Vapi's Architectural Approach

Modular Sandwich Platform

Provider flexibility:

• 10+ STT providers integrated
• 15+ LLM models supported
• 8+ TTS providers available
• Swap through dashboard configuration

Orchestration layer:

• Streaming between all components
• Automatic failover during provider outages
• Load balancing across endpoints
• Unified monitoring and logging

Infrastructure abstraction:

• Developers choose providers, not infrastructure
• Vapi manages servers, scaling, failover
• Focus on conversation logic, not DevOps

Why Vapi Chose Modularity

Flexibility: Different use cases need different provider combinations Innovation: New providers launch frequently, modularity enables adoption Optimization: Developers A/B test providers on real traffic Control: Developers optimize cost vs quality vs latency independently Future-proof: Add speech-to-speech as another provider option when mature

Migration Path to Speech-to-Speech

When speech-to-speech matures:

1. Add as provider option in Vapi platform
2. Enable A/B testing vs Sandwich configurations
3. Developers gradually migrate use cases benefiting from end-to-end
4. Maintain Sandwich for use cases requiring customization
5. Platform supports both architectures simultaneously