RAG

Cortex RAG

Your Documents, Answered Instantly

Cortex RAG is an enterprise RAG platform designed for Indian government and regulated enterprises. Organizations upload documents and get AI-powered question-answering with chunk-level source citations, powered by hybrid retrieval (dense + sparse + reranking) and 16 configurable pipeline modes from simple RAG to multi-agent orchestration.

  • Enterprise-grade with complete data isolation
  • Hybrid retrieval: dense + sparse + cross-encoder reranking
  • 16 pipeline modes from simple RAG to multi-agent
  • Embeddable chat widget (19KB, Shadow DOM)
Hybrid Retrieval
16 Pipeline Modes
Source Citations
Document Versioning

Key Features

Everything you need to transform your rag operations

Hybrid Retrieval

BGE-M3 dual-encoder producing dense and sparse vectors in a single pass, fused with Reciprocal Rank Fusion and cross-encoder reranking.

16 Pipeline Modes

From simple RAG to HyDE, Corrective RAG, Self-Reflective RAG, Multi-Hop, Graph RAG, and multi-agent orchestration.

Source Citations

Every answer comes with chunk-level citations linking back to the exact source document and page.

Document Versioning

Full version history with supersedes tracking and automatic file numbering for government document conventions.

Document Preview

In-browser PDF and DOCX preview with page-level navigation. View source documents alongside AI answers without leaving the platform.

Embeddable Widget

19KB zero-dependency chat widget with Shadow DOM isolation. Embed document Q&A into any web application.

Compliance Ready

Enterprise-grade data isolation with tenant-scoped encryption. Built for regulated industries with full compliance controls.

Modules

Powerful Components

Specialized modules designed for specific business needs

Ingestion Engine

Multi-format document processing with 11 chunking strategies

PDF/DOCX/Excel11 Chunking StrategiesLate ChunkingCrash Recovery

Retrieval Core

Hybrid dense + sparse retrieval with cross-encoder reranking

Dense SearchBM25 SparseRRF FusionCross-Encoder Rerank

Pipeline Engine

16 configurable RAG pipeline modes per collection

Simple/StandardHyDE/CRAG/Self-RAGMulti-AgentGraph RAG

Evaluation Suite

RAGAS-powered evaluation with faithfulness and relevancy metrics

FaithfulnessAnswer RelevancyContext PrecisionContext Recall

Chat Widget

Zero-dependency embeddable chat with SSE streaming

Shadow DOMSSE Streaming19KB BundleCustom Styling

Impact by Numbers

Measurable results that matter

16

Pipeline Modes

11

Chunking Strategies

19KB

Widget Size

0

Data Egress

Architecture

How It Works

Enterprise-grade architecture designed for scale and reliability

01

Frontend

Next.js AppChat UIEval DashboardMonaco Editor
02

API Layer

FastAPISSO AuthRBACSSE Streaming
03

RAG Core

IngestionChunkingRetrievalGeneration
04

AI Layer

BGE-M3 EmbedderCross-EncoderLiteLLM GatewayRAGAS Eval
05

Storage

PostgreSQLQdrantRedis CacheFile Storage
Technology

Built With the Best

Modern, battle-tested technologies powering Cortex RAG

Python
Language
FastAPI
Backend
Next.js
Frontend
React
Frontend
TypeScript
Language
Qdrant
Vector Store
PostgreSQL
Database
Redis
Cache
BGE-M3
Embeddings
LiteLLM
LLM Gateway
RAGAS
Evaluation
Docker
Container

Ready to Transform with Cortex RAG?

Schedule a personalized demo and discover how we can help your organization.