# Configuration reference
<!-- Code generated from internal/platform/config/schema.go by "scripts/task.sh go:config-docs". DO NOT EDIT. -->


The server reads each setting from these sources, highest precedence first:

1. The environment variable `MY_<KEY>` (the key in upper case).
2. The TOML file named by `MY_CONFIG` (or `my config check -config <path>`), with the keys below at the top level. A value may be a string, a whole number, or a boolean; `mail_smtp_port = 587` and `mail_smtp_port = "587"` mean the same.
3. The default.

Secrets are read only from the environment: either the variable itself or `<NAME>_FILE`, the path of a file that holds the value
(trailing newlines are removed), for Docker and Kubernetes secrets. Setting a secret in the TOML file is an error.

Adapters are always selected by an explicit key; nothing is inferred from a URL scheme.

`my config check` validates the configuration and lists every problem at once; the server refuses to start on any of them.
The problems are described in the language of the process locale (`LANGUAGE`, `LC_ALL`, `LC_MESSAGES`, `LANG`; English for `C` or an unsupported language),
followed by the technical detail in English.

A setting that a release renamed or removed is one of those problems, in the config file and in the environment alike:
the error names the new key, or tells why the setting was removed ([renamed and removed settings](#renamed-and-removed-settings)).
An unknown key in the config file is an error too. An environment variable that starts with `MY_` but sets nothing
is ignored with a warning, in the log of `my serve` and in the output of `my config check`, because the platform may add such variables:
Kubernetes adds `MY_SERVICE_HOST`, `MY_PORT` and the like for a Service named `my`, and those are ignored without a warning, as are `MY_VERSION` and `MY_HTTP_PORT`, which the Docker Compose files read from the env file that also reaches the container.

A Kubernetes Service named `my-mail-smtp` is the exception: the variable `MY_MAIL_SMTP_PORT=tcp://...` that it adds is the name of a setting, so the value fails its check and the server does not start.
Rename the Service, or set `enableServiceLinks: false` on the pod.

In the `production` profile, `my config check` then checks that the valid settings suit a production instance:
errors, such as a `host` that other servers cannot reach or a `kek` that is not random, make it exit 1, and warnings, such as `mail = none`, do not.
`my serve` logs both at start without refusing to start. The list of checks is in the [production checklist](https://mycroblog.dev/next/operations/production-checklist/).

## Configuration and instance settings

The settings on this page configure the deployment: what a process runs, which adapters it uses, where they are, and its secrets.
These values are fixed for the life of a process and change only with a restart.
The policies of the instance that administrators change while it runs (registration mode, maximum post length,
default locale, which timelines are enabled) and its localized texts (name, descriptions, terms, privacy policy)
are instance settings instead: they are stored in the database with every change kept as a revision,
and are changed with `my instance set` ([instance settings](https://mycroblog.dev/next/operations/instance-settings/)).

The two never share a key, so neither overrides the other: no environment variable or config file key sets an
instance setting, and the database never holds a setting from this page. A test keeps the two key sets disjoint.
An instance setting that has never been set reads as its default, which is defined in code.

