For charts before 0.15.1, Insights and Chat are not enabled by default. Use the configuration supported by your pinned chart, or upgrade before following this guide.
langsmith_config.yaml. Continue with Configure model access.
Set the encryption keys
To generate one key per feature, see Keys and secrets. Do not commit encryption keys to version control. Provide the keys in one of two ways:- Inline: Set
insights.encryptionKeyandpolly.encryptionKeyinlangsmith_config.yaml, as in the minimum configuration. - Existing Secret: Add
insights_encryption_keyandpolly_encryption_keyto your existing LangSmith app Secret. Reference it throughconfig.existingSecretName.
Configure model access
Insights and Chat call models through your provider credentials and model configurations. For keyless Vertex AI on GKE, see Configure GKE workload identity.Disable Insights or Chat
Disable either feature when your installation does not need it. Set the corresponding top-level flag tofalse in your langsmith_config.yaml. You do not need an encryption key for a disabled feature.
engine.enabled is true.
Upgrade existing configurations
This section applies only to installations upgrading from earlier chart versions. Current charts reject the legacyconfig.insights, config.polly, and backend.agentBootstrap settings. Remove these settings rather than setting them to false. Contact technical support through the Support Portal for legacy deployment migration guidance.
Starting with chart 0.16.0, Insights and Engine share the langsmith-insights-engine image and deployment. This does not enable Engine. Move legacy insights.apiServer, insights.queue, insights.postgres, insights.redis, and insights.namePrefix values to engineInsightsAgent. Keep insights.enabled and insights.encryptionKey under insights.
If you pin the retired langsmith-clio image, remove or update that pin before upgrading. For private registries, mirror langsmith-insights-engine. See Mirroring images.
For the general upgrade procedure, see Upgrade an installation.
See also
Connect these docs to your agent of choice via MCP for real-time answers.

