Product Glossary

This glossary defines the core concepts, product lineup, and technical terms of the Lovrabet platform, so everyone works from the same vocabulary.

Products & Brands

Term	Description
Lovrabet	An AI-native platform for generating enterprise systems. Connect a database, let AI understand the data, and a working business system is generated automatically. Not low-code, not BI, not a chatbot.
Qizhi Yuntu (启智云图)	The company brand. Lovrabet is its flagship product.
Lovrabet (云兔)	The customer-facing brand at runtime — the platform name enterprises see when running their business on it.
RabetBase	The enterprise-grade intelligent foundation, built in three layers: the enterprise base layer (permissions/security/governance), the AI business comprehension engine (DB Agent + semantic layer), and the open extension layer (SDK/CLI/MCP/OpenAPI).
AI-Native	AI is built into the system, not bolted on. AI is the core engine underneath, not an add-on feature. The stronger the model, the stronger the platform — the opposite of the "one prompt generates an app" approach.
AI-Embedded	AI attached to an existing system as an add-on feature — the traditional approach. Lovrabet is not this.

Core Philosophy

Principle	Description
AI-Ready Data (Fuel)	Make AI understand your enterprise data. Even a ten-year-old database with no documentation and no primary or foreign keys can be understood automatically by DB Agent. Data silos (CRM/ERP/e-commerce/warehousing) are connected in one click.
AI-Driven Processes (Engine)	Let AI run the enterprise. Three levels: AI-assisted development (available now — text-to-page generation), AI-assisted runtime (in progress — data Q&A pages), and AI-assisted process orchestration (planned — AI-generated workflows).
Awakening Legacy Assets	No rip-and-replace. Put dormant legacy data assets back to work.
Hybrid Coexistence	Standard products, AI self-service, and professional development coexist — each scenario uses the mode that fits.
Enterprise-Ready	Runtime governance must be in place: tenant isolation, permission boundaries, security, and compliance.
The 80/20 Rule	AI generates the 80% of standard features automatically; the 20% of custom requirements are built with Pro Code extensions.
The Evolution Flywheel	The system keeps evolving after delivery: business logic in production → AI distills knowledge through reverse engineering → runtime AI learns → high-frequency scenarios accumulate → continuous reuse and optimization. The more you use it, the smarter it gets.
PD Vibe Mode	Product managers and business users drive system evolution — shifting from "file a request → wait for dev → acceptance testing" to "build directly → iterate immediately".

Data & Models

Core Concepts

Term	Description
Application (App)	A project created on the platform, with a unique appCode. For example, "Customer Management System" is an app.
AppCode	The unique identifier of each app, in the format app-c4c89304. Required by almost every API call.
Dataset	The most central concept. It maps to a database table, but richer: database table + AI-understood business semantics + auto-generated API endpoints.
DatasetCode	The unique identifier of each dataset — a 32-character string. Accessed in the SDK as client.models.dataset_xxx.
tableName	The physical database table behind a dataset.
alias	A human-readable alias for a dataset, easier to work with in code. For example, client.models.customer is more intuitive than client.models.dataset_1000372.

Fields & Structure

Term	Description
Field	A column in a dataset, with attributes such as type (text/number/date/enum), required or not, primary key or not, enum options, and relations.
Business Entity	A business concept recognized by AI, such as "Customer", "Order", or "Product" — emphasizing business meaning beyond "table".
Semantic Layer	The business knowledge produced once AI understands your database. For example, AI knows cust_id is a customer ID and relates to the order table; this knowledge is automatically injected into API/SDK/MCP.
ER Diagram	The entity-relationship diagram, generated automatically by the platform. Access it at /app/{appCode}/data/er.

Multi-Database & Tenancy

Term	Description
Multi-Database	One app can connect to multiple databases at once. Customer data in MySQL, orders in PostgreSQL — AI understands them all in one model.
Multi-Tenant Isolation	When multiple organizations share a system, each tenant sees only its own data. Handled automatically by the data access layer.
Multi-Table JOIN	filter queries support cross-table relations in the relatedTable.field format. The SDK performs LEFT JOINs automatically, up to 5 levels deep.