## Summary

| Key | Environment variable | Default | Required | Secret |
| --- | --- | --- | --- | --- |
| [`host`](#host) | `MY_HOST` | none | yes | no |
| [`account_domain`](#account_domain) | `MY_ACCOUNT_DOMAIN` | value of `host` | no | no |
| [`source_url`](#source_url) | `MY_SOURCE_URL` | `https://github.com/mycroblog/my` | no | no |
| [`status_page_url`](#status_page_url) | `MY_STATUS_PAGE_URL` | none | no | no |
| [`maintenance`](#maintenance) | `MY_MAINTENANCE` | `false` | no | no |
| [`maintenance_server_name`](#maintenance_server_name) | `MY_MAINTENANCE_SERVER_NAME` | value of `host` | no | no |
| [`maintenance_message`](#maintenance_message) | `MY_MAINTENANCE_MESSAGE` | none | no | no |
| [`maintenance_until`](#maintenance_until) | `MY_MAINTENANCE_UNTIL` | none | no | no |
| [`maintenance_retry_after`](#maintenance_retry_after) | `MY_MAINTENANCE_RETRY_AFTER` | `10m` | no | no |
| [`web_origin`](#web_origin) | `MY_WEB_ORIGIN` | `https://{host}` | no | no |
| [`web_early_hints`](#web_early_hints) | `MY_WEB_EARLY_HINTS` | `false` | no | no |
| [`role`](#role) | `MY_ROLE` | `all` | no | no |
| [`worker_job_types`](#worker_job_types) | `MY_WORKER_JOB_TYPES` | none | no | no |
| [`worker_job_types_except`](#worker_job_types_except) | `MY_WORKER_JOB_TYPES_EXCEPT` | none | no | no |
| [`worker_job_concurrency`](#worker_job_concurrency) | `MY_WORKER_JOB_CONCURRENCY` | `1` | no | no |
| [`listen`](#listen) | `MY_LISTEN` | `:443` with `tls` other than `off` in the `web` and `all` roles, `:8080` otherwise | no | no |
| [`tls`](#tls) | `MY_TLS` | `off` | no | no |
| [`tls_acme_directory`](#tls_acme_directory) | `MY_TLS_ACME_DIRECTORY` | `https://acme-v02.api.letsencrypt.org/directory` | no | no |
| [`tls_acme_agree_terms`](#tls_acme_agree_terms) | `MY_TLS_ACME_AGREE_TERMS` | `false` | `tls = acme` (set to `true`) | no |
| [`tls_acme_email`](#tls_acme_email) | `MY_TLS_ACME_EMAIL` | none | no | no |
| [`tls_acme_profile`](#tls_acme_profile) | `MY_TLS_ACME_PROFILE` | none | no | no |
| [`tls_cert_file`](#tls_cert_file) | `MY_TLS_CERT_FILE` | none | `tls = files` | no |
| [`tls_key_file`](#tls_key_file) | `MY_TLS_KEY_FILE` | none | `tls = files` | no |
| [`tls_http_listen`](#tls_http_listen) | `MY_TLS_HTTP_LISTEN` | `:80` | no | no |
| [`trusted_proxies`](#trusted_proxies) | `MY_TRUSTED_PROXIES` | none | no | no |
| [`trusted_proxy_header`](#trusted_proxy_header) | `MY_TRUSTED_PROXY_HEADER` | `x-forwarded-for` | no | no |
| [`origin_secret`](#origin_secret) | `MY_ORIGIN_SECRET` | none | no | yes |
| [`origin_secret_previous`](#origin_secret_previous) | `MY_ORIGIN_SECRET_PREVIOUS` | none | no | yes |
| [`origin_secret_header`](#origin_secret_header) | `MY_ORIGIN_SECRET_HEADER` | none | when `origin_secret` is set | no |
| [`database`](#database) | `MY_DATABASE` | `sqlite` | no | no |
| [`database_url`](#database_url) | `MY_DATABASE_URL` | none | yes | yes |
| [`database_acquire_timeout`](#database_acquire_timeout) | `MY_DATABASE_ACQUIRE_TIMEOUT` | `60s` | no | no |
| [`database_dev_url`](#database_dev_url) | `MY_DATABASE_DEV_URL` | a temporary schema in the database of `database_url` | no | yes |
| [`queue`](#queue) | `MY_QUEUE` | `db` | no | no |
| [`queue_cloudtasks_routes`](#queue_cloudtasks_routes) | `MY_QUEUE_CLOUDTASKS_ROUTES` | none | no | no |
| [`queue_cloudtasks_queue`](#queue_cloudtasks_queue) | `MY_QUEUE_CLOUDTASKS_QUEUE` | none | when `queue = cloudtasks` | no |
| [`queue_cloudtasks_url`](#queue_cloudtasks_url) | `MY_QUEUE_CLOUDTASKS_URL` | `https://{host}` | no | no |
| [`queue_cloudtasks_service_account`](#queue_cloudtasks_service_account) | `MY_QUEUE_CLOUDTASKS_SERVICE_ACCOUNT` | none | when `queue = cloudtasks` | no |
| [`queue_cloudtasks_audience`](#queue_cloudtasks_audience) | `MY_QUEUE_CLOUDTASKS_AUDIENCE` | the URL of the endpoint called | no | no |
| [`queue_cloudtasks_dispatch_deadline`](#queue_cloudtasks_dispatch_deadline) | `MY_QUEUE_CLOUDTASKS_DISPATCH_DEADLINE` | `10m` | no | no |
| [`blob`](#blob) | `MY_BLOB` | `fs` | no | no |
| [`media_image`](#media_image) | `MY_MEDIA_IMAGE` | `native` | no | no |
| [`media_video`](#media_video) | `MY_MEDIA_VIDEO` | `native` | no | no |
| [`media_audio`](#media_audio) | `MY_MEDIA_AUDIO` | `native` | no | no |
| [`blob_fs_dir`](#blob_fs_dir) | `MY_BLOB_FS_DIR` | `blobs` | no | no |
| [`blob_s3_bucket`](#blob_s3_bucket) | `MY_BLOB_S3_BUCKET` | none | when `blob = s3` | no |
| [`blob_s3_endpoint`](#blob_s3_endpoint) | `MY_BLOB_S3_ENDPOINT` | none | no | no |
| [`blob_s3_region`](#blob_s3_region) | `MY_BLOB_S3_REGION` | none | no | no |
| [`blob_s3_path_style`](#blob_s3_path_style) | `MY_BLOB_S3_PATH_STYLE` | `false` | no | no |
| [`blob_s3_access_key_id`](#blob_s3_access_key_id) | `MY_BLOB_S3_ACCESS_KEY_ID` | none | no | no |
| [`blob_s3_secret_access_key`](#blob_s3_secret_access_key) | `MY_BLOB_S3_SECRET_ACCESS_KEY` | none | when `blob_s3_access_key_id` is set | yes |
| [`blob_gcs_bucket`](#blob_gcs_bucket) | `MY_BLOB_GCS_BUCKET` | none | when `blob = gcs` | no |
| [`media_origin`](#media_origin) | `MY_MEDIA_ORIGIN` | the server's own origin (`https://{host}`) | no | no |
| [`media_max_image_bytes`](#media_max_image_bytes) | `MY_MEDIA_MAX_IMAGE_BYTES` | `16777216` | no | no |
| [`media_max_image_side`](#media_max_image_side) | `MY_MEDIA_MAX_IMAGE_SIDE` | `8192` | no | no |
| [`media_max_image_pixels`](#media_max_image_pixels) | `MY_MEDIA_MAX_IMAGE_PIXELS` | `33177600` | no | no |
| [`media_unattached_retention`](#media_unattached_retention) | `MY_MEDIA_UNATTACHED_RETENTION` | `24h` | no | no |
| [`media_replaced_retention`](#media_replaced_retention) | `MY_MEDIA_REPLACED_RETENTION` | `720h` | no | no |
| [`media_max_av_bytes`](#media_max_av_bytes) | `MY_MEDIA_MAX_AV_BYTES` | `104857600` | no | no |
| [`media_max_video_duration`](#media_max_video_duration) | `MY_MEDIA_MAX_VIDEO_DURATION` | `5m` | no | no |
| [`media_max_audio_duration`](#media_max_audio_duration) | `MY_MEDIA_MAX_AUDIO_DURATION` | `10m` | no | no |
| [`custom_emoji_max_bytes`](#custom_emoji_max_bytes) | `MY_CUSTOM_EMOJI_MAX_BYTES` | `262144` | no | no |
| [`remote_custom_emoji_max_bytes`](#remote_custom_emoji_max_bytes) | `MY_REMOTE_CUSTOM_EMOJI_MAX_BYTES` | `1048576` | no | no |
| [`kv`](#kv) | `MY_KV` | `db` | no | no |
| [`valkey_url`](#valkey_url) | `MY_VALKEY_URL` | none | when `kv = valkey` or `stream = valkey` | yes |
| [`valkey_tls_ca_file`](#valkey_tls_ca_file) | `MY_VALKEY_TLS_CA_FILE` | none | no | no |
| [`valkey_key_prefix`](#valkey_key_prefix) | `MY_VALKEY_KEY_PREFIX` | `my:` | no | no |
| [`stream`](#stream) | `MY_STREAM` | `hub` | no | no |
| [`stream_postgres_url`](#stream_postgres_url) | `MY_STREAM_POSTGRES_URL` | value of `database_url` | no | yes |
| [`stream_sns_topic_arn`](#stream_sns_topic_arn) | `MY_STREAM_SNS_TOPIC_ARN` | none | when `stream = sns` | no |
| [`stream_sns_queue_prefix`](#stream_sns_queue_prefix) | `MY_STREAM_SNS_QUEUE_PREFIX` | none | when `stream = sns` | no |
