RAGtime

RAGtime is a lightweight and intuitive Retrieval-Augmented Generation (RAG) framework built for LLM workflows. It supports PDF and raw text embedding, storage, and retrieval using OpenAI embeddings and Qdrant vector search.

Text is automatically chunked and embedded with relevant summaries, and queries return contextually similar results via cosine similarity.

Features

• 📄 Embed PDFs and raw text content
• ⚙️ Chunking with optional summarisation
• 🧠 Vector embeddings with OpenAI
• 📦 Embedding storage in Qdrant
• 🔍 Querying across single or multiple collections
• 🧪 Tests double as usage examples

Requirements

• Bun runtime
• Docker + Docker Compose (for Qdrant)
• OpenAI API Key

Getting Started

1. Install Dependencies

bun install

2. Setup Qdrant and Run Tests

bun run setup

This will:

• Build and run Qdrant via Docker Compose
• Run the full test suite to verify embedding/query workflows

Usage

Embed a PDF

import OpenAI from 'openai'
import { EmbeddingProcessingService } from '@@services/EmbeddingProcessing.service'

const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY })
const embeddingProcessor = new EmbeddingProcessingService(openai)

const { embeddingId } = await embeddingProcessor.embedPDF('path/to/file.pdf')

Embed Raw Text

const input = [
  'Some historical account...',
  'Details about your hardware...',
]

const { embeddingId } = await embeddingProcessor.embedText(input)

Query a Collection

import { EmbeddingQueryService } from '@@services/EmbeddingQuery.service'
import { EmbeddingManagementService } from '@@services/EmbeddingManagement.service'

const queryService = new EmbeddingQueryService(
  new EmbeddingManagementService(),
  embeddingProcessor
)

const results = await queryService.query('What year did I get the Amiga 500?', embeddingId!)
console.log(results)

Query Multiple Collections

const results = await queryService.queryCollections(
  'What GPUs did I use and how is key exchange done?',
  [embeddingId1, embeddingId2],
  2
)

console.log(results)

Environment

Create a .env file in your project root with:

OPENAI_API_KEY=your-key-here

Tests

Run all tests with:

bun test

Test files in spec/ demonstrate full examples of:

• PDF embedding and querying
• Text chunking and embedding
• Multi-document collection search
• Expected RAG-based responses

License

MIT © Eugene Odeluga