Developer Toolchain

Tool	Description
OpenAPI	Standard HTTP endpoints callable from any language, with HMAC-SHA256 signature authentication. Ideal for backend integrations, mini-programs, and third-party systems.
TypeScript SDK @lovrabet/sdk	A type-safe data access toolkit for frontend/Node.js. Create a client with createClient(), then query data with client.models.xxx.filter().
Java SDK	A data access toolkit for Java backends, fitting the Java ecosystem (Spring Boot and more).
CLI @lovrabet/cli	The terminal tool. lovrabet init creates a project, lovrabet api pull pulls dataset configuration, and lovrabet dev starts the dev server.
MCP Server @lovrabet/dataset-mcp-server	Gives AI a pair of eyes that can see enterprise data. AI tools like Claude and Cursor understand dataset structure through it, raising code-generation accuracy from 60-70% to 95%+.
Skills	Rule packs installed into AI editors so AI better follows Lovrabet development conventions.

Authentication & Security

Auth Method	Environment	Description
WebAPI (Cookie)	Browser frontend	The browser sends cookies automatically — no extra configuration needed. API path: /api/{appCode}/{datasetCode}/{method}.
OpenAPI (AccessKey)	Server side	AccessKey + SecretKey generate an HMAC-SHA256 signature sent in an HTTP header. API path /openapi/data/{method}, POST only.

Term	Description
AccessKey / SecretKey	The API key pair. The AccessKey is a public identifier; the SecretKey must only be used server-side and must never be exposed in frontend code.
Token (pre-generated mode)	A pre-computed auth token valid for 10 minutes, ideal for temporary authorization and mobile scenarios.
SSO	An Enterprise Edition feature — employees sign in once to access all systems.
LDAP	Integrates your existing corporate identity system (such as Active Directory). An Enterprise Edition feature.

Data Operations

Common Methods

Method	Description	Notes
filter(params)	Query a list with conditions; supports where/orderBy/select/pagination	Most used
getOne(id)	Get a single record by primary key ID	—
create(data)	Create a new record	—
update(id, data)	Update records; up to 1000 per batch	—
delete(id)	Delete records; up to 1000 per batch	Not supported in OpenAPI mode
aggregate(params)	Aggregate statistics (SUM/COUNT/AVG), supports groupBy/having	WebAPI mode only
excelExport(params)	Export query results as an Excel file	WebAPI mode only
getSelectOptions(params)	Get the selectable values of an enum field	For frontend dropdowns

Filter Operator Quick Reference

Operator	Meaning	Example
$eq / $ne	Equals / not equals	{ status: { $eq: "active" } }
$gte / $lte	Greater than or equal / less than or equal	{ age: { $gte: 18 } }
$in	In a list	{ country: { $in: ["CN", "US"] } }
$contain	String contains	{ name: { $contain: "张" } }
$and / $or	Logical combination	{ $and: [...] }

Other Concepts

Term	Description
Soft Delete	Records are not physically deleted; a deleted field is flagged instead. Data is recoverable and the audit trail is preserved.
Rate Limits	Per app: 600 requests/minute, 10,000/hour, 100,000/day. Up to 100 records per batch operation.

Business Logic Extensions

Concept	Description
Backend Function / BFF	JS/TS functions running on Lovrabet servers, handling complex business logic that standard CRUD cannot.
Before Hook	Runs automatically before a data operation. Used for validation (deduplicating phone numbers), auto-fill (populating department/creator), and permission interception (tenant isolation).
After Hook	Runs automatically after a data operation. Used for data masking (hiding the middle four digits of a phone number), derived values (birthday → age), and related-data enrichment (order → customer info).
Standalone Endpoint (ENDPOINT)	An API endpoint not bound to any dataset. Ideal for cross-table logic, aggregate reports, and third-party integrations.
Custom SQL	Write SQL once, save it, and call it later by sqlCode. Ideal for multi-table joins, complex aggregation, and leaderboards. AI cannot write DELETE or DDL, preventing accidental damage.
sqlCode	The unique identifier of each custom SQL, in the format xxxxx-xxxxx. Called via client.sql.execute({ sqlCode: 'xxx' }).
safe()	An SDK convenience wrapper. Wraps an async operation, never throws, and returns { data, error } — no try-catch needed.
Pro Code Extension Points	Extend with real code instead of low-code building blocks: React child apps, BFFs, custom SQL, or third-party integrations via OpenAPI.

