandbox events to production CRM write paths.
* Never load unmasked customer PII in sandbox.
* Never enable production webhook destinations before replay testing.

<Steps>
  <Step title="Validate in sandbox with test data">
    Run quickstart flow, confirm call outcomes, trace visibility, and webhook ingestion.
  </Step>

  <Step title="Run production-readiness verification">
    Validate script quality, suppression logic, schedule windows, and fallback handling with business stakeholders.
  </Step>

  <Step title="Promote credentials and endpoints">
    Provision production keys, configure production webhook URLs, and lock receiver authentication and observability.
  </Step>

  <Step title="Launch in controlled ramp">
    Start with a limited campaign cohort and monitor delivery, call outcomes, and CRM write quality.
  </Step>
</Steps>

## Promotion checklist

* Auth scopes reviewed and least privilege confirmed.
* Number pool tested for spam health and region fit.
* Webhook dedupe and retry handling verified.
* CRM field mapping validated against sample calls.
* On-call ownership and rollback plan documented.
* Stakeholder sign-off recorded.

## Rollback triggers

Rollback production launch if any of these appear:

* sustained webhook processing failure rate above internal threshold
* CRM mismatch between call outcomes and logged records
* abnormal spam flags or carrier rejection on numbers
* critical script regression causing compliance risk

<Warning>
  Knowledge bases, prompts, and action tools are environment-scoped artifacts. Never allow draft sandbox content to serve live production callers.
</Warning>