Enable blob storage

By default, LangSmith stores run inputs, outputs, errors, manifests, extras, and events in ClickHouse. If you so choose, you can instead store this information in blob storage, which has a couple of notable benefits. For the best results in production deployments, we strongly recommend using blob storage, which offers the following benefits:

1. In high trace environments, inputs, outputs, errors, manifests, extras, and events may balloon the size of your databases.
2. If using LangSmith Managed ClickHouse, you may want sensitive information in blob storage that resides in your environment. To alleviate this, LangSmith supports storing run inputs, outputs, errors, manifests, extras, events, and attachments in an external blob storage system.

For cloud-specific setup, choose your platform:

• Amazon S3 (AWS)
• Google Cloud Storage (GCP)
• Azure Blob Storage

For complete cloud-specific setup and architecture guides, see AWS, GCP, or Azure.

Requirements

Azure blob storage is available in Helm chart versions 0.8.9 and greater. Deleting trace projects is supported in Azure starting in Helm chart version 0.10.43.

Native GCS blob storage engine support (using engine: "GCS") is available in Helm chart versions 0.13.29 and greater. For earlier versions, GCS is supported via the S3-compatible API by setting engine: "S3" with HMAC credentials.

• Access to a valid blob storage service

  • Amazon S3
  • Google Cloud Storage (GCS)
  • Azure Blob Storage

• A bucket/directory in your blob storage to store the data. We highly recommend creating a separate bucket/directory for LangSmith data.

  • If you are using TTLs, you will need to set up a lifecycle policy to delete old data. For more information, see configuring TTLs. These policies should mirror the TTLs you have set in your LangSmith configuration, or you may experience data loss. See TTL configuration for blob storage for how to set up the lifecycle rules.

• Credentials to permit LangSmith Services to access the bucket/directory

  • You will need to provide your LangSmith instance with the necessary credentials to access the bucket/directory. Read the authentication section below for more information.

• If using S3 or GCS, an API URL for your blob storage service

  • This will be the URL that LangSmith uses to access your blob storage system
  • For Amazon S3, this will be the URL of the S3 endpoint. Something like: https://s3.amazonaws.com or https://s3.us-west-1.amazonaws.com if using a regional endpoint.
  • For Google Cloud Storage, this will be the URL of the GCS endpoint. Something like: https://storage.googleapis.com

Authentication

AWS

Amazon S3

To authenticate to Amazon S3, you will need to create an IAM policy granting the following permissions on your bucket.

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "s3:GetObject",
        "s3:PutObject",
        "s3:DeleteObject",
        "s3:ListBucket"
      ],
      "Resource": [
        "arn:aws:s3:::your-bucket-name",
        "arn:aws:s3:::your-bucket-name/*"
      ]
    }
  ]
}

Once you have the correct policy, there are three ways to authenticate with Amazon S3:

1. IAM Roles for Service Accounts (IRSA) (Recommended): You can create an IAM role for your LangSmith instance and attach the policy to that role. This is the recommended way to authenticate with Amazon S3 in production.

   1. You will need to create an IAM role with the policy attached.
   2. You will need to allow LangSmith service accounts to assume the role. The langsmith-queue, langsmith-backend, langsmith-platform-backend, and langsmith-ingest-queue service accounts will need to be able to assume the role.
      The service account names will be different if you are using a custom release name. You can find the service account names by running kubectl get serviceaccounts in your cluster.

   3. You will need to provide the role ARN to LangSmith. You can do this by adding the eks.amazonaws.com/role-arn: "<role_arn>" annotation to the queue, backend, platform-backend, and ingest-queue services in your Helm Chart installation.

2. Access Key and Secret Key: You can provide LangSmith with an access key and secret key. This is the simplest way to authenticate with Amazon S3. However, it is not recommended for production use as it is less secure.

   1. You will need to create a user with the policy attached. Then you can provision an access key and secret key for that user.

3. VPC Endpoint Access: You can enable access to your S3 bucket via a VPC endpoint, which allows traffic to flow securely from your VPC to your S3 bucket.

   1. You’ll need to provision a VPC endpoint and configure it to allow access to your S3 bucket.
   2. You can refer to our public Terraform modules for guidance and an example of configuring this.

KMS encryption header support

Starting with LangSmith Helm chart version 0.11.24, you can pass a KMS encryption key header and enforce a specific KMS key for writes by providing its ARN. To enable this, set the following values in your Helm chart:

config:
  blobStorage:
    kmsEncryptionEnabled: true
    kmsKeyArn: <your_kms_key_arn>

GCP

Google Cloud Storage

To authenticate with Google Cloud Storage, you will need to create a service account with the necessary permissions to access your bucket.

Your service account will need the Storage Admin role or a custom role with equivalent permissions. This can be scoped to the bucket that LangSmith will be using.

Once you have a provisioned service account, you will need to generate a HMAC key for that service account. This key and secret will be used to authenticate with Google Cloud Storage.

