Skip to main content
Version: Next

Getting Started

Constellaxion is a CLI tool for deploying and fine-tuning open-source models on your private cloud. Whether you’re working with foundation models or fine-tuning with your own dataset, Constellaxion helps you go from config to cloud jobs in just a few commands.

📦 Installation​

pip install constellaxion

🔧 Prerequisites​

Constellaxion is designed to be cloud-agnostic, but each environment has its own setup.

📁 Initializing a Project​

To begin, create a directory with a model.yaml configuration file.

For model deployment:​

  • Only the model.yaml file is needed

For fine-tuning:​

  • Include train.csv, val.csv, and test.csv
  • These CSVs should contain two columns: prompt, response

Then run:

constellaxion init

This generates a job.json file based on your model.yaml.

🌟 Deploying a Foundation Model​

To deploy a foundation model, run:

constellaxion model deploy

Example model.yaml for foundation model deployment:​

model:
  id: cxn-foundation-model
  base: "tiiuae/falcon-7b-instruct"
deploy:
  gcp:
    project_id: your-project-id
    region: europe-west2

What happens under the hood →

🎯 Fine-Tuning a Model​

Train a model on your dataset with:

constellaxion model train

This kicks off a custom training job on Vertex AI using your local data and model base.

Then serve it with:

constellaxion model serve

Example model.yaml for fine-tuning:​

model:
  id: crypto-sentiment-v2
  base: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
dataset:
  train: ./train.csv
  val: ./val.csv
  test: ./test.csv
training:
  epochs: 1
  batch_size: 16
deploy:
  gcp:
    project_id: your-project-id
    region: us-central1

💬 Prompting a Model​

Chat directly with a deployed model using:

constellaxion model prompt

Your terminal becomes a chat window for the deployed model (defined in job.json). Type exit or quit to stop.

🧠 Supported Models

We are actively expanding support for more models and cloud environments.

View supported models and platforms →