Skip to main content
LangSmith uses Redis to back our queuing/caching operations. By default, LangSmith Self-Hosted will use an internal Redis instance. However, you can configure LangSmith to use an external Redis instance. By configuring an external Redis instance, you can more easily manage backups, scaling, and other operational tasks for your Redis instance.

​
Requirements

  • A provisioned Redis instance that your LangSmith instance will have network access to. We recommend using a managed Redis service like:
  • Note: We only officially support Redis versions >= 5.
  • By default, we recommend an instance with at least 2 vCPUs and 8GB of memory. However, the actual requirements will depend on your tracing workload. We recommend monitoring your Redis instance and scaling up as needed.
We now support Redis Cluster for LangSmith as of helm version 0.12.25. You can find the helm values to configure under redis.external.cluster, and an example configuration here

​
Connection String

We use redis-py to connect to Redis. This library supports a variety of connection strings. You can find more information on the connection string format here. You will need to assemble the connection string for your Redis instance. This connection string should include the following information:
  • Host
  • Database
  • Port
  • URL params
This will take the form of:
An example connection string might look like:
To use SSL, you can use the rediss:// prefix. An example connection string with SSL might look like:

​
Configuration

With your connection string in hand, you can configure your LangSmith instance to use an external Redis instance. You can do this by modifying the values file for your LangSmith Helm Chart installation or the .env file for your Docker installation.
Once configured, you should be able to reinstall your LangSmith instance. If everything is configured correctly, your LangSmith instance should now be using your external Redis instance.
Edit the source of this page on GitHub.
Connect these docs programmatically to Claude, VSCode, and more via MCP for real-time answers.