| [`stream_pubsub_topic`](#stream_pubsub_topic) | `MY_STREAM_PUBSUB_TOPIC` | none | when `stream = pubsub` | no |
| [`stream_pubsub_subscription_prefix`](#stream_pubsub_subscription_prefix) | `MY_STREAM_PUBSUB_SUBSCRIPTION_PREFIX` | none | when `stream = pubsub` | no |
| [`key_manager`](#key_manager) | `MY_KEY_MANAGER` | `local` | no | no |
| [`kek`](#kek) | `MY_KEK` | none | when `key_manager = local` | yes |
| [`kek_previous`](#kek_previous) | `MY_KEK_PREVIOUS` | none | no | yes |
| [`kms_key`](#kms_key) | `MY_KMS_KEY` | none | when `key_manager` is `gcpkms` or `awskms` | no |
| [`kms_key_previous`](#kms_key_previous) | `MY_KMS_KEY_PREVIOUS` | none | no | no |
| [`kms_dek_cache_ttl`](#kms_dek_cache_ttl) | `MY_KMS_DEK_CACHE_TTL` | `10m` | no | no |
| [`kms_dek_cache_entries`](#kms_dek_cache_entries) | `MY_KMS_DEK_CACHE_ENTRIES` | `1024` | no | no |
| [`mail`](#mail) | `MY_MAIL` | `none` | no | no |
| [`mail_from`](#mail_from) | `MY_MAIL_FROM` | none | when `mail` is not `none` | no |
| [`mail_smtp_host`](#mail_smtp_host) | `MY_MAIL_SMTP_HOST` | none | when `mail = smtp` | no |
| [`mail_smtp_port`](#mail_smtp_port) | `MY_MAIL_SMTP_PORT` | `587` | no | no |
| [`mail_smtp_username`](#mail_smtp_username) | `MY_MAIL_SMTP_USERNAME` | none | no | no |
| [`mail_smtp_password`](#mail_smtp_password) | `MY_MAIL_SMTP_PASSWORD` | none | no | yes |
| [`mail_ses_region`](#mail_ses_region) | `MY_MAIL_SES_REGION` | none | no | no |
| [`mail_ses_configuration_set`](#mail_ses_configuration_set) | `MY_MAIL_SES_CONFIGURATION_SET` | none | no | no |
| [`mail_ses_sns_topic_arn`](#mail_ses_sns_topic_arn) | `MY_MAIL_SES_SNS_TOPIC_ARN` | none | no | no |
| [`captcha`](#captcha) | `MY_CAPTCHA` | `none` | no | no |
| [`captcha_site_key`](#captcha_site_key) | `MY_CAPTCHA_SITE_KEY` | none | when `captcha` is not `none` | no |
| [`captcha_secret_key`](#captcha_secret_key) | `MY_CAPTCHA_SECRET_KEY` | none | when `captcha` is not `none` | yes |
| [`vapid_private_key`](#vapid_private_key) | `MY_VAPID_PRIVATE_KEY` | none | no | yes |
| [`profile`](#profile) | `MY_PROFILE` | `production` | no | no |
| [`http_allow_private_networks`](#http_allow_private_networks) | `MY_HTTP_ALLOW_PRIVATE_NETWORKS` | `false` | no | no |
| [`webhook_allowed_destinations`](#webhook_allowed_destinations) | `MY_WEBHOOK_ALLOWED_DESTINATIONS` | none | no | no |
| [`login_provider_allowed_destinations`](#login_provider_allowed_destinations) | `MY_LOGIN_PROVIDER_ALLOWED_DESTINATIONS` | none | no | no |
| [`push_allowed_destinations`](#push_allowed_destinations) | `MY_PUSH_ALLOWED_DESTINATIONS` | none | no | no |
| [`relay_max_activities_per_hour`](#relay_max_activities_per_hour) | `MY_RELAY_MAX_ACTIVITIES_PER_HOUR` | `1000` | no | no |
| [`billing_sources`](#billing_sources) | `MY_BILLING_SOURCES` | none | no | no |
| [`billing_aws_max_requests_per_day`](#billing_aws_max_requests_per_day) | `MY_BILLING_AWS_MAX_REQUESTS_PER_DAY` | `4` | no | no |
| [`billing_aws_max_requests_per_month`](#billing_aws_max_requests_per_month) | `MY_BILLING_AWS_MAX_REQUESTS_PER_MONTH` | `120` | no | no |
| [`billing_aws_metric`](#billing_aws_metric) | `MY_BILLING_AWS_METRIC` | `NetUnblendedCost` | no | no |
| [`billing_aws_filter`](#billing_aws_filter) | `MY_BILLING_AWS_FILTER` | `none` | no | no |
| [`billing_aws_tag`](#billing_aws_tag) | `MY_BILLING_AWS_TAG` | none | when `billing_aws_filter = tag` | no |
| [`billing_aws_linked_accounts`](#billing_aws_linked_accounts) | `MY_BILLING_AWS_LINKED_ACCOUNTS` | none | when `billing_aws_filter = linked_account` | no |
| [`billing_aws_role_arn`](#billing_aws_role_arn) | `MY_BILLING_AWS_ROLE_ARN` | none | no | no |
| [`billing_gcp_method`](#billing_gcp_method) | `MY_BILLING_GCP_METHOD` | `bigquery` | no | no |
| [`billing_gcp_table`](#billing_gcp_table) | `MY_BILLING_GCP_TABLE` | none | when `billing_sources` includes `gcp` and `billing_gcp_method = bigquery` | no |
| [`billing_gcp_query_project`](#billing_gcp_query_project) | `MY_BILLING_GCP_QUERY_PROJECT` | none | no | no |
| [`billing_gcp_projects`](#billing_gcp_projects) | `MY_BILLING_GCP_PROJECTS` | none | no | no |
| [`billing_gcp_label`](#billing_gcp_label) | `MY_BILLING_GCP_LABEL` | none | no | no |
| [`billing_gcp_max_bytes_billed`](#billing_gcp_max_bytes_billed) | `MY_BILLING_GCP_MAX_BYTES_BILLED` | `1073741824` | no | no |
| [`billing_gcp_budget_subscription`](#billing_gcp_budget_subscription) | `MY_BILLING_GCP_BUDGET_SUBSCRIPTION` | none | when `billing_sources` includes `gcp` and `billing_gcp_method = budget` | no |
| [`billing_cloudflare_account_id`](#billing_cloudflare_account_id) | `MY_BILLING_CLOUDFLARE_ACCOUNT_ID` | none | when `billing_sources` includes `cloudflare` | no |
| [`billing_cloudflare_api_token`](#billing_cloudflare_api_token) | `MY_BILLING_CLOUDFLARE_API_TOKEN` | none | when `billing_sources` includes `cloudflare` | yes |
| [`payments`](#payments) | `MY_PAYMENTS` | `none` | no | no |
| [`payments_stripe_secret_key`](#payments_stripe_secret_key) | `MY_PAYMENTS_STRIPE_SECRET_KEY` | none | when `payments = stripe` | yes |
| [`payments_stripe_webhook_secret`](#payments_stripe_webhook_secret) | `MY_PAYMENTS_STRIPE_WEBHOOK_SECRET` | none | when `payments = stripe` | yes |
| [`payments_payjp_secret_key`](#payments_payjp_secret_key) | `MY_PAYMENTS_PAYJP_SECRET_KEY` | none | when `payments = payjp` | yes |
| [`payments_payjp_public_key`](#payments_payjp_public_key) | `MY_PAYMENTS_PAYJP_PUBLIC_KEY` | none | when `payments = payjp` | no |
| [`payments_payjp_webhook_token`](#payments_payjp_webhook_token) | `MY_PAYMENTS_PAYJP_WEBHOOK_TOKEN` | none | when `payments = payjp` | yes |
| [`payments_reconcile_interval`](#payments_reconcile_interval) | `MY_PAYMENTS_RECONCILE_INTERVAL` | `6h` | no | no |
| [`payments_reconcile_max_requests`](#payments_reconcile_max_requests) | `MY_PAYMENTS_RECONCILE_MAX_REQUESTS` | `1000` | no | no |
| [`translator_libretranslate_url`](#translator_libretranslate_url) | `MY_TRANSLATOR_LIBRETRANSLATE_URL` | none | no | no |
| [`translator_libretranslate_api_key`](#translator_libretranslate_api_key) | `MY_TRANSLATOR_LIBRETRANSLATE_API_KEY` | none | no | yes |
| [`translator_deepl_api_key`](#translator_deepl_api_key) | `MY_TRANSLATOR_DEEPL_API_KEY` | none | no | yes |
| [`translator_deepl_url`](#translator_deepl_url) | `MY_TRANSLATOR_DEEPL_URL` | `https://api-free.deepl.com` for a key of the free plan, `https://api.deepl.com` otherwise | no | no |
| [`translator_google_project`](#translator_google_project) | `MY_TRANSLATOR_GOOGLE_PROJECT` | none | no | no |
| [`translator_google_location`](#translator_google_location) | `MY_TRANSLATOR_GOOGLE_LOCATION` | `global` | no | no |
| [`translator_aws_region`](#translator_aws_region) | `MY_TRANSLATOR_AWS_REGION` | none | no | no |
| [`translator_fake`](#translator_fake) | `MY_TRANSLATOR_FAKE` | `false` | no | no |
| [`search`](#search) | `MY_SEARCH` | `none` | no | no |