As of Helm chart version 0.13.29, you can set the blob storage engine to "GCS" directly. This supports two authentication methods:

1. GCP Workload Identity (recommended): Leave accessKey and accessKeySecret empty. LangSmith will use Application Default Credentials. You will need to add the workload identity annotation to the backend, platform-backend, queue, and ingest-queue service accounts.
2. HMAC keys: Set accessKey and accessKeySecret to your GCS HMAC credentials.

For both methods, set apiURL to https://storage.googleapis.com and bucketName to your GCS bucket name.

For Helm chart versions prior to 0.13.29, GCS is supported via the S3-compatible API by setting engine: "S3" with HMAC credentials.

Azure

Azure Blob Storage

To authenticate with Azure Blob Storage, you will need to use one of the following methods to grant LangSmith workloads permission to access your container (listed in order of precedence):

1. Storage account and access key
2. Connection string
3. Workload identity (recommended), managed identity, or environment variables supported by DefaultAzureCredential. This is the default authentication method when configuration for either option above is not present.
   1. To use workload identity, add the label azure.workload.identity/use: true to the queue, backend, platform-backend, and ingest-queue deployments. Additionally, add the azure.workload.identity/client-id annotation to the corresponding service accounts, which should be an existing Azure AD Application’s client ID or user-assigned managed identity’s client ID. See Azure’s documentation for additional details.

Some deployments may need further customization of the connection configuration using a Service URL Override instead of the default service URL (https://<storage_account_name>.blob.core.windows.net/). For example, this override is necessary in order to use a different blob storage domain (e.g. government or china).

CH search

By default, LangSmith will still store tokens for search in ClickHouse. If you are using LangSmith Managed Clickhouse, you may want to disable this feature to avoid sending potentially sensitive information to ClickHouse. You can do this in your blob storage configuration.

Configuration

After creating your bucket and obtaining the necessary credentials, you can configure LangSmith to use your blob storage system.

Helm

config:
  blobStorage:
    enabled: true
    engine: "S3" # Or "GCS" or "Azure". This is case-sensitive.
    chSearchEnabled: true # Set to false if you want to disable CH search (Recommended for LangSmith Managed Clickhouse)
    bucketName: "your-bucket-name"
    apiURL: "Your connection url"
    accessKey: "Your access key" # Optional. Only required if using S3 access key and secret key
    accessKeySecret: "Your access key secret" # Optional. Only required if using access key and secret key
    # The following blob storage configuration values are for Azure and require blobStorage.engine = "Azure". Omit otherwise.
    azureStorageAccountName: "Your storage account name" # Optional. Only required if using storage account and access key.
    azureStorageAccountKey: "Your storage account access key" # Optional. Only required if using storage account and access key.
    azureStorageContainerName: "your-container-name" # Required
    azureStorageConnectionString: "" # Optional.
    azureStorageServiceUrlOverride: "" # Optional
  backend: # Optional, only required if using IAM role for service account on AWS, workload identity on GKE, or workload identity on AKS
    deployment: # Azure only
      labels:
        azure.workload.identity/use: true
    serviceAccount:
      annotations:
        azure.workload.identity/client-id: "<client_id>" # Azure only
        eks.amazonaws.com/role-arn: "<role_arn>" # AWS only
        iam.gke.io/gcp-service-account: "<gsa_name>@<project_id>.iam.gserviceaccount.com" # GCP only
  platformBackend: # Optional, only required if using IAM role for service account on AWS, workload identity on GKE, or workload identity on AKS
    deployment: # Azure only
      labels:
        azure.workload.identity/use: true
    serviceAccount:
      annotations:
        azure.workload.identity/client-id: "<client_id>" # Azure only
        eks.amazonaws.com/role-arn: "<role_arn>" # AWS only
        iam.gke.io/gcp-service-account: "<gsa_name>@<project_id>.iam.gserviceaccount.com" # GCP only
  queue: # Optional, only required if using IAM role for service account on AWS, workload identity on GKE, or workload identity on AKS
    deployment: # Azure only
      labels:
        azure.workload.identity/use: true
    serviceAccount:
      annotations:
        azure.workload.identity/client-id: "<client_id>" # Azure only
        eks.amazonaws.com/role-arn: "<role_arn>" # AWS only
        iam.gke.io/gcp-service-account: "<gsa_name>@<project_id>.iam.gserviceaccount.com" # GCP only
  ingestQueue: # Optional, only required if using IAM role for service account on AWS, workload identity on GKE, or workload identity on AKS
    deployment: # Azure only
      labels:
        azure.workload.identity/use: true
    serviceAccount:
      annotations:
        azure.workload.identity/client-id: "<client_id>" # Azure only
        eks.amazonaws.com/role-arn: "<role_arn>" # AWS only
        iam.gke.io/gcp-service-account: "<gsa_name>@<project_id>.iam.gserviceaccount.com" # GCP only

If using an access key and secret, you can also provide an existing Kubernetes secret that contains the authentication information. This is recommended over providing the access key and secret key directly in your config. See the generated secret template for the expected secret keys.

TTL configuration

If using the TTL feature with LangSmith, you’ll also have to configure TTL rules for your blob storage. Trace information stored on blob storage is stored on a particular prefix path, which determines the TTL for the data. When a trace’s retention is extended, its corresponding blob storage path changes to ensure that it matches the new extended retention.

