Skip to content

datazen-vn/datazen-ai

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 

Repository files navigation

Datazen-AI

Vietnam-first AI chatbot SaaS for social commerce, customer support automation, ecommerce conversations, and knowledge base RAG.

AI Chatbot SaaS Vietnam-first Knowledge Base RAG Human Takeover

Datazen-AI helps businesses answer customer messages 24/7 across social messaging channels, ground AI replies in approved business knowledge, recommend products from catalogs, and hand conversations to human staff when needed.

It is designed for ecommerce shops, F&B brands, clinics, education providers, local service businesses, and teams that sell or support customers through Facebook, Instagram, Zalo, and chat-first channels.

Table of Contents

Why Datazen-AI

Most small and mid-sized businesses lose time answering the same questions again and again: price, stock, size, booking, delivery, refund policy, service details, opening hours, and follow-up support.

Datazen-AI turns those repeated conversations into a supervised AI workflow:

Problem Datazen-AI approach
Customers expect fast replies AI chatbot automation answers routine questions 24/7
Staff repeat the same answers Knowledge base RAG retrieves approved answers from documents and FAQs
Product questions need catalog context Ecommerce product search grounds replies in product, variant, price, stock, and category data
AI should not handle every situation Human takeover inbox lets staff step in, transfer, release, and continue conversations
SaaS operations need control Multi-tenant isolation, subscriptions, quotas, usage tracking, audit logs, and token health workflows

Who It Helps

Ecommerce and retail

Datazen-AI helps shops answer questions about product availability, size, color, variant, shipping, warranty, and returns while keeping replies connected to product data.

F&B and local brands

Restaurants, coffee shops, salons, and local chains can share menu details, branch information, opening hours, booking intent, and service details without forcing staff to reply manually to every repeated question.

Clinics, education, consulting, and services

Service teams can explain packages, pre-qualify leads, collect follow-up details, and escalate sensitive or high-value conversations to staff.

Support and operations teams

Managers get an inbox, audit trail, usage visibility, and handoff workflow instead of a black-box chatbot.

Product Capabilities

Multi-channel AI conversations

Datazen-AI is built for social messaging workflows where customers already ask questions.

  • Facebook Messenger chatbot automation
  • Instagram DM automation
  • Facebook comment reply automation
  • Zalo OA messaging flows
  • Telegram bridge and staff inbox workflows
  • Webhook-driven tenant routing and platform delivery

Knowledge Base and RAG chatbot

The bot can answer from approved business knowledge instead of relying only on a generic prompt.

  • Upload documents, FAQs, menus, policies, catalogs, and guides
  • Extract, chunk, embed, and search knowledge
  • Use pgvector-backed semantic retrieval
  • Store files and rendered assets with S3-compatible storage such as Cloudflare R2
  • Review document status, readiness, chunks, and knowledge audit events

Ecommerce chatbot and product search

Datazen-AI connects conversations to product data so replies can be useful for buying decisions.

  • Product catalog, variants, categories, prices, and stock
  • Product search function calling
  • Shopping-oriented response templates
  • Catalog-backed answers for fashion, cosmetics, F&B, retail, and service catalogs

Human takeover inbox

AI can automate routine conversations, but staff keep control.

  • Conversation inbox
  • Customer profile sidebar
  • AI context sidebar
  • Takeover and release bot control
  • Staff transfer and manual reply flows
  • Read state, typing state, conversation events, and realtime updates

CRM and customer data foundation

Datazen-AI includes a CRM foundation for customer-aware chatbot workflows.

  • Customer records and channels
  • Interactions and timelines
  • Notes, tasks, tags, and segments
  • Consent and audit models
  • Encrypted PII support
  • AI tool handler for saving user information during chat

Subscription, quota, and usage control

The platform is built as a SaaS foundation, not a single-tenant automation script.

  • Plans, features, subscriptions, quotas, payments, and addons
  • OpenAI usage tracking
  • Tenant-level quota checks
  • Admin provisioning and operational dashboards
  • Audit logs and token health workflows

Use Cases

Industry Common questions Datazen-AI workflow
Fashion ecommerce "Is this available in size M?", "Do you have this in black?", "How much is shipping?" Product search + policy knowledge + human handoff for order details
F&B "Can I see the menu?", "Are you open tonight?", "Can I book a table?" Menu knowledge + branch info + lead capture
Clinics and spa "How much is this service?", "Which package fits me?", "Can someone call me?" Service FAQ + pre-qualification + CRM capture + staff handoff
Education "What course should I take?", "What is the schedule?", "How much is tuition?" Course knowledge + lead qualification + follow-up
Support teams "Where is my order?", "How do I return this?", "I need a human" Knowledge base answer + escalation workflow

Architecture

Datazen-AI combines a Laravel + Vue SaaS control plane with a NestJS AI runtime and a lightweight mobile staff inbox.