| [`search_elasticsearch_url`](#search_elasticsearch_url) | `MY_SEARCH_ELASTICSEARCH_URL` | none | when `search = elasticsearch` | no |
| [`search_elasticsearch_api_key`](#search_elasticsearch_api_key) | `MY_SEARCH_ELASTICSEARCH_API_KEY` | none | no | yes |
| [`search_elasticsearch_username`](#search_elasticsearch_username) | `MY_SEARCH_ELASTICSEARCH_USERNAME` | none | no | no |
| [`search_elasticsearch_password`](#search_elasticsearch_password) | `MY_SEARCH_ELASTICSEARCH_PASSWORD` | none | when `search_elasticsearch_username` is set | yes |
| [`search_elasticsearch_index`](#search_elasticsearch_index) | `MY_SEARCH_ELASTICSEARCH_INDEX` | `my-statuses` | no | no |
| [`search_elasticsearch_index_settings_file`](#search_elasticsearch_index_settings_file) | `MY_SEARCH_ELASTICSEARCH_INDEX_SETTINGS_FILE` | none | no | no |
| [`search_elasticsearch_ca_file`](#search_elasticsearch_ca_file) | `MY_SEARCH_ELASTICSEARCH_CA_FILE` | none | no | no |
| [`search_opensearch_url`](#search_opensearch_url) | `MY_SEARCH_OPENSEARCH_URL` | none | when `search = opensearch` | no |
| [`search_opensearch_index`](#search_opensearch_index) | `MY_SEARCH_OPENSEARCH_INDEX` | `my-statuses` | no | no |
| [`search_opensearch_index_settings_file`](#search_opensearch_index_settings_file) | `MY_SEARCH_OPENSEARCH_INDEX_SETTINGS_FILE` | none | no | no |
| [`search_opensearch_username`](#search_opensearch_username) | `MY_SEARCH_OPENSEARCH_USERNAME` | none | no | no |
| [`search_opensearch_password`](#search_opensearch_password) | `MY_SEARCH_OPENSEARCH_PASSWORD` | none | when `search_opensearch_username` is set | yes |
| [`search_opensearch_aws_signing`](#search_opensearch_aws_signing) | `MY_SEARCH_OPENSEARCH_AWS_SIGNING` | `off` | no | no |
| [`search_opensearch_aws_region`](#search_opensearch_aws_region) | `MY_SEARCH_OPENSEARCH_AWS_REGION` | the region of the AWS SDK's default chain (`AWS_REGION`) | no | no |
| [`search_meilisearch_url`](#search_meilisearch_url) | `MY_SEARCH_MEILISEARCH_URL` | none | when `search = meilisearch` | no |
| [`search_meilisearch_api_key`](#search_meilisearch_api_key) | `MY_SEARCH_MEILISEARCH_API_KEY` | none | no | yes |
| [`search_meilisearch_index`](#search_meilisearch_index) | `MY_SEARCH_MEILISEARCH_INDEX` | `my_statuses` | no | no |
| [`search_meilisearch_settings_file`](#search_meilisearch_settings_file) | `MY_SEARCH_MEILISEARCH_SETTINGS_FILE` | none | no | no |
| [`voice`](#voice) | `MY_VOICE` | `off` | no | no |
| [`voice_p2p_max_participants`](#voice_p2p_max_participants) | `MY_VOICE_P2P_MAX_PARTICIPANTS` | `10` | no | no |
| [`voice_sfu_max_participants`](#voice_sfu_max_participants) | `MY_VOICE_SFU_MAX_PARTICIPANTS` | `10` | no | no |
| [`voice_max_room_duration`](#voice_max_room_duration) | `MY_VOICE_MAX_ROOM_DURATION` | `4h` | no | no |
| [`voice_max_open_rooms`](#voice_max_open_rooms) | `MY_VOICE_MAX_OPEN_ROOMS` | `10` | no | no |
| [`voice_stun_urls`](#voice_stun_urls) | `MY_VOICE_STUN_URLS` | none | no | no |
| [`voice_turn`](#voice_turn) | `MY_VOICE_TURN` | `none` | no | no |
| [`voice_turn_coturn_urls`](#voice_turn_coturn_urls) | `MY_VOICE_TURN_COTURN_URLS` | none | when `voice = p2p` and `voice_turn = coturn` | no |
| [`voice_turn_coturn_secret`](#voice_turn_coturn_secret) | `MY_VOICE_TURN_COTURN_SECRET` | none | when `voice = p2p` and `voice_turn = coturn` | yes |
| [`voice_turn_cloudflare_key_id`](#voice_turn_cloudflare_key_id) | `MY_VOICE_TURN_CLOUDFLARE_KEY_ID` | none | when `voice = p2p` and `voice_turn = cloudflare` | no |
| [`voice_turn_cloudflare_api_token`](#voice_turn_cloudflare_api_token) | `MY_VOICE_TURN_CLOUDFLARE_API_TOKEN` | none | when `voice = p2p` and `voice_turn = cloudflare` | yes |
| [`voice_turn_credential_ttl`](#voice_turn_credential_ttl) | `MY_VOICE_TURN_CREDENTIAL_TTL` | `6h` | no | no |
| [`voice_ice_transport_policy`](#voice_ice_transport_policy) | `MY_VOICE_ICE_TRANSPORT_POLICY` | `all` | no | no |
| [`voice_livekit_url`](#voice_livekit_url) | `MY_VOICE_LIVEKIT_URL` | none | when `voice = livekit` | no |
| [`voice_livekit_api_url`](#voice_livekit_api_url) | `MY_VOICE_LIVEKIT_API_URL` | `voice_livekit_url` with `https:` (`http:` for `ws:`) | no | no |
| [`voice_livekit_api_key`](#voice_livekit_api_key) | `MY_VOICE_LIVEKIT_API_KEY` | none | when `voice = livekit` | no |
| [`voice_livekit_api_secret`](#voice_livekit_api_secret) | `MY_VOICE_LIVEKIT_API_SECRET` | none | when `voice = livekit` | yes |
| [`voice_sora_mode`](#voice_sora_mode) | `MY_VOICE_SORA_MODE` | none | when `voice = sora` | no |
| [`voice_sora_signaling_urls`](#voice_sora_signaling_urls) | `MY_VOICE_SORA_SIGNALING_URLS` | none | when `voice = sora` | no |
| [`voice_sora_api_url`](#voice_sora_api_url) | `MY_VOICE_SORA_API_URL` | none | when `voice = sora` and `voice_sora_mode = self_hosted` | no |
| [`voice_sora_api_header_name`](#voice_sora_api_header_name) | `MY_VOICE_SORA_API_HEADER_NAME` | none | no | no |
| [`voice_sora_api_header_value`](#voice_sora_api_header_value) | `MY_VOICE_SORA_API_HEADER_VALUE` | none | when `voice = sora` and `voice_sora_api_header_name` is set | yes |
| [`voice_sora_webhook_username`](#voice_sora_webhook_username) | `MY_VOICE_SORA_WEBHOOK_USERNAME` | none | when `voice = sora` and `voice_sora_mode = self_hosted` | no |
| [`voice_sora_webhook_password`](#voice_sora_webhook_password) | `MY_VOICE_SORA_WEBHOOK_PASSWORD` | none | when `voice = sora` and `voice_sora_mode = self_hosted` | yes |
| [`voice_sora_cloud_project_id`](#voice_sora_cloud_project_id) | `MY_VOICE_SORA_CLOUD_PROJECT_ID` | none | when `voice = sora` and `voice_sora_mode = cloud` | no |
| [`voice_sora_cloud_api_key`](#voice_sora_cloud_api_key) | `MY_VOICE_SORA_CLOUD_API_KEY` | none | when `voice = sora` and `voice_sora_mode = cloud` | yes |
| [`voice_sora_join_secret`](#voice_sora_join_secret) | `MY_VOICE_SORA_JOIN_SECRET` | none | when `voice = sora` | yes |
| [`dead_job_retention`](#dead_job_retention) | `MY_DEAD_JOB_RETENTION` | `720h` | no | no |
| [`job_attempt_retention`](#job_attempt_retention) | `MY_JOB_ATTEMPT_RETENTION` | `720h` | no | no |
| [`dynamic_client_unredeemed_retention`](#dynamic_client_unredeemed_retention) | `MY_DYNAMIC_CLIENT_UNREDEEMED_RETENTION` | `24h` | no | no |
| [`dynamic_client_idle_retention`](#dynamic_client_idle_retention) | `MY_DYNAMIC_CLIENT_IDLE_RETENTION` | `2160h` | no | no |
| [`metrics_exporter`](#metrics_exporter) | `MY_METRICS_EXPORTER` | `none` | no | no |
| [`metrics_listen`](#metrics_listen) | `MY_METRICS_LISTEN` | `127.0.0.1:9464` | no | no |
| [`metrics_emf_namespace`](#metrics_emf_namespace) | `MY_METRICS_EMF_NAMESPACE` | `my` | no | no |
| [`metrics_emf_dimensions`](#metrics_emf_dimensions) | `MY_METRICS_EMF_DIMENSIONS` | none | no | no |
| [`traces_exporter`](#traces_exporter) | `MY_TRACES_EXPORTER` | `none` | no | no |
| [`traces_sample_ratio`](#traces_sample_ratio) | `MY_TRACES_SAMPLE_RATIO` | `0.01` | no | no |
| [`traces_flush`](#traces_flush) | `MY_TRACES_FLUSH` | `batch` | no | no |
