You can deploy the Literal AI platform anywhere by provisioning the infrastructure, setting up the environment variables and running the Docker image.

​
Get the Image

docker login --username literalai
# When prompted for a password, use the token you have been provided with
docker pull literalai/platform:latest

​
Provision the Infrastructure

To run the Literal AI platform, you need to set up the following services:

  1. PostgreSQL database >= 15
  2. File storage (AWS S3, Google Cloud Storage, Azure Blob Storage)
  3. [Optional] Redis cache
  4. [Optional] SMTP server

​
Disclaimer

It is your responsibility to ensure that the infrastructure is secure and compliant with your company standards. Here are some recommendations:

  • Run the database and redis cache in a private network so that only the container running the Literal AI platform can access them.
  • Disallow public access to the file storage.
  • Disable credential authentication and use OAuth providers for authentication.

​
Configure Environment Variables

Before running the container, you need to set up the environment variables.

​
1. Database Configuration

The pgcrypto module is required for PostgreSQL. When using Azure, you will need to manually add the extension.

​
Required

You can either pass the full connection string or the individual components.

VariableDescriptionExample
DATABASE_URLURL for database communication (PostgreSQL supported).postgresql://username:password@host:port/db_name

OR

VariableDescription
DATABASE_HOSTHost of the database.
DATABASE_USERNAMEDatabase username.
DATABASE_PASSWORDPassword for the DB user.
DATABASE_NAMEDatabase name.

​
Optional

VariableDescriptionExample
DATABASE_DIRECT_URLDirect URL for database migrations (cannot use pg-bouncer).postgresql://username:password@host:port/db_name
DATABASE_SSLSet to ‘true’ if using a secure network.false

​
2. Authentication Configuration

​
Provider Agnostic

VariableDescriptionExample
NEXTAUTH_SECRETSecret for encrypting JWT and hashing tokens.your secret
NEXTAUTH_URLURL for your domain.https://yourdomain.com
ENABLE_CREDENTIALS_AUTHEnable credentials (email password) authentication.false

​
Provider Specific

Remember you will have to allow the OAuth redirect URL in your provider’s settings.
VariableDescriptionOAuth Redirect URL
AzureAZURE_AD_CLIENT_ID AZURE_AD_CLIENT_SECRET AZURE_AD_TENANT_ID/api/auth/callback/azure-ad
GoogleGOOGLE_CLIENT_ID GOOGLE_CLIENT_SECRET/api/auth/callback/google
OktaOKTA_CLIENT_ID OKTA_CLIENT_SECRET OKTA_ISSUER/api/auth/callback/okta
GitHubGITHUB_ID GITHUB_SECRET/api/auth/callback/github

​
3. File Storage Configuration

​
Provider Agnostic

VariableDescriptionExample
BUCKET_NAMEName of the bucketmy_bucket

​
Provider Specific

ProviderVariablesComment
AWS S3APP_AWS_ACCESS_KEY APP_AWS_SECRET_KEY APP_AWS_REGION
Google Cloud StorageAPP_GCS_PROJECT_ID APP_GCS_CLIENT_EMAIL APP_GCS_PRIVATE_KEYPrivate key should be base64 encoded.
Azure Blob StorageAPP_AZURE_STORAGE_ACCOUNT APP_AZURE_STORAGE_ACCESS_KEY

​
Configure CORS for your storage solution:

​
4. Cache Configuration [Optional]

Setting up a cache is optional, but recommended to decrease latency.

Either REDIS_URL or REDISHOST and REDISPORT can be used.

VariableDescriptionExample
REDIS_URLFull Redis URL.redis://127.0.0.1:6379/0
REDISHOSTRedis host (useful with Google Cloud Run).127.0.0.1
REDISPORTRedis port.6379

The cache allows for asynchronous task handling such as:

  • step ingestion
  • AI evals monitoring
  • experiment runs

Control over the queues and workers handling the asynchronous tasks requires an authentication token defined via the ADMIN_AUTH_TOKEN environment variable. Set it to a non-empty value to be able to investigate your cache-based queueing system.

​
5. SMTP Configuration [Optional]

Setting up SMTP is optional. It enables Literal AI to send emails for password resets and project invitations.
VariableDescriptionExample
EMAIL_SERVER_HOSTMail server host."servicehost"
EMAIL_SERVER_PORTMail server port."4000"
EMAIL_SERVICEEmail service provider."gmail"
EMAIL_USEREmail username."username"
EMAIL_PASSEmail password."password"
EMAIL_FROMSender email address."noreply@service.com"

​
Configure the Port and Start the Container

The server within the Docker image runs on port 3000. Ensure this port is exposed.

​
Start the Container

You should now be able to start the container and access the Literal AI Platform.