Craveva AI
Solutions/Data Layer

Data Layer: Connect Everything

Universal data connections to POS systems, databases, cloud warehouses, APIs, Google Drive, and files. Automatic schema discovery and unified data warehouse.

Data Layer Overview

The Data Layer is the foundation of Craveva AI. It connects all your enterprise data sources into one unified system, enabling AI agents to understand and query your data seamlessly.

Supported Data Sources

Rendering via Kroki...

Usually takes 1-2 seconds

POS Systems

  • Qashier
  • Eats365
  • Raptor
  • Micros
  • Toast
  • Lightspeed
  • StoreHub
  • Square

Databases (12 Total)

  • PostgreSQL
  • MySQL
  • MongoDB
  • SQL Server
  • Oracle
  • DuckDB
  • BigQuery
  • Snowflake
  • Redshift
  • Athena
  • ClickHouse
  • Trino

Cloud Warehouses

  • BigQuery
  • Snowflake
  • Redshift
  • Athena
  • ClickHouse
  • Trino

APIs

  • REST APIs
  • GraphQL
  • Webhooks

Google Workspace

Beta V2.0 (Feb 14, 2026)
  • Google Drive
  • Google Docs
  • Google Sheets

Files

  • CSV
  • Excel
  • JSON
  • PDF
  • Word
  • Parquet

Deployment Modes

Rendering via Kroki...

Usually takes 1-2 seconds

Offline-Only

Process local files without internet connection

  • Upload files directly
  • Process data locally
  • Generate reports offline
  • No external API calls

Best for: Sensitive data, air-gapped environments

Online-Only

Real-time data from connected sources

  • Real-time synchronization
  • Live query execution
  • Automatic updates
  • Multi-source aggregation

Best for: Real-time operations, live dashboards

Hybrid

Combine historical files with live data

  • Merge offline and online data
  • Unified query interface
  • Data enrichment
  • Comprehensive analysis

Best for: Trend analysis, historical comparisons

Automatic Schema Discovery

Our AI automatically discovers your data structure without manual configuration:

Table/Collection Analysis

Automatically detects all tables, collections, and their structures

Column/Field Type Detection

Identifies data types, constraints, and relationships automatically

Constraint Detection

Detects primary keys, foreign keys, and indexes

Data Sampling

Analyzes sample data to understand patterns and business logic

RAG Implementation (For Files)

Beta V2.0 (Feb 14, 2026)

Beta V2.0: PostgreSQL pgvector migration coming Feb 14, 2026 for 70-80% faster vector searches.

Offline file uploads (PDF, Word, Excel, CSV, JSON) are processed using RAG (Retrieval-Augmented Generation) with MongoDB Atlas Vector Search. Beta V2.0 (Feb 14, 2026) will migrate to PostgreSQL pgvector for 70-80% faster vector searches.

1

Parse

Extract text from files

2

Chunk

Split into semantic chunks

3

Embed

Generate vector embeddings

4

Store

Store in vector database

Enables semantic search over uploaded documents for chat-based queries, separate from SQL-based semantic layer used for databases.

Semantic Layer (For Databases)

Our custom Craveva AI Semantic Layer converts natural language to SQL:

  • Self-hosted semantic layer (no external dependencies)
  • MDL (Modeling Definition Language) stored in MongoDB
  • LLM-powered SQL generation using Craveva LLM Router
  • Automatic schema analysis to create MDL definitions
  • Multi-tenant support with full tenant isolation

Copy-Paste JavaScript Deployment

Deploy the AI Data Warehouse anywhere with a simple copy-paste JavaScript snippet:

<script>
  (function() {
    const cravevaScript = document.createElement('script');
    cravevaScript.src = 'https://cdn.craveva.ai/v1/data-warehouse.js';
    cravevaScript.setAttribute('data-api-key', 'YOUR_API_KEY');
    cravevaScript.setAttribute('data-mode', 'online+offline');
    cravevaScript.setAttribute('data-company-id', 'YOUR_COMPANY_ID');
    document.head.appendChild(cravevaScript);
  })();
</script>
← Back to SolutionsAI Layer →