| [`errors_exporter`](#errors_exporter) | `MY_ERRORS_EXPORTER` | `none` | no | no |
| [`errors_sentry_dsn`](#errors_sentry_dsn) | `MY_ERRORS_SENTRY_DSN` | none | when `errors_exporter` is `sentry` | yes |
| [`errors_environment`](#errors_environment) | `MY_ERRORS_ENVIRONMENT` | the value of `profile` | no | no |
| [`pprof_listen`](#pprof_listen) | `MY_PPROF_LISTEN` | none | no | no |
| [`plugin_translator_address`](#plugin_translator_address) | `MY_PLUGIN_TRANSLATOR_ADDRESS` | none | no | no |
| [`plugin_translator_protocol`](#plugin_translator_protocol) | `MY_PLUGIN_TRANSLATOR_PROTOCOL` | `json` | no | no |
| [`plugin_translator_token`](#plugin_translator_token) | `MY_PLUGIN_TRANSLATOR_TOKEN` | none | no | yes |
| [`plugin_translator_tls_ca_file`](#plugin_translator_tls_ca_file) | `MY_PLUGIN_TRANSLATOR_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_translator_tls_cert_file`](#plugin_translator_tls_cert_file) | `MY_PLUGIN_TRANSLATOR_TLS_CERT_FILE` | none | no | no |
| [`plugin_translator_tls_key_file`](#plugin_translator_tls_key_file) | `MY_PLUGIN_TRANSLATOR_TLS_KEY_FILE` | none | no | no |
| [`plugin_translator_timeout`](#plugin_translator_timeout) | `MY_PLUGIN_TRANSLATOR_TIMEOUT` | `10s` | no | no |
| [`plugin_translator_name`](#plugin_translator_name) | `MY_PLUGIN_TRANSLATOR_NAME` | none | when `plugin_translator_address` is set | no |
| [`plugin_search_address`](#plugin_search_address) | `MY_PLUGIN_SEARCH_ADDRESS` | none | when `search = plugin` | no |
| [`plugin_search_protocol`](#plugin_search_protocol) | `MY_PLUGIN_SEARCH_PROTOCOL` | `json` | no | no |
| [`plugin_search_token`](#plugin_search_token) | `MY_PLUGIN_SEARCH_TOKEN` | none | no | yes |
| [`plugin_search_tls_ca_file`](#plugin_search_tls_ca_file) | `MY_PLUGIN_SEARCH_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_search_tls_cert_file`](#plugin_search_tls_cert_file) | `MY_PLUGIN_SEARCH_TLS_CERT_FILE` | none | no | no |
| [`plugin_search_tls_key_file`](#plugin_search_tls_key_file) | `MY_PLUGIN_SEARCH_TLS_KEY_FILE` | none | no | no |
| [`plugin_search_timeout`](#plugin_search_timeout) | `MY_PLUGIN_SEARCH_TIMEOUT` | `10s` | no | no |
| [`plugin_billing_address`](#plugin_billing_address) | `MY_PLUGIN_BILLING_ADDRESS` | none | when `billing_sources` includes `plugin` | no |
| [`plugin_billing_protocol`](#plugin_billing_protocol) | `MY_PLUGIN_BILLING_PROTOCOL` | `json` | no | no |
| [`plugin_billing_token`](#plugin_billing_token) | `MY_PLUGIN_BILLING_TOKEN` | none | no | yes |
| [`plugin_billing_tls_ca_file`](#plugin_billing_tls_ca_file) | `MY_PLUGIN_BILLING_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_billing_tls_cert_file`](#plugin_billing_tls_cert_file) | `MY_PLUGIN_BILLING_TLS_CERT_FILE` | none | no | no |
| [`plugin_billing_tls_key_file`](#plugin_billing_tls_key_file) | `MY_PLUGIN_BILLING_TLS_KEY_FILE` | none | no | no |
| [`plugin_billing_timeout`](#plugin_billing_timeout) | `MY_PLUGIN_BILLING_TIMEOUT` | `2m0s` | no | no |
| [`plugin_billing_location`](#plugin_billing_location) | `MY_PLUGIN_BILLING_LOCATION` | none | when `billing_sources` includes `plugin` | no |
| [`plugin_billing_provider`](#plugin_billing_provider) | `MY_PLUGIN_BILLING_PROVIDER` | none | when `billing_sources` includes `plugin` | no |
| [`plugin_mailer_address`](#plugin_mailer_address) | `MY_PLUGIN_MAILER_ADDRESS` | none | when `mail = plugin` | no |
| [`plugin_mailer_protocol`](#plugin_mailer_protocol) | `MY_PLUGIN_MAILER_PROTOCOL` | `json` | no | no |
| [`plugin_mailer_token`](#plugin_mailer_token) | `MY_PLUGIN_MAILER_TOKEN` | none | no | yes |
| [`plugin_mailer_tls_ca_file`](#plugin_mailer_tls_ca_file) | `MY_PLUGIN_MAILER_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_mailer_tls_cert_file`](#plugin_mailer_tls_cert_file) | `MY_PLUGIN_MAILER_TLS_CERT_FILE` | none | no | no |
| [`plugin_mailer_tls_key_file`](#plugin_mailer_tls_key_file) | `MY_PLUGIN_MAILER_TLS_KEY_FILE` | none | no | no |
| [`plugin_mailer_timeout`](#plugin_mailer_timeout) | `MY_PLUGIN_MAILER_TIMEOUT` | `30s` | no | no |
| [`plugin_media_image_address`](#plugin_media_image_address) | `MY_PLUGIN_MEDIA_IMAGE_ADDRESS` | none | when `media_image = plugin` | no |
| [`plugin_media_image_protocol`](#plugin_media_image_protocol) | `MY_PLUGIN_MEDIA_IMAGE_PROTOCOL` | `proto` | no | no |
| [`plugin_media_image_token`](#plugin_media_image_token) | `MY_PLUGIN_MEDIA_IMAGE_TOKEN` | none | no | yes |
| [`plugin_media_image_tls_ca_file`](#plugin_media_image_tls_ca_file) | `MY_PLUGIN_MEDIA_IMAGE_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_media_image_tls_cert_file`](#plugin_media_image_tls_cert_file) | `MY_PLUGIN_MEDIA_IMAGE_TLS_CERT_FILE` | none | no | no |
| [`plugin_media_image_tls_key_file`](#plugin_media_image_tls_key_file) | `MY_PLUGIN_MEDIA_IMAGE_TLS_KEY_FILE` | none | no | no |
| [`plugin_media_image_timeout`](#plugin_media_image_timeout) | `MY_PLUGIN_MEDIA_IMAGE_TIMEOUT` | `4m0s` | no | no |
| [`plugin_media_video_address`](#plugin_media_video_address) | `MY_PLUGIN_MEDIA_VIDEO_ADDRESS` | none | when `media_video = plugin` | no |
| [`plugin_media_video_protocol`](#plugin_media_video_protocol) | `MY_PLUGIN_MEDIA_VIDEO_PROTOCOL` | `proto` | no | no |
| [`plugin_media_video_token`](#plugin_media_video_token) | `MY_PLUGIN_MEDIA_VIDEO_TOKEN` | none | no | yes |
| [`plugin_media_video_tls_ca_file`](#plugin_media_video_tls_ca_file) | `MY_PLUGIN_MEDIA_VIDEO_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_media_video_tls_cert_file`](#plugin_media_video_tls_cert_file) | `MY_PLUGIN_MEDIA_VIDEO_TLS_CERT_FILE` | none | no | no |
| [`plugin_media_video_tls_key_file`](#plugin_media_video_tls_key_file) | `MY_PLUGIN_MEDIA_VIDEO_TLS_KEY_FILE` | none | no | no |
| [`plugin_media_video_timeout`](#plugin_media_video_timeout) | `MY_PLUGIN_MEDIA_VIDEO_TIMEOUT` | `4m0s` | no | no |
| [`plugin_media_audio_address`](#plugin_media_audio_address) | `MY_PLUGIN_MEDIA_AUDIO_ADDRESS` | none | when `media_audio = plugin` | no |
| [`plugin_media_audio_protocol`](#plugin_media_audio_protocol) | `MY_PLUGIN_MEDIA_AUDIO_PROTOCOL` | `proto` | no | no |
| [`plugin_media_audio_token`](#plugin_media_audio_token) | `MY_PLUGIN_MEDIA_AUDIO_TOKEN` | none | no | yes |
| [`plugin_media_audio_tls_ca_file`](#plugin_media_audio_tls_ca_file) | `MY_PLUGIN_MEDIA_AUDIO_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_media_audio_tls_cert_file`](#plugin_media_audio_tls_cert_file) | `MY_PLUGIN_MEDIA_AUDIO_TLS_CERT_FILE` | none | no | no |
| [`plugin_media_audio_tls_key_file`](#plugin_media_audio_tls_key_file) | `MY_PLUGIN_MEDIA_AUDIO_TLS_KEY_FILE` | none | no | no |
| [`plugin_media_audio_timeout`](#plugin_media_audio_timeout) | `MY_PLUGIN_MEDIA_AUDIO_TIMEOUT` | `4m0s` | no | no |
| [`plugin_key_manager_address`](#plugin_key_manager_address) | `MY_PLUGIN_KEY_MANAGER_ADDRESS` | none | when `key_manager = plugin` | no |
| [`plugin_key_manager_protocol`](#plugin_key_manager_protocol) | `MY_PLUGIN_KEY_MANAGER_PROTOCOL` | `json` | no | no |
