Flask Application with pgvector

This project is a Flask web application that uses pgvector for vector storage and retrieval, along with langchain for language model interactions.

Prerequisites

• Docker
• Python 3.8+
• pip
• ollama

Setup Instructions

Step 1: Run pgvector in a Docker Container

1. Pull the Docker image for PostgreSQL with pgvector:

   docker pull ankane/pgvector

2. Run the Docker container:

   docker run --name pgvector -e POSTGRES_USER=mirror -e POSTGRES_PASSWORD=mysecretpassword -e POSTGRES_DB=mirror -d -p 6024:5432 pgvector/pgvector:pg16

   Replace mysecretpassword with a secure password of your choice.

Step 2: Install Python Dependencies

1. Create a virtual environment:

   python3 -m venv venv

2. Activate the virtual environment:

   source venv/bin/activate

3. Install the required packages:

   pip install -r requirements.txt

Step 3: Configure the Database

1. Set the DATABASE_URL environment variable with your database connection string. For example:

   export DATABASE_URL="postgresql://mirror:mysecretpassword@localhost:6024/mirror"

Step 4: Build the UI

1. Navigate to the frontend directory:

   cd frontend

2. Install the frontend dependencies:

   npm install

3. Build the UI:

   npm run build

Step 5: Run the Application

1. Start the Flask application:

   python app.py

2. Access the application at http://localhost:5000.

Step 6: Install Ollama models

1. Install the necessary models

   ollama pull phi3:mini
   ollama pull mistral

Running the API in Development Mode

To run the API in development mode, set the FLASK_DEBUG environment variable to True:

export FLASK_DEBUG=True

Then start the Flask application as usual:

python app.py

Endpoints

• POST /generate: Generate a response using the LLM based on the provided prompt. The response includes the generated text and an optional array of sources.
• GET /system_prompt: Retrieve the current system prompt.
• PUT /system_prompt: Update the system prompt.
• POST /embed: Embed a title and description and save it to the database.
• POST /rag: Retrieve and generate an answer for a given question using RAG. The response includes the generated answer and an array of sources.

Building the Docker Image

To build the Docker image for the application, run the following command in the root directory of the project:

docker build -t n3rdgir1/elysian-mirror:latest .

Publishing the Docker Image to GitHub's Registry

To publish the Docker image to GitHub's registry, follow these steps:

1. Authenticate with GitHub's Docker registry:

   echo $GITHUB_TOKEN | docker login ghcr.io -u n3rdgir1 --password-stdin

2. Tag the Docker image:

   docker tag n3rdgir1/elysian-mirror:latest ghcr.io/n3rdgir1/elysian-mirror:latest

3. Push the Docker image to GitHub's registry:

   docker push ghcr.io/n3rdgir1/elysian-mirror:latest

Running the Docker Container

To run the Docker container, use the following command:

docker-compose up -d

This will start the containers and expose the application on port 5012.