Insight - Jacked RAG

RAG - Retrieval Augmented Generation. This is the Jacked initiative for that within the PHP ecosystem.

This solution streams back the text chunks of the response coming from the LLM. There is a WebSocket experiment that accomplishes the same in the branch websocket-experiment.

Usage

Sample credentials:

Username: [email protected]

Password: password

Installation

Before start you need to make sure you have the following dependencies:

• php8.2 or higher
• php openswoole extension
• docker (for postgres db)

Step 1

Clone the repository:

git clone https://github.com/Jacked-PHP/insight my-project

Then cd my-project

Step 2

Install dependencies (php composer and npm):

composer install

npm install

Step 3

npm run build

(or npm run dev if you are developing and want the hot reload)

Step 4

Prepare the database:

docker-compose up -d

Step 5

Make sure your .env is properly set:

• have db credentials
• app key (php artisan key:generate)
• app url (default is http://localhost:8000)

Then run the migrations:

php artisan migrate

Run the server (this is running jacked server):

php artisan serve

Now you can visit http://localhost:8000 and see the application running.

Vector DB

The free opensource solution considers the following:

• https://github.com/pgvector/pgvector

Embedding Solution

This application uses the following embedding solution:

• ollama embedding model: mxbai-embed-large

Ollama details

To run Ollama, a suggested way would be to run through docker:

docker run -d -v ollama:$HOME/.ollama -p 11434:11434 --name ollama ollama/ollama

Reference: https://github.com/ollama/ollama/blob/main/docs/docker.md