| [`plugin_key_manager_token`](#plugin_key_manager_token) | `MY_PLUGIN_KEY_MANAGER_TOKEN` | none | no | yes |
| [`plugin_key_manager_tls_ca_file`](#plugin_key_manager_tls_ca_file) | `MY_PLUGIN_KEY_MANAGER_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_key_manager_tls_cert_file`](#plugin_key_manager_tls_cert_file) | `MY_PLUGIN_KEY_MANAGER_TLS_CERT_FILE` | none | no | no |
| [`plugin_key_manager_tls_key_file`](#plugin_key_manager_tls_key_file) | `MY_PLUGIN_KEY_MANAGER_TLS_KEY_FILE` | none | no | no |
| [`plugin_key_manager_timeout`](#plugin_key_manager_timeout) | `MY_PLUGIN_KEY_MANAGER_TIMEOUT` | `5s` | no | no |
| [`plugin_inbound_status_address`](#plugin_inbound_status_address) | `MY_PLUGIN_INBOUND_STATUS_ADDRESS` | none | no | no |
| [`plugin_inbound_status_protocol`](#plugin_inbound_status_protocol) | `MY_PLUGIN_INBOUND_STATUS_PROTOCOL` | `json` | no | no |
| [`plugin_inbound_status_token`](#plugin_inbound_status_token) | `MY_PLUGIN_INBOUND_STATUS_TOKEN` | none | no | yes |
| [`plugin_inbound_status_tls_ca_file`](#plugin_inbound_status_tls_ca_file) | `MY_PLUGIN_INBOUND_STATUS_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_inbound_status_tls_cert_file`](#plugin_inbound_status_tls_cert_file) | `MY_PLUGIN_INBOUND_STATUS_TLS_CERT_FILE` | none | no | no |
| [`plugin_inbound_status_tls_key_file`](#plugin_inbound_status_tls_key_file) | `MY_PLUGIN_INBOUND_STATUS_TLS_KEY_FILE` | none | no | no |
| [`plugin_inbound_status_timeout`](#plugin_inbound_status_timeout) | `MY_PLUGIN_INBOUND_STATUS_TIMEOUT` | `3s` | no | no |
| [`plugin_inbound_follow_address`](#plugin_inbound_follow_address) | `MY_PLUGIN_INBOUND_FOLLOW_ADDRESS` | none | no | no |
| [`plugin_inbound_follow_protocol`](#plugin_inbound_follow_protocol) | `MY_PLUGIN_INBOUND_FOLLOW_PROTOCOL` | `json` | no | no |
| [`plugin_inbound_follow_token`](#plugin_inbound_follow_token) | `MY_PLUGIN_INBOUND_FOLLOW_TOKEN` | none | no | yes |
| [`plugin_inbound_follow_tls_ca_file`](#plugin_inbound_follow_tls_ca_file) | `MY_PLUGIN_INBOUND_FOLLOW_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_inbound_follow_tls_cert_file`](#plugin_inbound_follow_tls_cert_file) | `MY_PLUGIN_INBOUND_FOLLOW_TLS_CERT_FILE` | none | no | no |
| [`plugin_inbound_follow_tls_key_file`](#plugin_inbound_follow_tls_key_file) | `MY_PLUGIN_INBOUND_FOLLOW_TLS_KEY_FILE` | none | no | no |
| [`plugin_inbound_follow_timeout`](#plugin_inbound_follow_timeout) | `MY_PLUGIN_INBOUND_FOLLOW_TIMEOUT` | `3s` | no | no |
| [`plugin_signup_check_address`](#plugin_signup_check_address) | `MY_PLUGIN_SIGNUP_CHECK_ADDRESS` | none | no | no |
| [`plugin_signup_check_protocol`](#plugin_signup_check_protocol) | `MY_PLUGIN_SIGNUP_CHECK_PROTOCOL` | `json` | no | no |
| [`plugin_signup_check_token`](#plugin_signup_check_token) | `MY_PLUGIN_SIGNUP_CHECK_TOKEN` | none | no | yes |
| [`plugin_signup_check_tls_ca_file`](#plugin_signup_check_tls_ca_file) | `MY_PLUGIN_SIGNUP_CHECK_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_signup_check_tls_cert_file`](#plugin_signup_check_tls_cert_file) | `MY_PLUGIN_SIGNUP_CHECK_TLS_CERT_FILE` | none | no | no |
| [`plugin_signup_check_tls_key_file`](#plugin_signup_check_tls_key_file) | `MY_PLUGIN_SIGNUP_CHECK_TLS_KEY_FILE` | none | no | no |
| [`plugin_signup_check_timeout`](#plugin_signup_check_timeout) | `MY_PLUGIN_SIGNUP_CHECK_TIMEOUT` | `2s` | no | no |
| [`plugin_media_check_address`](#plugin_media_check_address) | `MY_PLUGIN_MEDIA_CHECK_ADDRESS` | none | no | no |
| [`plugin_media_check_protocol`](#plugin_media_check_protocol) | `MY_PLUGIN_MEDIA_CHECK_PROTOCOL` | `json` | no | no |
| [`plugin_media_check_token`](#plugin_media_check_token) | `MY_PLUGIN_MEDIA_CHECK_TOKEN` | none | no | yes |
| [`plugin_media_check_tls_ca_file`](#plugin_media_check_tls_ca_file) | `MY_PLUGIN_MEDIA_CHECK_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_media_check_tls_cert_file`](#plugin_media_check_tls_cert_file) | `MY_PLUGIN_MEDIA_CHECK_TLS_CERT_FILE` | none | no | no |
| [`plugin_media_check_tls_key_file`](#plugin_media_check_tls_key_file) | `MY_PLUGIN_MEDIA_CHECK_TLS_KEY_FILE` | none | no | no |
| [`plugin_media_check_timeout`](#plugin_media_check_timeout) | `MY_PLUGIN_MEDIA_CHECK_TIMEOUT` | `10s` | no | no |
| [`plugin_local_status_address`](#plugin_local_status_address) | `MY_PLUGIN_LOCAL_STATUS_ADDRESS` | none | no | no |
| [`plugin_local_status_protocol`](#plugin_local_status_protocol) | `MY_PLUGIN_LOCAL_STATUS_PROTOCOL` | `json` | no | no |
| [`plugin_local_status_token`](#plugin_local_status_token) | `MY_PLUGIN_LOCAL_STATUS_TOKEN` | none | no | yes |
| [`plugin_local_status_tls_ca_file`](#plugin_local_status_tls_ca_file) | `MY_PLUGIN_LOCAL_STATUS_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_local_status_tls_cert_file`](#plugin_local_status_tls_cert_file) | `MY_PLUGIN_LOCAL_STATUS_TLS_CERT_FILE` | none | no | no |
| [`plugin_local_status_tls_key_file`](#plugin_local_status_tls_key_file) | `MY_PLUGIN_LOCAL_STATUS_TLS_KEY_FILE` | none | no | no |
| [`plugin_local_status_timeout`](#plugin_local_status_timeout) | `MY_PLUGIN_LOCAL_STATUS_TIMEOUT` | `1s` | no | no |
| [`plugin_inbound_actor_address`](#plugin_inbound_actor_address) | `MY_PLUGIN_INBOUND_ACTOR_ADDRESS` | none | no | no |
| [`plugin_inbound_actor_protocol`](#plugin_inbound_actor_protocol) | `MY_PLUGIN_INBOUND_ACTOR_PROTOCOL` | `json` | no | no |
| [`plugin_inbound_actor_token`](#plugin_inbound_actor_token) | `MY_PLUGIN_INBOUND_ACTOR_TOKEN` | none | no | yes |
| [`plugin_inbound_actor_tls_ca_file`](#plugin_inbound_actor_tls_ca_file) | `MY_PLUGIN_INBOUND_ACTOR_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_inbound_actor_tls_cert_file`](#plugin_inbound_actor_tls_cert_file) | `MY_PLUGIN_INBOUND_ACTOR_TLS_CERT_FILE` | none | no | no |
| [`plugin_inbound_actor_tls_key_file`](#plugin_inbound_actor_tls_key_file) | `MY_PLUGIN_INBOUND_ACTOR_TLS_KEY_FILE` | none | no | no |
| [`plugin_inbound_actor_timeout`](#plugin_inbound_actor_timeout) | `MY_PLUGIN_INBOUND_ACTOR_TIMEOUT` | `3s` | no | no |
| [`plugin_local_profile_address`](#plugin_local_profile_address) | `MY_PLUGIN_LOCAL_PROFILE_ADDRESS` | none | no | no |
| [`plugin_local_profile_protocol`](#plugin_local_profile_protocol) | `MY_PLUGIN_LOCAL_PROFILE_PROTOCOL` | `json` | no | no |
| [`plugin_local_profile_token`](#plugin_local_profile_token) | `MY_PLUGIN_LOCAL_PROFILE_TOKEN` | none | no | yes |
| [`plugin_local_profile_tls_ca_file`](#plugin_local_profile_tls_ca_file) | `MY_PLUGIN_LOCAL_PROFILE_TLS_CA_FILE` | the system's roots | no | no |
| [`plugin_local_profile_tls_cert_file`](#plugin_local_profile_tls_cert_file) | `MY_PLUGIN_LOCAL_PROFILE_TLS_CERT_FILE` | none | no | no |
| [`plugin_local_profile_tls_key_file`](#plugin_local_profile_tls_key_file) | `MY_PLUGIN_LOCAL_PROFILE_TLS_KEY_FILE` | none | no | no |
| [`plugin_local_profile_timeout`](#plugin_local_profile_timeout) | `MY_PLUGIN_LOCAL_PROFILE_TIMEOUT` | `1s` | no | no |

