ollama.md

To create an Ollama Docker setup that uses the nomic-embed-text model, you need to ensure that the Docker container can download and use the model, as well as any necessary configuration files. Here’s a step-by-step guide to achieve this:

Create a Modelfile

First, create a Modelfile that specifies the base model and any additional parameters. Here is an example Modelfile for the nomic-embed-text model, although note that this model is specifically for generating embeddings and does not require the same parameters as a language model:

FROM nomic-embed-text

Since nomic-embed-text is an embedding model, it does not need additional parameters like temperature or num_ctx.

Create a Shell Script

Create a shell script (run_ollama.sh) that will download the model, create the necessary configuration, and start the Ollama service:

#!/bin/bash

# Pull the nomic-embed-text model
ollama pull nomic-embed-text

# Start Ollama service
ollama serve &

# Keep the service running
tail -f /dev/null

Dockerfile

Create a Dockerfile to copy the script and make it executable:

FROM ollama/ollama:latest

COPY run_ollama.sh /run_ollama.sh
RUN chmod +x /run_ollama.sh

Docker Compose File

Create a docker-compose.yml file to define the service and volumes:

version: '3'

services:
  ollama:
    build: .
    container_name: ollama
    ports:
      - "11434:11434"
    volumes:
      - ./model_files:/model_files
      - ollama_volume:/root/.ollama
    command: /run_ollama.sh

volumes:
  ollama_volume:

Directory Structure

Ensure your directory structure is as follows:

- model_files/
  - Modelfile
- run_ollama.sh
- docker-compose.yml
- Dockerfile

Running the Setup

To start the Ollama service with the nomic-embed-text model, run:

docker-compose up -d

This setup will pull the nomic-embed-text model, start the Ollama service, and keep it running in the background.

Using the Model

Once the service is running, you can use the nomic-embed-text model via the Ollama API. Here is an example of how to generate embeddings using curl:

curl http://localhost:11434/api/embeddings -d '{
  "model": "nomic-embed-text",
  "prompt": "The sky is blue because of Rayleigh scattering"
}'