Truss Push Action

This action deploys a Truss model or chain to Baseten. It pushes the deployment, waits for it to become active, and optionally validates it with a predict request.

Models are detected when truss-directory points to a directory containing config.yaml. Chains are detected when truss-directory points to a .py file containing a @chains.mark_entrypoint class.

Usage

Model

- uses: basetenlabs/action-truss-push@v0.1
  with:
    truss-directory: "./my-model"
    baseten-api-key: ${{ secrets.BASETEN_API_KEY }}

Chain

- uses: basetenlabs/action-truss-push@v0.1
  with:
    truss-directory: "./my_chain.py"
    baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
    predict-payload: '{"max_value": 5}'

Inputs

- uses: basetenlabs/action-truss-push@v0.1
  with:
    # Path to a model directory containing config.yaml,
    # or a .py file for chain deployments
    # Required
    truss-directory: ""

    # Baseten API key
    # Required
    baseten-api-key: ""

    # Override the model/chain name
    # For models: maps to truss push --model-name
    # For chains: sets the chain_name
    # Default: '' (uses model_name from config.yaml, or entrypoint class name for chains)
    model-name: ""

    # Deploy to a specific environment (implies publish)
    # Default: '' (no environment)
    environment: ""

    # Attach git versioning info (sha, branch, tag) to the deployment
    # Default: true
    include-git-info: ""

    # JSON string of labels as key-value pairs
    # Default: ''
    labels: ""

    # Name of the deployment. Defaults to 'PR-<number>_<sha>' on pull
    # requests or '<sha>' otherwise
    # Default: '' (auto-generated)
    deployment-name: ""

    # Whether to deactivate the deployment after validation
    # Default: true
    cleanup: ""

    # JSON predict payload. For models, defaults to
    # model_metadata.example_model_input from config.yaml.
    # For chains, must be provided explicitly.
    # Default: ''
    predict-payload: ""

    # Max minutes to wait for deployment to become active
    # Default: 45
    deploy-timeout-minutes: ""

    # Timeout in seconds for predict request
    # Default: 300
    predict-timeout: ""

    # Use regional endpoint format for predict requests. Set to true when
    # deploying to a regional environment that requires regional routing.
    # Requires `environment` to also be set.
    # Default: false
    regional-environment: ""

    # Baseten team name to deploy to. Required when the API key has access to
    # multiple teams.
    # Default: '' (uses the default team)
    team: ""

Scenarios

• Deploy a chain
• Deploy a model without cleanup
• Deploy with a custom predict payload
• Deploy to a specific environment
• Deploy to a regional environment
• Deploy with labels
• Run in CI on pull requests
• Deploy multiple models

Deploy a chain

Deploy a Baseten chain from a Python source file. The action auto-detects chains when the path ends in .py.

- uses: basetenlabs/action-truss-push@v0.1
  with:
    truss-directory: "./chains/my_chain.py"
    baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
    model-name: "my-rag-chain"
    cleanup: false
    predict-payload: '{"query": "What is Baseten?"}'

Deploy a model without cleanup

Keep the deployment running after validation for further inspection or manual promotion.

- uses: basetenlabs/action-truss-push@v0.1
  with:
    truss-directory: "./my-model"
    baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
    cleanup: false

Deploy with a custom predict payload

Override the example input defined in config.yaml with an inline JSON payload.

- uses: basetenlabs/action-truss-push@v0.1
  with:
    truss-directory: "./my-model"
    baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
    predict-payload: '{"prompt": "Hello, world!", "max_new_tokens": 128}'
    predict-timeout: 60

Deploy to a specific environment

Push to a named environment (e.g., staging).

- uses: basetenlabs/action-truss-push@v0.1
  with:
    truss-directory: "./my-model"
    baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
    environment: "staging"
    cleanup: false

Deploy to a regional environment

If your environment is a regional environment that requires regional routing, set regional-environment: true so the predict validation request uses the regional endpoint format (model-{id}-{env}.api.baseten.co/predict) instead of the deployment endpoint.

- uses: basetenlabs/action-truss-push@v0.1
  with:
    truss-directory: "./my-model"
    baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
    environment: "prod-us"
    regional-environment: true

Deploy with labels

Attach metadata labels to track deployments in your CI pipeline.

- uses: basetenlabs/action-truss-push@v0.1
  with:
    truss-directory: "./my-model"
    baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
    labels: '{"team": "ml-platform", "triggered-by": "ci"}'

Run in CI on pull requests

Validate model changes on every pull request without promoting to production.

name: Validate model

on:
  pull_request:
    branches: [main]

jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - uses: basetenlabs/action-truss-push@v0.1
        with:
          truss-directory: "./my-model"
          baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
          cleanup: true

Deploy multiple models

Deploy several models in parallel using a matrix strategy.

name: Deploy models

on:
  push:
    branches: [main]

jobs:
  deploy:
    runs-on: ubuntu-latest
    strategy:
      matrix:
        model:
          - path: llm/gpt-oss-20b/latency
          - path: video/stable-diffusion/quality
    steps:
      - uses: actions/checkout@v4

      - uses: basetenlabs/action-truss-push@v0.1
        with:
          truss-directory: ${{ matrix.model.path }}
          baseten-api-key: ${{ secrets.BASETEN_API_KEY }}
          cleanup: false

Recommended secrets configuration

Store your Baseten API key as an encrypted secret in your repository or organization. Never hardcode it in your workflow file.

permissions:
  contents: read

The action does not require any additional GitHub token permissions beyond reading repository contents.

License

This project is released under the MIT License.