## Settings

### `host`

Domain that serves the instance. Every ActivityPub `id` is under `https://{host}/`. A port is allowed for local development. Recorded in the database on first start and must not change afterwards: the server and the `my` commands refuse to run when the configuration differs from the recorded value.

- Environment variable: `MY_HOST`
- Default: none
- Required: yes

### `account_domain`

Domain of user handles (`@alice@{account_domain}`). Defaults to `host`. When it differs, `account_domain` must redirect `/.well-known/webfinger`, `/.well-known/host-meta`, and `/.well-known/nodeinfo` to `host`. Recorded in the database on first start and must not change afterwards: the server and the `my` commands refuse to run when the configuration differs from the recorded value.

- Environment variable: `MY_ACCOUNT_DOMAIN`
- Default: value of `host`
- Required: no

### `source_url`

Where users get the source code of the running server, as an `https` URL or a `mailto:` address to request it from. AGPL-3.0 section 13 requires offering the source to everyone who uses the server over the network; `GET /instance`, the information page (`/about`), the landing page, and NodeInfo (`metadata.sourceCode`, and `software.repository` for an `https` URL) publish it. At the default, a release's `GET /instance`, and so the information and landing pages, link to the source at the release's tag (`https://github.com/mycroblog/my/tree/v0.1.0`), and NodeInfo to the repository. Point it at your own source when you run a fork or a modified version ([running the server](https://mycroblog.dev/next/operations/running/)).

- Environment variable: `MY_SOURCE_URL`
- Default: `https://github.com/mycroblog/my`
- Required: no

### `status_page_url`

Address of the instance's status page, an absolute `https` URL on a service apart from the server. Empty means none. `GET /instance`, the sign-in screen, and the server-rendered error pages link to it, and so does the maintenance page, which is why it is a setting rather than an instance setting in the database: the link must show while the database is out of reach.

- Environment variable: `MY_STATUS_PAGE_URL`
- Default: none
- Required: no

### `maintenance`

Starts `my serve` in the maintenance mode, also set by `my serve -maintenance`. The process then opens no adapter and reads only the settings marked for it (the listen address, the host and origins, the proxies, `status_page_url`, and the `maintenance_*` settings), so that it starts while the database, the KV, or the queue is out of reach. Whatever the `role`, it serves HTTP only and answers every request with a `503` that says the server is under maintenance; `/healthz`, `/readyz`, `/health/live`, and `/health/ready` answer `200` with `{"status":"maintenance"}`. Switch it by redeploying ([maintenance](https://mycroblog.dev/next/operations/maintenance/)).

- Environment variable: `MY_MAINTENANCE`
- Default: `false`
- Required: no

| Value | Description |
| --- | --- |
| `false` | Normal operation. |
| `true` | Maintenance mode. |

### `maintenance_server_name`

Name of the server on the maintenance page and the exported maintenance pages (`my maintenance export`). The instance's own name is in the database, which the maintenance mode does not read.

- Environment variable: `MY_MAINTENANCE_SERVER_NAME`
- Default: value of `host`
- Required: no

### `maintenance_message`

Notice of the maintenance as plain text, at most 2,000 characters, shown on the maintenance page and in the `message` of the maintenance problem of the client API. The same text is shown in every locale.

- Environment variable: `MY_MAINTENANCE_MESSAGE`
- Default: none
- Required: no

### `maintenance_until`

When the maintenance is expected to end, as an RFC 3339 time such as `2026-10-24T03:00:00Z`. The maintenance page shows it in UTC, the maintenance problem has it as `until`, and while it is in the future `Retry-After` counts the seconds to it (at most 86,400).

- Environment variable: `MY_MAINTENANCE_UNTIL`
- Default: none
- Required: no

### `maintenance_retry_after`

The `Retry-After` of the maintenance responses when `maintenance_until` is unset or past, a Go duration from `1s` to `24h`.

- Environment variable: `MY_MAINTENANCE_RETRY_AFTER`
- Default: `10m`
- Required: no

### `web_origin`

Origin (`scheme://host[:port]`) of the web client. The client API allows only this origin by CORS. Set it when the web client is served from another origin, for example a development server at `http://localhost:5173`. An origin on `host` is where the server itself serves the web client, and the client served there calls the API at it: set `http://localhost:8080` (with `host` = `localhost:8080`) to run the server over plain HTTP on localhost, where the sign-in screens, the OAuth issuer, the client API, the links to the instance's pages in its responses, media URLs, and email and invite links are all on `http://localhost:8080`. ActivityPub ids stay under `https://{host}/` in every setup.

- Environment variable: `MY_WEB_ORIGIN`
- Default: `https://{host}`
- Required: no

### `web_early_hints`

Whether the web client's pages (the app and the sign-in screens) are preceded by a `103 Early Hints` response that names the stylesheets and scripts the page loads, so that the browser fetches them while the server prepares the page. The pages carry the same `Link` headers either way. Turn it on only when every proxy in front passes a 103 on: a Google Cloud external Application Load Balancer fails the request instead ([web client](https://mycroblog.dev/next/operations/running/#web-client)).

- Environment variable: `MY_WEB_EARLY_HINTS`
- Default: `false`
- Required: no

| Value | Description |
| --- | --- |
| `false` | No 103; the `Link` headers come with the page. |
| `true` | A 103 with the `Link` headers before each page. |

### `role`

What this process runs.

- Environment variable: `MY_ROLE`
- Default: `all`
- Required: no

| Value | Description |
| --- | --- |
| `all` | HTTP server and job worker in one process. |
| `web` | HTTP server only. |
| `worker` | Job worker only; the HTTP server answers only the health checks ([running the server](https://mycroblog.dev/next/operations/running/)). |

### `worker_job_types`

Comma-separated job types that this process runs, each a name such as `federation.deliver` or a prefix such as `federation.*`; the others wait for another worker. A name or prefix that matches no job type fails the start. Needs `queue = db` and `role = all` or `worker`, and excludes `worker_job_types_except`. Empty runs every type ([running the server](https://mycroblog.dev/next/operations/running/)).

