All integrationsAI Models

Baseten

Run and look after the AI models you deploy on Baseten: call a model, wake it, activate, deactivate or promote deployments, change autoscaling, read logs and manage environments and secrets.

actions
20
triggers
4

Triggers

Start a workflow when something happens in Baseten.

  • New modelRuns when a model is added to your Baseten workspace. Models already there when the workflow is turned on don't start it.
  • New deploymentRuns when a new deployment (version) of a model is pushed to Baseten, while it is still building. Pair it with 'Deployment status changed' to act once it is ready.
  • Deployment status changedRuns when a deployment's status changes, such as when it becomes active, fails to deploy or scales to zero. Choose a status to run only when a deployment reaches it.
  • Deployment promoted to productionRuns when a model's production deployment changes, such as when a new version is promoted or production is rolled back to an earlier one.

Actions

Steps a workflow can run in Baseten.

  • List modelsList the models in your Baseten workspace, with their production and development deployments.
  • Get modelGet one model: its name, how many deployments it has and which ones are in production and development.
  • List deploymentsList a model's deployments with their status, running replicas and autoscaling settings.
  • Get deploymentGet one deployment: its status (such as active or scaled to zero), running replicas, hardware and autoscaling settings.
  • Activate deploymentTurn an inactive deployment back on so it serves requests. Baseten bills for the hardware while it runs. Nothing changes if it is already active.
  • Deactivate deploymentTurn a deployment off: it stops serving requests and costing money until it is activated again.
  • Promote deploymentMake a deployment the one an environment serves, such as moving a tested deployment into production.
  • Update autoscalingChange how many replicas a deployment runs and when it adds or removes them. Only the settings you fill in change; the rest keep their values.
  • Retry failed deploymentBuild and deploy a failed deployment again. If it can't be retried, the result says why.
  • Delete deploymentDelete a deployment for good, such as an old version nobody uses. It can't be undone.
  • Get deployment logsGet a deployment's log lines, such as why it failed to start. Without a time window, the last 30 minutes.
  • List environmentsList a model's environments, such as production and staging, with the deployment each one serves.
  • Get environmentGet one environment: the deployment it serves, any promotion under way, and its autoscaling settings.
  • Create environmentAdd an environment to a model, such as staging, with an address of its own that deployments can be promoted to.
  • Delete environmentDelete an environment other than production. The deployment it served scales down to zero but is kept.
  • List secretsList the names of the secrets your models can read, such as access tokens. Their values are never shown.
  • Save secretAdd a secret your models can read, or replace the value of the one with that name. Running deployments get the new value at once.
  • Delete secretDelete a secret. Models that read it no longer get it.
  • Call modelSend input to one of your deployed models and get its answer, such as a summary or a transcript. Baseten bills for the time it runs.
  • Wake modelStart a model that has scaled to zero, ahead of the requests you expect, so nobody waits for it to load. Returns at once.

Automate Baseten with Lodol.

Connect Baseten, then hand Lodol the job in Slack, Claude or ChatGPT.

Free plan, no credit card