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

# Architecture & Data Strategy

> How Avoca calls your APIs in real time, and what that means for latency and resiliency

# Architectural Design

### Real-Time API Integration

Avoca's AI agents call your APIs **synchronously, in real time, during live conversations**. When a caller asks for an appointment, the agent looks up the customer, checks availability, and books the job against your system while the caller is still on the line. The data types fetched live include:

* **Live availability** – Appointment slot availability
* **Customer lookup** – Current customer records and account status
* **Service eligibility** – Eligibility checks based on your business rules
* **Instant confirmations** – Immediate booking confirmations
* **Booking modifications** – Cancellations, reschedules, and changes

Avoca does not maintain a local cache of your availability, pricing, or customer data for custom integrations — every conversation works from fresh responses to keep bookings accurate. Relatively static reference data (service catalogs, business hours, service-area rules) is typically configured in Avoca during onboarding and updated through the joint integration process rather than fetched per call.

### Latency Expectations

Because your API sits in the live conversation path, response time directly affects the caller experience:

* **Target response times:** under 200 ms for availability checks; under 500 ms for booking operations
* **Timeout:** Avoca times out requests after **10 seconds**
* **Retries:** failed requests are automatically retried once after a short delay; repeated failures trigger alerts to Avoca's integration team

The AI agent is designed to hold the conversation naturally while a request is in flight, but consistently slow endpoints will produce awkward pauses — keep hot paths fast.

### Hosting & Connectivity

* Your APIs can be hosted on any cloud or on-premise environment reachable over the public internet via **HTTPS/TLS** — no VPN or special networking is required
* Avoca's platform is hosted on US-based cloud infrastructure; for US operations, hosting your API in a US region minimizes latency
* We defer to your organization's data residency requirements for customer data

### Failover & Redundancy (recommendations for your API)

* Deploy redundantly with automatic failover where possible — if your API is down, the AI can still take messages and capture bookings for manual entry, but real-time booking is degraded
* Expose a health check endpoint so issues can be detected quickly during testing and operations
* Return clear, fast error responses rather than hanging — Avoca's timeout/retry handles transient failures gracefully

<hr />