AI & Voice Integration

Overview

Lexi uses three main AWS services for AI and voice processing:

  1. Amazon Bedrock - AI conversation (Claude 3 Haiku or Nova Lite)
  2. Amazon Transcribe - Speech-to-Text (STT)
  3. Amazon Polly - Text-to-Speech (TTS)

Speaking Flow Process

User speaks
    ↓
Frontend records → sends audio stream
    ↓
Lambda receives audio
    ↓
Transcribe: Audio → Text
    ↓
Bedrock: Text + Context → AI Response
    ↓
Comprehend: Language analysis
    ↓
Polly: Response → Audio
    ↓
Lambda sends audio to frontend
    ↓
Frontend plays audio
    ↓
Save session to DynamoDB

AI & Voice Content

AI & Voice is divided into 3 main parts:

  1. Amazon Bedrock - AI conversation

    • List available models
    • Invoke models
    • Monitor usage
    • Cost optimization

  2. Amazon Transcribe - Speech-to-Text

    • Batch transcription
    • Real-time streaming
    • Check job status
    • Get transcription result

  3. Amazon Polly - Text-to-Speech

    • Synthesize speech
    • List available voices
    • SSML support
    • Cost optimization

Checklist

After completing the AI & Voice section, you should:

  • Understand the complete Speaking Flow process
  • Know how to use Amazon Bedrock for AI conversation
  • Understand how Transcribe converts speech to text
  • Know how Polly synthesizes speech from text
  • Understand how to integrate these 3 services with Lambda
  • Understand costs and optimization strategies

📸 TODO: Add detailed Speaking Flow diagram

📸 TODO: Add screenshot of Bedrock model invocation

Next Steps

You’ve completed the AI & Voice section! Continue to CI/CD Pipeline to set up automated deployment.