TensorZero UI Deployment Guide

The TensorZero UI is a self-hosted web application that streamlines the use of TensorZero with features like observability and optimization. It’s easy to get started with the TensorZero UI.

Setup

To use the TensorZero UI, you only need your ClickHouse database URL (TENSORZERO_CLICKHOUSE_URL) and TensorZero Gateway URL (TENSORZERO_GATEWAY_URL). Optionally, you can also provide credentials for fine-tuning APIs.

Credentials for Fine-tuning

The TensorZero UI integrates with model providers like OpenAI to streamline workflows like fine-tuning. To use these features, you need to provide credentials for the relevant model providers as environment variables. You don’t need to provide credentials if you’re not using the fine-tuning features for those providers.

The supported fine-tuning providers and their required credentials (environment variables) are:

Provider	Required Credentials
Fireworks AI	`FIREWORKS_ACCOUNT_ID` `FIREWORKS_API_KEY`
OpenAI	`OPENAI_API_KEY`

We’re planning to add support for more fine-tuning providers in the near future.

Deployment

The TensorZero UI is available on Docker Hub as tensorzero/ui.

Running with Docker Compose (Recommended)

You can easily run the TensorZero UI using Docker Compose:

services:
  ui:
    image: tensorzero/ui
    # Mount your configuration folder (e.g. tensorzero.toml) to /app/config
    volumes:
      - ./config:/app/config:ro
    # Add your environment variables the .env file
    env_file:
      - ${ENV_FILE:-.env}
    # Publish the UI to port 4000
    ports:
      - "4000:4000"
    restart: unless-stopped

For more details, see the example docker-compose.yml file in the GitHub repository.

Running with Docker

Alternatively, you can launch the UI directly with the following command:

docker run \
    --volume ./config:/app/config:ro \
    --env TENSORZERO_CLICKHOUSE_URL="http://chuser:[email protected]:8123/tensorzero" \
    --env OPENAI_API_KEY="sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
    --publish 4000:4000 \
    tensorzero/ui

Make sure to replace the TENSORZERO_CLICKHOUSE_URL with the URL of your ClickHouse instance, and add any other API credentials as needed like we did above with OPENAI_API_KEY.

Building from source

Alternatively, you can build the UI from source. See our GitHub repository for more details.