Complete End-to-End Architecture

See how data flows from sources through our unified data layer, AI-powered agent builder, deployment layer, and finally to end users.

Complete E2E Architecture Flow

Loading diagram...

Connect databases, APIs, Google Drive, webhooks, or upload files. No complex setup required.

POS systems (Qashier, Eats365, Raptor, Micros, Toast, Lightspeed, StoreHub, Square)
Databases (PostgreSQL, MySQL, MongoDB, SQL Server, Oracle, DuckDB)
Cloud warehouses (BigQuery, Snowflake, Redshift, Athena, ClickHouse, Trino)
File uploads (CSV, Excel, JSON, PDF, Word, PowerPoint, Markdown, HTML, TXT)
Google Drive files (Docs, Sheets, Slides, PDFs)
REST APIs (Bearer, API key, OAuth2)
GraphQL APIs (schema introspection)
Webhooks for event-driven data

Our AI automatically discovers schema, detects relationships, and builds a semantic model. No manual configuration needed.

Automatic schema discovery (tables, fields, types)
Relationship detection and MDL generation
CRAVEVA AI Semantic Layer (natural language to SQL/NoSQL)
File RAG pipeline (parse → chunk → embed)
Vector storage (PostgreSQL pgvector migration in Beta V2.0, MongoDB Atlas Vector Search currently)
Hybrid mode (online sources + offline files)

Build agents with schema-aware configuration, prompt templates, and output settings.

Prompt Engineering templates + output settings
Craveva LLM Router (multi-provider models)
Semantic models (MDL) for SQL/NoSQL generation
17 output formats (Text, Chart, Infographic, PDF, PowerPoint, Word, Markdown, CSV, Excel, JSON, HTML, YAML, RTF, XML, SQL, ZIP, Diagram, Flowchart, Org Chart, Process)
12 specialized agent types (Sales, Customer Service, Data Analysis, SEO, Design & Social, Internal Coach, Procurement, Search Platform, Search Brand, Auto Acquisition, Delivery Platform, Content Generation)
Context-aware prompt assembly from data layer

Deploy agents to messaging, e-commerce, or custom systems with multiple integration options.

End users interact with agents using natural language and receive answers backed by live or document data.

Loading diagram...

This sequence diagram shows how data flows through the system from initial connection through agent execution and user interaction.

Unified access for databases, APIs, files, and Google Drive

Connectors for databases, REST/GraphQL, webhooks, Google Drive, and files
Auto schema discovery + relationship detection
File RAG pipeline with worker-thread parsing + chunking
Vector store routing (PostgreSQL pgvector migration in Beta V2.0, MongoDB Atlas Vector Search currently)
CRAVEVA AI Semantic Layer (NL → SQL/NoSQL)
Hybrid query engine (online + offline sources)