iLlama

A web interface built with Nuxt for interacting with any OLlama language model locally.

Prerequisites

• Docker
• Docker Compose
• Node.js (for local development)
• pnpm
• oLlama (for local development)

Technologies & Dependencies

• Nuxt.js 3
• Nuxt UI 3
• Tailwind CSS v4
• TailwindCSS Motion
• Marked

Environment Setup

1. Create a .env file in the root directory:

cp .env.example .env

2. Configure your model in the .env file:

# Model Configuration
NUXT_PUBLIC_LLAMA_MODEL="deepseek-r1:1.5b"
LLAMA_MODEL="deepseek-r1:1.5b"

Note: Both environment variables should typically use the same model name.

Docker Setup

1. Clone the repository:

git clone <repository-url>
cd iLlama

2. Start the application using Docker Compose:

docker compose up -d

3. Stop the application using Docker Compose:

docker compose down

The application will be available at http://localhost:3000

Local Development Setup

1. Install dependencies and start dev server:

pnpm install
pnpm dev

2. Install oLlama

3. Pull and run your desired model:

ollama run deepseek-r1:1.5b  # Replace 'deepseek-r1:1.5b' with your actual model name

The development server will be available at http://localhost:3000

Available Models

You can use any model supported by oLlama. Some popular options include:

• deepseek
• codellama
• mistral
• llama2

The following models were recently tested in a containerized environment and ran smoothly:

• deepseek-r1:1.5b
• llama3.2:1b
• qwen2.5:0.5b

The model name in your .env file must match exactly with the model name from oLlama's library. For example: deepseek-r1:1.5b not just deepseek.

Check oLlama's model library for more options.

View Installed Models

To view installed model(s):

echo 'Installed models: ' && docker run --rm -v illama_ollama_data:/data alpine ls /data/models/manifests/registry.ollama.ai/library

Environment Variables

• NUXT_PUBLIC_LLAMA_MODEL: Model name used by the frontend
• LLAMA_MODEL: Model name for Docker container to pull

Both variables should typically match and use the exact model name from oLlama's library.