RANA vs LlamaIndex

LlamaIndex pioneered RAG frameworks. RANA builds on those patterns with a simpler API, TypeScript-first design, and production features.

70%

Less Code

100%

TypeScript

Built-in

Cost Tracking

~50KB

Bundle Size

Feature	LlamaIndex	RANA
RAG Pipeline Both provide RAG capabilities	✓	✓
Vector Storage Both support vector stores	✓	✓
Document Loaders Both support multiple document types	✓	✓
TypeScript Native RANA is TypeScript-first	✗	✓
Cost Tracking RANA tracks costs automatically	✗	✓
Built-in Testing RANA has @rana/testing	✗	✓
Security Features RANA includes PII detection, injection prevention	✗	✓
Multi-Provider Both support multiple providers	✓	✓
Observability RANA has built-in tracing	✗	✓
Agent Framework Both support agent patterns	✓	✓

Basic RAG Setup

LlamaIndex26 lines

import { Document, VectorStoreIndex, serviceContextFromDefaults } from "llamaindex";

// Create documents
const documents = [
  new Document({ text: "Your document content here..." }),
  new Document({ text: "Another document..." }),
];

// Create service context
const serviceContext = serviceContextFromDefaults({
  chunkSize: 512,
  chunkOverlap: 50,
});

// Create index
const index = await VectorStoreIndex.fromDocuments(
  documents,
  { serviceContext }
);

// Create query engine
const queryEngine = index.asQueryEngine();

// Query
const response = await queryEngine.query("Your question here");
console.log(response.toString());

RANA13 lines

import { createRana } from '@rana/core';
import { createRAG } from '@rana/rag';

const rana = createRana();
const rag = createRAG({ rana });

await rag.ingest([
  "Your document content here...",
  "Another document..."
]);

const response = await rag.query("Your question here");
console.log(response.content);

Custom Embeddings

LlamaIndex22 lines

import {
  Document,
  VectorStoreIndex,
  OpenAIEmbedding,
  serviceContextFromDefaults
} from "llamaindex";

const embedModel = new OpenAIEmbedding({
  model: "text-embedding-3-small",
  dimensions: 1536,
});

const serviceContext = serviceContextFromDefaults({
  embedModel,
});

const documents = [new Document({ text: content })];

const index = await VectorStoreIndex.fromDocuments(
  documents,
  { serviceContext }
);

RANA13 lines

import { createRana } from '@rana/core';
import { createRAG } from '@rana/rag';

const rana = createRana();
const rag = createRAG({
  rana,
  embedding: {
    model: 'text-embedding-3-small',
    dimensions: 1536,
  },
});

await rag.ingest(content);

Streaming Responses

LlamaIndex25 lines

import {
  Document,
  VectorStoreIndex,
  OpenAI,
  serviceContextFromDefaults
} from "llamaindex";

const llm = new OpenAI({
  model: "gpt-4",
  temperature: 0.7,
});

const serviceContext = serviceContextFromDefaults({ llm });

const index = await VectorStoreIndex.fromDocuments(docs, { serviceContext });
const queryEngine = index.asQueryEngine();

const stream = await queryEngine.query(
  "Your question",
  { streaming: true }
);

for await (const chunk of stream) {
  process.stdout.write(chunk.response);
}

RANA11 lines

import { createRana } from '@rana/core';
import { createRAG } from '@rana/rag';

const rana = createRana({ model: 'gpt-4' });
const rag = createRAG({ rana });

await rag.ingest(docs);

for await (const chunk of rag.queryStream("Your question")) {
  process.stdout.write(chunk);
}

When to Choose Each

Choose RANA if you:

✓Want simpler, more readable RAG code
✓Need TypeScript-first with great type inference
✓Want built-in cost tracking and security
✓Prefer convention over configuration

Choose LlamaIndex if you:

•Need Python as your primary language
•Have existing LlamaIndex integrations
•Need highly specialized index types

Try RANA Now→