Upgrading Self-Hosted LangSmith to v0.15: Fleet rename and migration guide

Applies to: Self-hosted LangSmith (Helm / Kubernetes) upgrading to v0.15.x from v0.14 or earlier

Audience: Platform / infra admins who manage the LangSmith Helm deployment

Summary

LangSmith Self-Hosted v0.15 renames Agent Builder to Fleet. If you have Fleet (formerly Agent Builder) enabled, this rename changes several Helm configuration keys that you must update in your values file before running helm upgrade. There are also a few other breaking changes to review, and a separate migration path for anyone who deployed Fleet through the now-deprecated agent-bootstrap script.

No action is required for the hosted product or for your existing Fleet agents, configurations, and integrations — they continue to work. This guide is about the self-hosted Helm configuration keys and infrastructure.

Before you upgrade

• Downgrades are not supported. v0.15 includes database migrations that are not backward-compatible. If you must roll back, contact support first.

• Review the full v0.15.0 changelog.

• Check the minimum dependency versions for v0.15.

Step 1 — Rename the Fleet config keys

In your Helm values file, rename the following keys (old → new). These are the keys called out by the Agent Builder → Fleet rename:

More broadly, any agentBuilder* section/value in your configuration is renamed to the equivalent fleet* key. Search your values file for agentBuilder and migrate each occurrence.

Step 2 — Update workload identity / service accounts (if applicable)

If you use workload identity, the Fleet rename may change the service account names that the Fleet tool/trigger servers run under. Update any IAM bindings or serviceAccount references that pointed at the old Agent Builder workloads so the new fleet* workloads can authenticate.

Step 3 — Review other v0.15 breaking changes

These are independent of the Fleet rename but ship in the same release:

• agent-bootstrap script deprecated. LangSmith agents are now standalone services that deploy with the Helm chart instead of through the LangSmith Deployment control plane. If you previously deployed Fleet/Agent Builder via the agent-bootstrap script, you need a guided migration — contact support before upgrading.

• RBAC – retention permission split. projects:update-retention is replaced by projects:increase-trace-tier and projects:decrease-trace-tier. Existing roles were backfilled, so no change is needed for them; use the new permissions for any new roles.

• RBAC – new fleet-admin:read permission gates the new Fleet Admin section. Backfilled to existing roles; grant it to new roles that need Fleet Admin access.

• POST /workspaces/current/members now requires role_id for RBAC-enabled orgs. Requests without it return 400 instead of silently defaulting to WORKSPACE_ADMIN.

• USAGE_EXPORT_ADMIN_EMAILS env var deprecated — use INSTANCE_ADMIN_EMAILS instead.

Step 4 — Run the upgrade

# Add/refresh the chart repo
helm repo add langchain https://langchain-ai.github.io/helm/
helm repo update

# Find available versions
helm search repo langchain/langsmith --versions

# Upgrade (use -n <namespace> if not using the default namespace)
helm upgrade <release-name> langchain/langsmith \
  --version <0.15.x version> \
  --values <path-to-values-file> \
  --wait --debug

Step 5 — Verify

helm status <release-name>
kubectl get pods

• All pods should be Running.

• The langsmith-pg-migrations and langsmith-ch-migrations jobs should show Completed.

• Confirm the version: curl <frontend-external-ip>/api/info and check the version field matches the version you upgraded to.

References

• Upgrade an installation

• Self-hosted changelog — v0.15.0

• Enable Fleet, Insights, and Polly (self-hosted)

• Minimum versions for self-hosting dependencies