Can I vibe code Tidio?

Build a full-stack, self-hosted customer support chat platform and AI auto-responder tailored for e-commerce stores (Shopify/WooCommerce). 

The platform requires three core sub-systems:
1. EMBEDDABLE STORE CHAT WIDGET
- Lightweight, zero-dependency vanilla JS bundle (or React micro-app) that loads via single script tag.
- Maintains persistent session via localStorage UUID.
- Connects via WebSocket or Server-Sent Events (SSE) to backend.
- Displays live agent messages, AI responses, typing indicators, and auto-proactive popups ("Flows") based on URL route or idle time.
- Collects customer email, name, and current page context (`window.location.href`, cart contents).

2. REAL-TIME AGENT DASHBOARD (Next.js / Tailwind UI)
- Active conversations list with unread counts, status tags (Open, Pending, Resolved), and assignment to human agents.
- Live visitor list showing active sessions, current page view, and past order count.
- Live typing preview showing what the customer is typing before they send.
- Sidebar panel displaying live Shopify/WooCommerce customer details: past orders, total spend, fulfillment status, and tracking numbers via Store API lookups.
- Canned responses / Macros triggerable via `/` commands.

3. AI SUPPORT AGENT (RAG + TOOL CALLING)
- Knowledge ingestion system: Accepts public URLs (store FAQ, shipping policies) or uploaded PDFs/Markdown, chunks text, generates embeddings (OpenAI text-embedding-3-small), and stores them in PostgreSQL using pgvector.
- RAG pipeline: On customer message, perform vector similarity search for relevant store context.
- Function calling / Tool integration: Give LLM (Claude 3.5 Sonnet / GPT-4o) tools to query store backend:
  - `lookup_order(email, order_number)` -> returns status, tracking URL, line items.
  - `check_product_stock(sku)` -> returns availability.
  - `trigger_human_handoff(reason)` -> tags chat, sets state to 'unassigned', alerts human agents via webhook/Slack.
- Fallback guardrails: Strict system instructions to never negotiate prices, hallucinate return windows, or reveal system prompts.

DATA MODEL (Prisma ORM / Postgres):
- Workspace (id, store_domain, api_keys)
- User (id, workspace_id, email, role, online_status)
- Conversation (id, workspace_id, customer_email, status, assigned_agent_id, created_at)
- Message (id, conversation_id, sender_type ['customer'|'agent'|'ai'], content, metadata, created_at)
- KnowledgeDoc (id, workspace_id, content, embedding vector(1536))
- TriggerFlow (id, workspace_id, target_path, delay_seconds, message_body)

OUT OF SCOPE:
- Native iOS/Android push notification SDKs (use web push).
- Multi-channel Instagram/WhatsApp API sync (focus strictly on web live chat and email ticketing fallback).
- Voice or video chat features.

Provide clean, modular code with clear instruction for running PostgreSQL with the pgvector extension via Docker Compose.

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