AURA Voice Assistant
Private AI voice workflow backend that connects Vapi tool calls to onboarding email and calendar actions with quick responses and background processing.
Shows practical AI workflow integration beyond demos: voice-agent tool calls, webhook reliability, email, calendar scheduling, and secret-driven deployment.
AURA Voice Assistant is a private AI automation webhook service built around Vapi.ai tool calls. The backend receives structured voice-agent actions, returns a quick success response so the call experience does not stall, then completes the real work in the background.
The public value is practical AI integration. Instead of a chatbot demo, AURA connects a voice assistant to transactional onboarding email through Postmark and calendar scheduling through Google Calendar, with secrets and service credentials isolated through environment configuration.
Tech Stack
Key Features
- Express webhook endpoint for Vapi.ai tool-call payloads
- Immediate JSON response path to avoid blocking the voice assistant
- Background processing for real email and calendar actions
- Postmark onboarding email flow for send_link intent
- Google Calendar invite creation for book_call intent
- Health endpoint and environment-driven service configuration
Technical Highlights
- Node.js and Express service with JSON middleware
- Vapi tool-call parsing and response formatting
- Postmark ServerClient integration for transactional email
- Google Calendar API service-account flow for event creation
- process.nextTick used to decouple webhook acknowledgement from slower side effects
- Public-safe architecture representation for private automation work
Architecture
Voice-Agent Webhook
- Receive a Vapi message payload
- Extract tool call ID, business name, email, phone, and intent
- Return the tool result immediately to keep the call moving
Automation Actions
- send_link routes to the Postmark onboarding email path
- book_call routes to Google Calendar event creation
- Background errors are logged without breaking the initial voice response
Challenges & Solutions
Designing AI voice automation around latency-sensitive webhooks
Keeping external side effects out of the synchronous response path
Handling transactional email and calendar credentials responsibly
Explaining private automation work without exposing sensitive deployment details
Gallery
