Project Settings
Many settings are also available at the Project level, allowing them to be configured for specific tenants on the server rather than globally. Only Super Admin users are allowed to edit Project settings.
Additional details are available in the full Project resource schema.
| Setting | Description | Default |
|---|---|---|
superAdmin | Users belonging to a Project with this flag are granted Super Admin access to the server. Multiple Projects can have this set. | false |
checkReferencesOnWrite | If true, the the server will reject any create or write operations to FHIR resources with a reference to a resource that does not exist. | false |
features | A list of optional features that are enabled for the project. Possible values are listed below. | |
defaultPatientAccessPolicy | Deprecated. Prefer the Patient entry in defaultAccessPolicies. Used only when that entry is absent, for patient invites, SCIM patient provisioning, and open patient registration. Open registration also requires the openPatientRegistration setting. | |
link | Additional Projects whose contents should be accessible to users in the current Project. | |
defaultProfile | Resource profiles that will be added to resources written in the Project that do not specify a profile directly. This enables automatic custom resource validation. See Default Profiles for details and examples. | |
setting | Arbitrary key-value pairs available to anyone in the Project, can be set by Project Admins. | |
secret | Key-value pairs similar to setting, that can only be read by Project Admins. These can be used to pass secrets to Bots | |
systemSetting | Server settings related to the Project: visible to anyone, but can only be set by Super Admins. | |
systemSecret | Key-value pairs that can only be accessed by Super Admins. |
Project feature flags
Medplum server exposes settings to control access to specific features on a per-Project basis. The available features are:
| Feature | Description |
|---|---|
bots | Project is allowed to create and run Bots |
cron | Can run Bots periodically on CRON timers |
email | Bots in this project can send emails |
google-auth-required | Google authentication is the only method allowed |
graphql-introspection | Allows potentially-expensive GraphQL schema introspection queries |
terminology | Enable full standards-compliant implementation for the ValueSet/$expand operation |
websocket-subscriptions | Allows setting up a Subscription over Websockets |
transaction-bundles | Enable atomic processing of transaction Bundles in a self-hosted deployment. Unavailable on hosted Medplum. See configuration and limits. |
range-search | Date searches compare the full period or range. See Searching by due date range. |
Atomic transaction Bundles
Atomic transaction processing is currently unavailable on Medplum-hosted deployments. Self-hosted operators can enable transaction-bundles in Project.features using a super administrator account.
Enable this only for workflows that require several writes to commit together or all roll back. Transactions can increase database contention and affect other workloads. Test the actual resource mix, concurrency, and failure recovery before enabling the feature in production. Prefer batches of conditional upserts when partial results can be reconciled.
Without the flag, type: "transaction" is processed with batch semantics, without a warning or rollback of successful entries. A transaction-response label does not establish atomicity. See FHIR Batch Requests.
Transactions have these limits; requests that exceed them are rejected with 400 Bad Request:
- At most 50 update operations.
- Transactions containing conditional creates, updates, or deletes use serializable isolation and are limited to 8 entries.
- Atomic transactions must be synchronous. Requests with
Prefer: respond-asyncare rejected when this feature is enabled.
Keep each transaction limited to the resources that must commit together. Do not replace conditional writes with search-then-create to avoid the limits: another writer can change the result between the search and the write.
Project system settings
The supported options that can be specified by a Super Admin in Project.systemSetting:
| systemSetting | Type | Description | Default |
|---|---|---|---|
authRateLimit | integer | The maximum requests per minute allowed for authentication endpoints per IP address | 160 |
graphqlBatchedSearchSize | integer | For some GraphQL nested searches, the maximum number of searches to batch per SQL query | 0 |
graphqlMaxDepth | integer | The maximum allowed depth of a GraphQL query | 12 |
graphqlMaxSearches | integer | The maximum number of searches allowed in a GraphQL query | none |
legacyExternalAuth | boolean | If true, ClientApplication identity providers in this Project may keep using the shared /auth/external callback instead of /auth/external/<clientId>. Temporary, for migration | false |
legacyFhirJsonResponseFormat | boolean | If true, plain JSON formatting is incorrectly used instead of FHIR JSON for some responses with content type application/fhir+json | false |
rateLimit | integer | The maximum requests per minute allowed per IP address | 60000 |
userFhirQuota | integer | The maximum number of FHIR interactions that can be performed in a minute by any single User in the project. See Fhir Interaction Quota. To enforce, set enableFhirQuota to true. | 50000 |
totalFhirQuota | integer | Similar to userFhirQuota, but calculated as a sum of all Users' FHIR interactions in the project. To enforce, set enableFhirQuota to true. | 500000 |
enableFhirQuota | boolean | If true, the totalFhirQuota limit will be enforced, returning 429 Too Many Requests errors when the limit is exceeded over a minute. Please note that as of v4.1.6, FHIR quotas are enabled by default. | true |
searchOnReader | boolean | If true, FHIR search requests (except in batch requests) are served by the reader database pool if available | false |
chainedSearchOnReader | boolean | If true, FHIR search requests with chained search parameters (except in batch requests) are served by the reader database pool if available | false |
redactAuditEvents | boolean | If true, remove human-readable detail strings from AuditEvent resources saved to the database and logs | false |
Open patient registration
Set the openPatientRegistration key in Project.setting to valueBoolean: true to allow patients to self-register. The project must also have a Patient access policy: the Patient entry in defaultAccessPolicies, or the deprecated defaultPatientAccessPolicy when that entry is absent. Self-registered patients receive that policy. A Patient policy without openPatientRegistration set to true does not enable registration. See Open Patient Registration.
Multi-factor authentication
Set the mfaRequired key in Project.setting to valueBoolean: true to require MFA for all existing and new users who
sign in to the Project with a username and password. Users who have not enrolled will be required to enroll during their
next password login. This project-wide requirement takes precedence over an individual user's User.mfaRequired value;
setting the user value to false does not create an exception. See
Requiring MFA for all Project users for configuration examples and
details.
The MFA methods users in a Project can enroll in are controlled by the allowedMfaMethods key in Project.setting, a comma-delimited list of totp and/or email. When unset, only authenticator-app (TOTP) MFA is offered. See Configuring allowed MFA methods for details and examples, including how to enable email-based MFA.
The appName key in Project.setting white-labels the MFA content Medplum generates: emailed verification codes and the authenticator app entries created by enrollment QR codes. When unset, both name Medplum. See Branding MFA emails and authenticator apps.
Project SMTP
Projects can send email through their own SMTP relay instead of the server-wide email provider, configured via Project.secret entries. See Project SMTP for the full configuration reference. Operators can disable this fleet-wide with the allowProjectSmtp server config setting.