Can I vibe code Gladly?

Build a customer support inbox using Next.js 14, TailwindCSS, Supabase, and LangChain that mimics Gladly's 'people, not tickets' core concept.

Data Architecture:
1. 'customers': id, email, phone, shopify_customer_id, full_name, total_spent, orders_count, ltv_score, created_at.
2. 'conversations': id, customer_id, channel (email, chat, sms, voice), status (open, pending, resolved, AI_handled), assigned_agent_id, created_at, updated_at.
3. 'messages': id, conversation_id, sender_type (customer, agent, ai_bot), body, channel, metadata (jsonb for carrier logs, voice transcript, or email headers), created_at.
4. 'guides': id, name, system_instructions, routing_rules (jsonb), scope.

Key Functionality:
1. Unified Timeline View: Never split customer inquiries into isolated tickets. Route all incoming messages (chat, email via SendGrid, SMS via Twilio) into a single chronological feed attached to 'customer_id'.
2. Shopify Integration: On conversation load, query the Shopify Admin API for the customer's recent orders, shipping status, tracking links, and lifetime spend. Display this in a sidebar panel next to the message thread.
3. Autonomous AI Resolution (Sidekick equivalent):
   - Use OpenAI GPT-4o function calling to process incoming messages.
   - Define tools: fetch_order_status(order_id), initiate_return(order_id, item_id), issue_discount_code(customer_id, percentage), transfer_to_human(reason).
   - If AI confidence score > 0.85 and tool execution succeeds, auto-resolve conversation and tag status as 'AI_handled'.
   - If AI encounters sentiment threshold < 0.3 or customer asks for human, emit a Realtime Supabase socket event to update the agent queue immediately with 'pending' status.
4. Real-time Agent UI:
   - Built with Supabase Realtime channels to allow multi-agent typing indicators, presence, and lock status (preventing two agents from replying simultaneously).
   - Rich text editor for agent responses with quick-insert macros for standard e-commerce queries.

Out of Scope:
- Native telephony/SIP trunk handling (use Twilio WebRTC client SDK for voice handling).
- Multi-tenant enterprise SSO / SAML configuration.
- Complex workforce management and shift scheduling analytics.

Failure Modes to Guard Against:
- Infinite loop prevention: Ensure AI bot does not reply to automated auto-responders or bounce messages.
- Race conditions: Implement optimistic locking on conversation assignment to prevent collision between AI auto-resolution and human agent pickup.

$ each button prefixes agent-specific run instructions · build your own product, never copy proprietary code, trademarks or designs