AI Capabilities

Capability	Description
DB Agent	The core technology. AI reads and understands databases automatically — no documentation or data dictionary required. Supports databases without primary or foreign keys, without data dictionaries, and heterogeneous multi-database setups.
Semantic Injection	Once DB Agent understands the data, business knowledge is automatically injected into API, SDK, and dev tools — no manual JOINs needed.
Text-to-Page Generation	Describe what you need in natural language and AI generates a set of pages automatically (list page + detail page + create form + edit form).
Vibe Coding	AI-assisted programming — describe the requirement in natural language and generate runnable code directly, with help from MCP.
Runtime AI	AI is there in day-to-day use. For example, on the data Q&A page ("What were this month's sales in East China?"), AI queries the database and returns the answer directly.
Incremental Sync	When the source database schema changes (new fields, type changes), no full re-parse is needed — changes sync in minutes.
Assetization	Turn SQL, BFFs, and page templates into reusable platform assets that other projects can use directly.

Pages & UI

Page Type	Description
RabetPage	The standard data management page (list + form), AI-generated or manually configured.
RabetSearch	Data search and filtering page with multi-condition combined queries.
RabetChats	The AI Q&A page for data and business questions — users interact with the system in natural language.
RabetReport	The AI report page — AI picks the visualization based on data characteristics and delivers insights.
List Page	Data table view with pagination/sorting/filtering/bulk actions, adjustable at any time.
Form	Create/edit form with field-level control (which fields are editable, which are read-only).
Dashboard	Data visualization board showing metrics, trends, leaderboards, and more.
Standard Suite	The baseline system generated by AI — CRUD management, basic reports, login and permissions, and audit logs. Your delivery starting point (v0).
Child App	A developer-built extension app covering roughly the 30% custom part. Integrated seamlessly with the parent app — users never notice the boundary.

Runtime & Permissions

Concept	Description
Dev Mode (开发态)	The stage where developers and product managers build features — AI-assisted development, full data visibility.
Runtime (运行态)	The stage where business users work with the system day to day — AI-assisted business, visibility limited to permitted data.
Proxy Layer	The data access control layer that enforces permission rules and tenant isolation. BFF/SQL cannot bypass permissions.
Data Permissions	Row-level data isolation — by department (own department and its sub-departments), by employee ownership (own data), or by role.
RBAC	Role-based access control: user → role → permission.
ABAC	Attribute-based access control, decided dynamically from user attributes, resource attributes, and environmental conditions.
RLS (Row-Level Security)	Row-level data isolation enforced at the database layer.

Editions & Pricing

Edition	Description
Professional Edition	For small and medium businesses. Up to 500 tables and 3 databases; excludes the advanced BFF runtime and SSO.
Enterprise Edition	For mid-to-large enterprises. Dedicated deployment, SSO/LDAP, high availability, full BFF, and asset version management.
AI Credits	The virtual quota for AI features. Covers development (text-to-page generation, DB Agent) and runtime (data Q&A, MCP). 1 CNY = 3.6 credits; top up when they run out.

Technical Quick Reference

Item	Value
WebAPI path	/api/{appCode}/{datasetCode}/{method}, cookie authentication, GET/POST supported
OpenAPI path	/openapi/data/{method}, signature authentication, POST only
Token validity	10 minutes
Environments	production (runtime.lovrabet.com), daily (daily-runtime.lovrabet.com), development (dev.lovrabet.com)
Batch limits	update/delete: up to 1000 records per call
Multi-table JOIN limit	Up to 5 levels of relations in filter
Default pagination	currentPage: 1, pageSize: 20
traceparent	All HTTP requests carry the W3C distributed tracing header

---

This document is updated regularly. For questions or suggestions, contact the team.