Skip to main content

Customer Support

AI-powered email support — inbound customer emails are processed with full org knowledge base context, generating intelligent replies or escalating to your team.

How It Works​

Customer Support AI reuses the same email architecture as Lira's inbound reply engine:

  1. Configure a custom support domain (e.g. support@yourcompany.com) via the Resend domain API
  2. Inbound emails are captured by AWS SES → stored in S3 → published to SNS
  3. SNS triggers POST /lira/v1/email/inbound
  4. Backend decodes the JWT reply token, loads thread history, runs Qdrant vector search against the org knowledge base
  5. GPT-4o-mini generates a grounded response or escalates to an org admin

Email Configuration​

# Register custom sending domain
POST /lira/v1/email/domain
Authorization: Bearer <jwt>

{
  "domain": "yourcompany.com"
}

# Check domain verification status
GET /lira/v1/email/domain/status

# Update email config (AI reply toggle, notification prefs)
PUT /lira/v1/email/config
{
  "aiReplyEnabled": true,
  "escalationEmail": "team@yourcompany.com"
}

Thread Management​

# List all email threads
GET /lira/v1/email/threads

# Get thread with full message history
GET /lira/v1/email/threads/:threadId

Knowledge Base Grounding​

AI replies are grounded in your organization's knowledge base:

  • Website crawl — Crawlee scrapes your org's website; pages are summarised and embedded in Qdrant
  • Document uploads — PDFs and DOCX files are parsed, chunked, and embedded

Before generating a reply, the system performs a vector search to find the most relevant knowledge base entries, injecting them into the system prompt for accurate, contextual responses.

Escalation​

When the AI cannot confidently answer a customer query (low relevance scores from vector search), it automatically escalates to the configured escalation email address with the full thread context.