Answerify - AI-Powered Customer Support Platform

Answerify is a modern customer support platform that uses AI to automatically answer customer questions, saving your team time while maintaining quality responses.

Features

🤖 AI-Powered Auto-Replies

Answerify uses Google Gemini AI to generate intelligent, contextually relevant responses based on your knowledge base. The system:

• Analyzes incoming customer emails using semantic search
• Generates accurate responses with source citations
• Learns from human feedback to improve over time

📊 Intelligent Knowledge Base Management

• Vector Embeddings: Content is chunked and embedded for semantic search
• Multiple Data Sources: Support URLs, documentation, FAQs, and more
• Internal Knowledge Base: Auto-learns from approved responses
• Real-time Indexing: New content is automatically indexed when added

✉️ Email Thread Management

• Unified Inbox: View all customer conversations in one place
• Real-time Updates: New emails appear instantly via Supabase Realtime
• Status Tracking: Open/closed ticket management
• Rich Text Editor: Compose replies with Tiptap-powered formatting

🎯 Intent Detection & Tagging

The system automatically classifies incoming emails into categories:

Category	Icon	Description
Billing	💳	Payment and invoice questions
Bug Report	🐛	Software issues and errors
Feature Request	✨	Enhancement suggestions
General Support	💬	Standard support queries
Account Access	🔐	Login and access issues
Account Settings	⚙️	Configuration questions
Account Deletion	🗑️	Data removal requests
Refund Request	💰	Money return inquiries
Payment Failed	❌	Transaction failures
Subscription Change	🔄	Plan modifications
Invoice Request	📄	Billing document requests
Performance Issue	⚡	Speed and responsiveness
Integration Issue	🔌	Third-party connectivity
Data Export	📤	Information retrieval
Upgrade Inquiry	🚀	Product upgrades
Demo Request	🎯	Product demonstrations
Onboarding Help	🎓	Getting started assistance
How-To Question	❓	Usage instructions
Security Concern	🛡️	Safety and privacy
Abuse Report	🚫	Policy violations
Privacy Request	🔒	GDPR and data requests
Complaint	😤	Dissatisfaction
Compliment	🌟	Positive feedback
Spam / Irrelevant	🚮	Noise filtering

🔄 Workflow Automation

Create automated rules to handle common scenarios:

Trigger Types:

• subject_contains - Match emails by keyword in subject
• sender_domain - Match emails from specific domains
• has_tag - Match emails with specific intent tags
• any_email - Trigger on all incoming emails

Automation Steps:

• Add Tag: Automatically categorize emails
• Auto Reply: Send predefined responses
• Escalate: Flag high-priority items
• Call Webhook: Integrate with external systems

🛡️ Autopilot Mode

Smart automation with human oversight:

• High Confidence (≥65%): Auto-sends AI-generated replies
• Low Confidence: Creates drafts for human review
• Configurable Threshold: Adjust sensitivity per organization
• Safety Guards: Never sends empty or unclear responses

📝 Learning Loop

Continuous improvement from human feedback:

1. Agent edits an AI draft
2. System tracks the human response
3. If approved, the response is chunked and embedded
4. Future similar queries use this learned content
5. Reply quality improves over time

👥 Team Collaboration

• Role-Based Access: Member, Admin, and Owner roles
• Invite System: Add team members by email
• Organization Management: Multiple workspaces supported

---

Architecture

Database Schema

┌─────────────────┐ ┌─────────────────┐
│ organization │────<│ member │
└─────────────────┘ └─────────────────┘
│
├────<│ thread │
│ └─────────────────┘
│ │
│ │
├────<│ email │
│ └─────────────────┘
│
├────<│ reply │
│ └─────────────────┘
│
├────<│ datasource │
│ │
│ │
│ └────<│ section │
│ └─────────────────┘
│
├────<│ workflow │
│ └─────────────────┘
│
└────<│ mcp_server │
└─────────────────┘

Key Tables

Table	Description
organization	Tenant workspace with settings
member	User membership with roles
thread	Email conversation thread
email	Individual messages
reply	AI-generated or human drafts
datasource	Content sources (URLs)
section	Chunked content with embeddings
workflow	Automation rules
mcp_server	MCP tool integrations

API Flow

Inbound Email
│
▼
┌─────────────────┐
│ Webhook Handler │
└────────┬────────┘
│
▼
┌─────────────────┐
│ Intent Detection│ ───► Tags added to thread
└────────┬────────┘
│
▼
┌─────────────────┐ ┌─────────────────┐
│ Vector Search │────>│ Match Sections │
└────────┬────────┘ └─────────────────┘
│
▼
┌─────────────────┐
│ Generate Reply │
└────────┬────────┘
│
▼
┌─────────────────┐
│ Workflow Runner │
└────────┬────────┘
│
┌────┴────┐
│ │
▼ ▼
┌───────┐ ┌────────┐
│ Auto │ │ Draft │
│ Send │ │ Review │
└───────┘ └────────┘

---

Environment Variables

Required

Variable	Description
NEXT_PUBLIC_SUPABASE_URL	Supabase project URL
NEXT_PUBLIC_SUPABASE_ANON_KEY	Supabase anonymous key
SUPABASE_SERVICE_KEY	Supabase service role key
NEXT_PUBLIC_BASE_URL	Application base URL
RESEND_API_KEY	Resend email API key
GOOGLE_GENERATIVE_AI_API_KEY	Google Gemini API key

Optional - Cloudflare AI Gateway

Variable	Description
CLOUDFLARE_ACCOUNT_ID	Cloudflare account ID
CLOUDFLARE_GATEWAY_NAME	AI Gateway identifier
CF_AIG_TOKEN	AI Gateway auth token

---

Development

Setup

# Install dependencies
pnpm install
# Copy environment variables
cp .env.example .env.local
# Start development server
pnpm dev

Database Migrations

Apply migrations via Supabase CLI or dashboard:

supabase db push

Key Scripts

Command	Description
pnpm dev	Start development server
pnpm build	Build for production
pnpm lint	Run ESLint
pnpm tsc	TypeScript type check

---

Configuration

Autopilot Settings

-- Disable autopilotUPDATE organization
SET autopilot_enabled = false
WHERE id ='<org-id>';
-- Adjust confidence thresholdUPDATE organization
SET autopilot_threshold =0.75WHERE id ='<org-id>';

Tone Policy

Customize AI response tone:

UPDATE organization
SET tone_policy ='friendly'WHERE id ='<org-id>';

---

Security

• Row-Level Security (RLS): All tables enforce organization-based access
• Service Role: Webhooks use service key for elevated access
• Email Sanitization: All HTML content is XSS-sanitized
• API Authentication: Demo endpoints require Bearer tokens

---

Support

For issues and feature requests, please open a GitHub issue.