Layer Responsibility Stack
SaaS control plane Tenant management, onboarding, subscriptions, social connection OAuth, knowledge UI, bot config, inbox, billing, admin tools Laravel, Vue 3, Inertia, Tailwind CSS, Reverb
AI runtime Webhooks, queues, tenant routing, OpenAI orchestration, RAG, function calling, streaming responses, platform adapters NestJS, TypeScript, Prisma, TypeORM, BullMQ, Redis
Mobile staff inbox Staff inbox workflows and notification entry points React, TypeScript, Vite, Telegram Mini App

High-level runtime flow:

Customer message or comment
  -> platform webhook
  -> tenant routing
  -> queue, debounce, and preprocessing
  -> moderation and intent detection
  -> agent runtime, prompt templates, and tools
  -> OpenAI response or streaming response
  -> knowledge, product, or CRM function calls when needed
  -> platform delivery
  -> dashboard, inbox, audit, quota, and notifications

Data model:

  • Central database: tenants, subscriptions, platform metadata, capabilities, prompt templates, OpenAI model/project metadata, and shared configuration.
  • Tenant databases: connected pages, bot configuration, conversations, knowledge documents, ecommerce data, CRM data, notifications, and audit events.
  • Redis: queues, routing cache, quota checks, debounce windows, realtime coordination, and runtime cache.

Tech Stack

Area Technology
SaaS app Laravel, PHP, Vue 3, Inertia.js, Tailwind CSS
Runtime API NestJS, TypeScript, Prisma, TypeORM
AI OpenAI, prompt templates, function calling, streaming responses
Retrieval PostgreSQL, pgvector, embeddings, RAG
Queue/cache Redis, BullMQ
Realtime Laravel Reverb
Storage Cloudflare R2 / S3-compatible object storage
Staff mobile inbox React, Vite, Telegram Mini App

Deployment Shape

Typical production deployments use containerized services:

  • SaaS web app
  • Queue workers
  • Scheduler
  • Realtime server
  • Message subscriber
  • AI runtime instances
  • Redis
  • PostgreSQL with pgvector
  • Reverse proxy and TLS termination

All production secrets, API keys, database credentials, and platform tokens should be configured through private environment variables. They should never be committed to GitHub.

Security Principles

  • Tenant data isolation
  • API-key protected internal services
  • Encrypted platform tokens and sensitive customer data
  • Webhook verification
  • Usage and quota controls
  • Audit-friendly operational records

Status

Datazen-AI has implemented core foundations for multi-tenant chatbot SaaS, knowledge base/RAG, ecommerce search, social messaging, inbox operations, subscription/quota tracking, and AI runtime orchestration.

Some advanced areas, such as CRM automation depth, channel rollout scope, analytics breadth, and deployment automation, may vary by environment and should be validated for each production launch.

FAQ

Is Datazen-AI only a chatbot?

No. It is a SaaS platform around chatbot operations: onboarding, social connections, knowledge management, prompt/runtime configuration, inbox, audit logs, quota usage, and staff handoff.

Does Datazen-AI support knowledge base answers?

Yes. Datazen-AI is designed as a knowledge base chatbot and RAG chatbot. Business documents, FAQs, policies, menus, catalogs, and guides can be processed into searchable knowledge for AI responses.

Can staff take over a conversation?

Yes. Human takeover is a core workflow. Staff can supervise conversations, take over, reply manually, transfer, and release the bot when appropriate.

Is it built for Vietnamese businesses?

Yes. Datazen-AI is Vietnam-first and built around social commerce workflows common to Vietnamese ecommerce, F&B, service, clinic, education, and local business teams.

Is this repository the full product source code?

No. This public repository is a product landing and GitHub SEO repository. The production application code is maintained separately.

Keywords

AI chatbot SaaS, Vietnamese AI chatbot, Facebook Messenger chatbot, Instagram DM automation, Facebook comment automation, Zalo chatbot, ecommerce chatbot, knowledge base chatbot, RAG chatbot, multi-tenant chatbot platform, OpenAI chatbot platform, Laravel Vue SaaS, NestJS chatbot, customer support automation, social commerce automation, human takeover inbox, CRM chatbot, product search chatbot.

License

Proprietary. Contact Datazen for commercial use, partnership, and deployment inquiries.

Releases

No releases published

Packages

 
 
 

Contributors