Easy RAG logo

Guides

Introduction

What is EasyRAG and how it works

Welcome to EasyRAG! This guide will help you understand what EasyRAG is and how it can help you build AI-powered applications.

What is EasyRAG?

EasyRAG is a Retrieval-Augmented Generation (RAG) service that makes it easy to build AI applications that can understand and answer questions about your documents.

Think of it as giving ChatGPT knowledge of your specific documents without having to build the complex infrastructure yourself.

What problems does EasyRAG solve?

Without EasyRAG

Building a document Q&A system typically requires:

  1. ✅ Setting up a vector database (Qdrant, Pinecone, etc.)
  2. ✅ Implementing document parsing for PDFs, Word docs, etc.
  3. ✅ Creating chunking strategies for large documents
  4. ✅ Managing embeddings and vector storage
  5. ✅ Building search and retrieval logic
  6. ✅ Integrating with LLMs for answer generation
  7. ✅ Handling transcription for audio/video files
  8. ✅ Managing multi-tenant data isolation

EasyRAG handles all of this for you.

With EasyRAG

javascript
// 1. Upload your documents
await uploadFile(dataset, pdfFile);

// 2. Ask questions
const answer = await query(dataset, "What are the key points?");

// Done! 🎉

How does it work?

EasyRAG follows a simple 3-step process:

1. Upload & Index

Your Document → Chunking → Embeddings → Vector Storage

When you upload a document:

  • EasyRAG breaks it into smaller chunks (default: 300 tokens)
  • Each chunk is converted to a vector embedding
  • Embeddings are stored in Qdrant for fast searching

2. Search

Your Question → Embedding → Vector Search → Relevant Chunks

When you search:

  • Your question is converted to an embedding
  • EasyRAG finds the most relevant chunks using similarity search
  • Returns ranked results with scores

3. Generate Answers (Optional)

Question + Context → GPT-4 → Grounded Answer

When you query:

  • EasyRAG retrieves relevant chunks (step 2)
  • Sends them to GPT-4 with your question
  • Returns an AI-generated answer based on your documents

Key Features

📄 Document Support

Upload virtually any document type:

  • PDFs, Word documents (.docx)
  • Excel spreadsheets (.xlsx)
  • PowerPoint presentations (.pptx)
  • CSV files
  • Audio files (.mp3, .wav) - automatically transcribed
  • Video files (.mp4) - audio extracted and transcribed

🔍 Semantic Search

Find relevant information even if the exact words don't match:

Question: "How do I reset my password?"
Matches: "To recover your account credentials..."

💬 AI-Powered Answers

Get ChatGPT-like answers grounded in your documents:

javascript
const answer = await query(
  'support-docs',
  'What is your refund policy?'
);
// Returns: "According to our policy documents, refunds are 
// available within 30 days of purchase for any reason..."

🏢 Multi-Tenant Architecture

Built-in support for multiple users/projects:

  • Create separate datasets for each user
  • Or use metadata filters in shared datasets
  • Complete data isolation

🔒 Secure Browser Access

Frontend tokens allow safe API access from browsers:

  • Short-lived (1-24 hours)
  • Dataset-scoped
  • No API key exposure

Common Use Cases

Customer Support Chatbot

javascript
// Upload help docs once
await uploadFiles('support-kb', [
  'faq.pdf',
  'user-manual.pdf',
  'troubleshooting.pdf'
]);

// Users ask questions
const answer = await query(
  'support-kb',
  'How do I reset my password?'
);

Document Management System

javascript
// Each user has their own dataset
const userDataset = `user-${userId}`;

// User uploads their documents
await uploadFile(userDataset, document);

// User searches their documents
const results = await search(userDataset, query);

Knowledge Base

javascript
// Upload company policies, procedures, contracts
await uploadFiles('company-kb', documents);

// Employees can search
const results = await search(
  'company-kb',
  'What is the vacation policy?',
  { filters: [{ key: 'department', match: { value: 'HR' } }] }
);

Educational Platform

javascript
// Upload course materials
await uploadFiles('course-101', lectures);

// Students ask questions
const answer = await query(
  'course-101',
  'Explain the concept of recursion'
);

What's Next?

Ready to get started?

  1. Quick Start - Get up and running in 5 minutes
  2. Authentication - Learn about API keys and tokens
  3. Uploading Files - Upload your first document
  4. API Reference - Detailed API documentation

Pricing

EasyRAG uses a simple pay-as-you-go model:

  • File Upload: 1 credit per file
  • Search/Query: 0.1 credit per request
  • Transcription: 0.1 credit per minute

Free tier: 20 credits on signup (~20 uploads or 200 queries)

See the dashboard for credit packs and pricing.

Support

Need help?