Set up TensorZero Autopilot - Tensorzero

TensorZero Autopilot is an automated AI engineer that analyzes LLM observability data, optimizes prompts and models, sets up evals, and runs A/B tests. It's an optional complementary service that runs on top of your self-hosted TensorZero deployment.

<Note>

TensorZero Autopilot is currently in a private beta. Schedule a demo →

</Note>

Deploy

Visit autopilot.tensorzero.com to generate an API key.

</Step> <Step title="Set the environment variable for the TensorZero Gateway">

Set the environment variable TENSORZERO_AUTOPILOT_API_KEY for your TensorZero Gateway:

bash

export TENSORZERO_AUTOPILOT_API_KEY="sk-t0-..."

</Step> <Step title="Deploy TensorZero">

TensorZero Autopilot requires the TensorZero Gateway, TensorZero UI, and Postgres.

Make sure the gateway has the TENSORZERO_AUTOPILOT_API_KEY environment variable.

</Warning> <Tip>

Learn more about how to:

</Tip> </Step> <Step title="Allow TensorZero Autopilot to edit your local configuration (optional)">

TensorZero Autopilot can edit your local TensorZero configuration.

When this feature is enabled, an "Apply changes" button appears in the UI, allowing you to easily apply any configuration changes proposed by Autopilot.

We recommend enabling this feature only in a local setup. Do not enable this feature if the TensorZero UI is shared by multiple users.

</Warning>

To enable it, mount the configuration directory as a writable volume and pass the --config-file flag to the TensorZero UI container. For example:

yaml

# ...
ui:
  image: tensorzero/ui
  # ...
  command: --config-file /app/config/tensorzero.toml
  volumes:
    - ./config:/app/config
  # ...

</Step> <Step title="Use TensorZero Autopilot">

Visit /autopilot in the self-hosted TensorZero UI to use Autopilot.

</Step> </Steps>