Minimalistic RAG Starter

Starter project for your GenAI RAG application.

🎬 Demo

The demo is using nonsense made up QuantumFlux software documentation generated by ChatGPT

🛠️ Tech Stack

The project is built using the following open source tools:

• Frontend: Vue 3, HeadlessUI, Prism
• Backend: Redis, Nest.js
• DevOps: Docker
• Others: LangChain, Socket.io
• Database: Qdrant

📦 Prerequisites

In order to run the app you should have the following installed:

• OpenAI API Key
• Node 20
• Docker (or Colima)

⚙️ Installation

1. Clone the repository:

   git clone https://github.com/zaikinv/rag-starter.git
   cd rag-starter

2. Install the dependencies:

   npm i

📊 Ingest the data

1. Create a docs folder

   mkdir docs

2. Add your docs to folder.

   🚨 Currently supports .md files only. Adjust packages/scripts/loader if needed.

3. Run the script

   npm run ingest

🤖 Run Locally

Clone the project

1. Create .env in the root folder:

   touch .env

2. Set environment variables in .env:

   OPENAI_API_KEY=
   OPENAI_API_BASE_URL=
   
   VECTOR_STORE_DEFAULT_COLLECTION_NAME=
   VECTOR_STORE_API_KEY=
   VECTOR_STORE_URL=
   
   REDIS_HOST=
   REDIS_PORT=
   
   DOCS_FOLDER_NAME=

   Example

   OPENAI_API_KEY=sk-JsdnaErYVIPuej4kqweT7BmzkGH7FXIDPR6u4fGyUF8W1ET
   OPENAI_API_BASE_URL=https://api.openai.com/v1/
   
   VECTOR_STORE_DEFAULT_COLLECTION_NAME=my-embeddings
   VECTOR_STORE_API_KEY=JsdnaErYVIPuej4kqweT7Bmzk
   VECTOR_STORE_URL=http://localhost:6333
    
   REDIS_HOST=localhost
   REDIS_PORT=6379
   
   DOCS_FOLDER_NAME=docs

3. Configure the assistant web component:

   <ask-docs-widget
     ui-theme=""
     api-url=""
     vector-store-collection-name=""
     external-source-base-url="">
   </ask-docs-widget>

   Parameter	Example	Description
   ui-theme	light	Can be customized in src/styles/themes.scss
   api-url	http://localhost:3000	Base URL of Nest backend
   vector-store-collection-name	my-embeddings	Name of Qdrant collection
   external-source-base-url	https://google.com	Optional, used to prefix links

4. Run the containers:

   docker compose -f docker-compose.yml up --build --remove-orphans

5. Run the app:

   npm run start:all