1. IraVoice API
Epicode docs
  • IraVoice API
    • IraVoice API (2.0.0)
    • makecall (Make outbound call)
      • Make Call
    • makecalldirect
      • Make Call Direct
    • dropcall
      • Drop Call
    • stopstream
      • Stop Stream
    • bridgecalls
      • Bridge Calls
    • transfercall
      • Transfer Call
    • Schemas
      • BridgeCalls
      • DropCall
      • HTTPValidationError
      • MakeCall
      • MakeCallDirect
      • StopStream
      • TransferCall
      • ValidationError
  • Epicode Onboarding Guide
    • 📞 IraVoice Onboarding Guide
    • Sandbox Environment
  1. IraVoice API

IraVoice API (2.0.0)

IraVoice is a bidirectional media server that brings together Dialer, Recorder, and BotStream modules to enable seamless telephony services for both legacy CX systems and modern conversational AI applications.

IraVoice - Key Advantages#

1. High-Concurrency AI Orchestration#

Problem: Low-level audio processing workloads (resampling, noise suppression, VAD), whether DSP or ML-based, bottleneck AI orchestration layers in Python/Node.js, constraining concurrency and scale.
IraVoice Approach:
Offloads all audio processing to a high-performance C++ media layer
Handles:
Real-time resampling (Developers can independently configure audio sampling rates for both IraVoice → AI Orchestration, & AI Orchestration → IraVoice).
Deep Learning based fast noise suppression.
Server-side VAD with "speech" / "silence" events sent over websocket.
Result:
~10x higher concurrency per instance of orchestrator.
Up to 90% lower compute cost for orchestration layer.
Cleaner orchestration code (no DSP handling)

2. Unified Multi-Trunk Orchestration#

Problem: Multi-vendor telephony = fragmented APIs, inconsistent events/CDRs, manual failover.
IraVoice Approach:
Single API across multiple SIP trunks and geographies.
Campaign Manager portal to configure campaigns with multiple trunks, routing rules, and DID range to be used.
Built-in:
CPS-aware routing & queuing
Capacity-based load distribution
Automatic failover on trunk outages
Horizontal scaling of dialer clusters
Result:
High availability across providers
Zero-touch failover
One integration, multi-operator backend
Reliable large-scale campaign execution

3. Real-Time Telemetry#

Problem:
Operational blind spots in voice systems make it difficult to diagnose failures (e.g., WebSocket issues), quantify latency, or detect degraded bot performance in real time—leading to delayed debugging and poor user experience.
IraVoice Approach:
Deep integration with IraWatch for live observability
Real-time dashboard with a 30-second moving window tracking:
WebSocket connection failures with stage-level granularity (DNS, TCP, SSL, WSS)
Average bot response delay (user silence → bot audio received)
Bot error rate (responses exceeding 4 seconds)
Correlates signaling, media, and orchestration-layer metrics in a single view
Result:
Immediate detection of latency spikes and failure patterns
Faster root-cause analysis with protocol-level visibility
Proactive performance monitoring for conversational AI systems
Improved SLA adherence and end-user experience
Modified at 2026-05-13 07:22:15
Next
Make Call
Built with