Twilio & Vonage
The OVA application integrates seamlessly with telephony providers like Twilio and Vonage through the OVA Connector. This integration enables businesses to handle customer calls efficiently using AI-powered virtual agents while providing a smooth handoff to human agents when needed.
The backend processing, combined with real-time voice streaming and session management, ensures a responsive and context-aware call experience.
This diagram illustrates the integration of the OVA application with popular telephony providers like Twilio and Vonage, allowing OVA to handle voice interactions seamlessly. Here’s a detailed breakdown of each component and the integration process:
Key Components and Flow
- Telephony Integration (Twilio/Vonage):
OVA connects with telephony services such as Twilio and Vonage, which manage the initial call setup and handle the voice data streaming.
Incoming calls from customers are routed through these providers, leveraging their robust telecommunication infrastructure for reliable voice communication.
- OVA Connector:
The voice data from Twilio/Vonage is sent to OVA through the OVA Connector, a key integration component that facilitates real-time data exchange.
The connector handles the call audio using protocols like WebRTC for live streaming and SIP for call session management.
OVA Backend:
The OVA Backend processes the call data and powers the virtual agent capabilities using AWS services:
AWS Lambda: Executes serverless functions to process call data and handle backend logic.
AWS S3: Stores call recordings and relevant files securely for later access.
Amazon DynamoDB: Manages session data and tracks the conversation context, allowing OVA to maintain state throughout the interaction.
- OVA Main Stack:
- The core AI processing occurs in the OVA Main Stack, which includes:
Speech-to-Text: Converts the audio input from the caller into text for analysis.
Language Model: Analyzes the text input to understand the user’s intent and generate appropriate responses.
Text-to-Speech: Converts the generated response back into audio, enabling OVA to reply to the caller naturally.
- Agent Integration (Fallback Option):
If the virtual agent determines that the call needs human intervention (e.g., complex requests), it triggers a transfer to a live agent.
The call is routed back through Twilio/Vonage into a queue, where it waits for an available agent.
The agent then receives the call with full conversation context, minimizing the need for the customer to repeat themselves.
- OVA Frontend:
The OVA Frontend provides a user interface for administrators and agents to monitor and manage the call sessions.
It displays real-time session data, call logs, and performance metrics, giving a comprehensive view of the virtual agent interactions.
- Telephony Integration (Twilio/Vonage):
Call Journey Example: Customer Support Scenario with Twilio
Step 1: Incoming Call via Twilio
A customer dials a support number, and the call is received by Twilio, which handles the initial voice data streaming.
Step 2: Voice Data Routed to OVA
The call audio is sent to the OVA Connector, which streams the voice data in real-time to the OVA Backend.
Step 3: Virtual Agent Interaction
- The OVA virtual agent engages with the caller:
Speech-to-Text converts the caller’s voice into text.
The Language Model processes the text, identifies the intent (e.g., a request for product information), and generates a suitable response.
Text-to-Speech converts the response back into audio, replying to the caller (e.g., “Let me provide you with the details you requested.”).
Step 4: Human Agent Transfer (If Needed)
If the virtual agent cannot resolve the request, it triggers a transfer to a live agent.
The call is placed in the Twilio-managed queue and then transferred to an available human agent with full session context.
Step 5: Monitoring and Logging
The entire interaction is logged and tracked in the OVA Frontend, where administrators can review call logs and monitor the performance of both the virtual and human agents.
For any custom development and further queries please write us at [email protected].