The following TTL prefixes are used:

• ttl_s/: Short term (base) TTL, configured for 14 days.
• ttl_l/: Long term (extended) TTL, configured for 400 days by default.

Custom workspace-level retention prefixes

If you use workspace-level extended retention, LangSmith writes blob data to prefixes of the form ttl_XXd/, where XX is the number of days configured for that workspace. For example, if a workspace is configured with 90-day extended retention, blob data for that workspace is written to the ttl_90d/ prefix.

You must create a lifecycle rule for each custom retention period configured across your workspaces. Common examples:

• ttl_90d/ — 90-day retention
• ttl_180d/ — 180-day retention
• ttl_365d/ — 365-day retention

If a lifecycle rule is missing for a configured retention period, blob data under that prefix will never be automatically deleted. Ensure you add a matching lifecycle rule whenever you configure a new workspace retention period.

For example, if you have workspaces configured with 90-day and 180-day extended retention, you would add the following lifecycle rules in addition to the default ttl_s and ttl_l rules:

AWS

rule {
  id      = "ttl-90d"
  prefix  = "ttl_90d/"
  enabled = true
  expiration {
    days = 90
  }
}
rule {
  id      = "ttl-180d"
  prefix  = "ttl_180d/"
  enabled = true
  expiration {
    days = 180
  }
}

GCP

lifecycle_rule {
  condition {
    age            = 90
    matches_prefix = ["ttl_90d"]
  }
  action {
    type = "Delete"
  }
}
lifecycle_rule {
  condition {
    age            = 180
    matches_prefix = ["ttl_180d"]
  }
  action {
    type = "Delete"
  }
}

Azure

rule {
  name    = "ttl-90d"
  enabled = true
  type    = "Lifecycle"
  filters {
    prefix_match = ["my-container/ttl_90d"]
    blob_types   = ["blockBlob"]
  }
  actions {
    base_blob {
      delete_after_days_since_creation_greater_than = 90
    }
    snapshot {
      delete_after_days_since_creation_greater_than = 90
    }
    version {
      delete_after_days_since_creation_greater_than = 90
    }
  }
}
rule {
  name    = "ttl-180d"
  enabled = true
  type    = "Lifecycle"
  filters {
    prefix_match = ["my-container/ttl_180d"]
    blob_types   = ["blockBlob"]
  }
  actions {
    base_blob {
      delete_after_days_since_creation_greater_than = 180
    }
    snapshot {
      delete_after_days_since_creation_greater_than = 180
    }
    version {
      delete_after_days_since_creation_greater_than = 180
    }
  }
}

If you have customized the TTLs in your LangSmith configuration, you will need to adjust the TTLs in your blob storage configuration to match.

AWS

Amazon S3 lifecycle rules

If using S3 for your blob storage, you will need to setup a filter lifecycle configuration that matches the prefixes above. You can find information for this in the Amazon Documentation.

As an example, if you are using Terraform to manage your S3 bucket, you would setup something like this:

rule {
  id      = "short-term-ttl"
  prefix  = "ttl_s/"
  enabled = true
  expiration {
    days = 14
  }
}
rule {
  id      = "long-term-ttl"
  prefix  = "ttl_l/"
  enabled = true
  expiration {
    days = 400
  }
}

GCP

Google Cloud Storage lifecycle rules

You will need to setup lifecycle conditions for your GCS buckets that you are using. You can find information for this in the Google Documentation, specifically using matchesPrefix.

As an example, if you are using Terraform to manage your GCS bucket, you would setup something like this:

lifecycle_rule {
  condition {
    age            = 14
    matches_prefix = ["ttl_s"]
  }
  action {
    type = "Delete"
  }
}
lifecycle_rule {
  condition {
    age            = 400
    matches_prefix = ["ttl_l"]
  }
  action {
    type = "Delete"
  }
}

Azure

Azure blob storage lifecycle management

You will need to configure a lifecycle management policy on the container in order to expire objects matching the prefixes above.

As an example, if you are using Terraform to manage your blob storage container, you would setup something like this:

resource "azurerm_storage_management_policy" "example" {
  storage_account_id = "my-storage-account-id"
  rule {
    name = "base"
    enabled = true
    type = "Lifecycle"
    filters {
      prefix_match = ["my-container/ttl_s"]
      blob_types = ["blockBlob"]
    }
    actions {
      base_blob {
        delete_after_days_since_creation_greater_than = 14
      }
      snapshot {
        delete_after_days_since_creation_greater_than = 14
      }
      version {
        delete_after_days_since_creation_greater_than = 14
      }
    }
  }
  rule {
    name = "extended"
    enabled = true
    type = "Lifecycle"
    filters {
      prefix_match = ["my-container/ttl_l"]
      blob_types = ["blockBlob"]
    }
    actions {
      base_blob {
        delete_after_days_since_creation_greater_than = 400
      }
      snapshot {
        delete_after_days_since_creation_greater_than = 400
      }
      version {
        delete_after_days_since_creation_greater_than = 400
      }
    }
  }
}

---

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.

Edit this page on GitHub or file an issue.