- Environment variable: `MY_WORKER_JOB_TYPES`
- Default: none
- Required: no

### `worker_job_types_except`

Comma-separated job types that this process does not run, written like `worker_job_types`. Types that a later release adds are run. Needs `queue = db` and `role = all` or `worker`, and excludes `worker_job_types`. Empty runs every type ([running the server](https://mycroblog.dev/next/operations/running/)).

- Environment variable: `MY_WORKER_JOB_TYPES_EXCEPT`
- Default: none
- Required: no

### `worker_job_concurrency`

How many jobs this process runs at once, from 1 to 128. Raise it for jobs that mostly wait on other servers, usually together with `worker_job_types`. It must be below the pool maximum (`pool_max_conns` of `database_url`, by default the greater of 4 and the CPU count), or the server does not start. A value of 2 or more needs `queue = db` and `role = all` or `worker`.

- Environment variable: `MY_WORKER_JOB_CONCURRENCY`
- Default: `1`
- Required: no

### `listen`

TCP address the HTTP server listens on, as `host:port`. With `tls` other than `off`, the `web` and `all` roles serve HTTPS on it; the `worker` role always serves plain HTTP (its health checks).

- Environment variable: `MY_LISTEN`
- Default: `:443` with `tls` other than `off` in the `web` and `all` roles, `:8080` otherwise
- Required: no

### `tls`

Whether the server serves HTTPS itself, so that it runs without a reverse proxy ([serving HTTPS from the server](https://mycroblog.dev/next/operations/running/#serving-https-from-the-server)). Only the `web` and `all` roles listen with TLS.

- Environment variable: `MY_TLS`
- Default: `off`
- Required: no

| Value | Description |
| --- | --- |
| `off` | Plain HTTP on `listen`: a reverse proxy, a load balancer, or a CDN in front terminates TLS. |
| `acme` | HTTPS on `listen` with certificates that the server obtains itself from the ACME certificate authority `tls_acme_directory`, for `host` and, when it differs, the host of `web_origin`. It needs `tls_acme_agree_terms = true`, a `host` that is a DNS name with a dot, and a `kv` that every process shares (`kv = memory` only with `role = all`). The certificates and their keys are kept in the database, the keys sealed. The first certificate is ordered after the server starts listening, and handshakes for a host fail until it has one. The CA validates the order on `listen` (TLS-ALPN-01) or on `tls_http_listen` (HTTP-01), so both must be reachable from the internet on ports 443 and 80, or at least `listen` on port 443. |
| `files` | HTTPS on `listen` with the certificate `tls_cert_file` and its key `tls_key_file`, which another tool such as certbot keeps up to date. The server looks at their modification times every minute and loads them again when they change, without a restart. Port `tls_http_listen` redirects to `https`. |

### `tls_acme_directory`

Directory URL of the ACME certificate authority of `tls = acme`, an `https` URL. The default is Let's Encrypt; its staging environment, `https://acme-staging-v02.api.letsencrypt.org/directory`, issues certificates that browsers do not trust, under higher rate limits, for a first try ([serving HTTPS from the server](https://mycroblog.dev/next/operations/running/#serving-https-from-the-server)). The server reaches it directly and follows no redirect.

- Environment variable: `MY_TLS_ACME_DIRECTORY`
- Default: `https://acme-v02.api.letsencrypt.org/directory`
- Required: no

### `tls_acme_agree_terms`

Whether you agree to the terms of service of the certificate authority `tls_acme_directory`, which `tls = acme` requires. Read them first: the directory links them.

- Environment variable: `MY_TLS_ACME_AGREE_TERMS`
- Default: `false`
- Required: `tls = acme` (set to `true`)

| Value | Description |
| --- | --- |
| `false` | Not agreed: `tls = acme` is refused. |
| `true` | Agreed: the server registers its account with the certificate authority on its first order. |

### `tls_acme_email`

Email address that the account of `tls = acme` gives the certificate authority as its contact, such as `ops@example.com`. Empty gives none. Let's Encrypt sends no email about expiring certificates.

- Environment variable: `MY_TLS_ACME_EMAIL`
- Default: none
- Required: no

### `tls_acme_profile`

Profile of the certificates that `tls = acme` orders, among those the certificate authority offers (its directory lists them), such as `shortlived` at Let's Encrypt. Empty takes the authority's default.

- Environment variable: `MY_TLS_ACME_PROFILE`
- Default: none
- Required: no

### `tls_cert_file`

Path of the certificate chain in PEM, the server's certificate first and then its intermediates, for `tls = files`. It must cover `host`.

- Environment variable: `MY_TLS_CERT_FILE`
- Default: none
- Required: `tls = files`

### `tls_key_file`

Path of the private key of `tls_cert_file` in PEM, for `tls = files`.

- Environment variable: `MY_TLS_KEY_FILE`
- Default: none
- Required: `tls = files`

### `tls_http_listen`

TCP address of the plain HTTP port with `tls` other than `off`, as `host:port`. It sends `GET` and `HEAD` requests to the same path on `https` with `301` and answers `400` to every other method, so that no request body travels in plain text. Empty opens no such port.

- Environment variable: `MY_TLS_HTTP_LISTEN`
- Default: `:80`
- Required: no

### `trusted_proxies`

Reverse proxies and load balancers in front of the server whose forwarding header (`trusted_proxy_header`) tells the client IP, as comma-separated CIDRs or addresses such as `10.0.0.0/8, 2001:db8::/32, 192.0.2.10`. The client IP keys the rate limits of sign-in, registration, the client API, and the inbox. When the connection's peer is in this list, the client IP is the rightmost address of the header that is not in it; otherwise the header is ignored and the client IP is the peer. Empty trusts no one: behind a proxy every request then shares the proxy's limits. List only addresses that nothing but your proxies can connect from, and only proxies that append to the header rather than pass a client's value through.

- Environment variable: `MY_TRUSTED_PROXIES`
- Default: none
- Required: no

### `trusted_proxy_header`

Forwarding header that the `trusted_proxies` append the client IP to. Only this header is read, so that a client cannot add the other one past a proxy that does not know it.

- Environment variable: `MY_TRUSTED_PROXY_HEADER`
- Default: `x-forwarded-for`
- Required: no

| Value | Description |
| --- | --- |
| `x-forwarded-for` | `X-Forwarded-For`, which nginx, Caddy, Traefik, Google Cloud load balancers and Cloud Run, and AWS Application Load Balancers append to. |
| `forwarded` | The `for` parameters of `Forwarded` (RFC 7239). |

### `origin_secret`

Refuses the requests that did not come through your CDN: a request gets `403` unless it carries the header `origin_secret_header` with this value, which the CDN adds to every request it forwards. Only the health checks and the calls of the platform's queue pass without it. Empty checks nothing. A wrong setup refuses everyone while the health checks stay fine, so read [serving media through a CDN](https://mycroblog.dev/next/operations/cdn/) first. A random string of at least 32 bytes of visible ASCII without spaces, such as the output of `openssl rand -hex 32`. It does nothing for a platform feature that adds no header, such as a Cloud Run domain mapping.

- Environment variable: `MY_ORIGIN_SECRET` or `MY_ORIGIN_SECRET_FILE`
- Default: none
- Required: no

### `origin_secret_previous`

A second value that the check accepts while you rotate `origin_secret`, in the same format, since the CDN and the server cannot change at the same moment. Remove it once the CDN sends the new value ([serving media through a CDN](https://mycroblog.dev/next/operations/cdn/)).

- Environment variable: `MY_ORIGIN_SECRET_PREVIOUS` or `MY_ORIGIN_SECRET_PREVIOUS_FILE`
- Default: none
- Required: no

### `origin_secret_header`

Name of the request header that carries `origin_secret`, which the server removes before anything else reads the request. Letters, digits, and hyphens, and not a name that a CDN refuses to set or that the server or a proxy reads for itself, such as those starting `cf-`, `x-cf-`, `x-amz-`, or `x-forwarded-`, or `authorization`.

- Environment variable: `MY_ORIGIN_SECRET_HEADER`
- Default: none
- Required: when `origin_secret` is set

### `database`

Database adapter. The adapter is never inferred from `database_url`.

- Environment variable: `MY_DATABASE`
- Default: `sqlite`
- Required: no

| Value | Description |
| --- | --- |
| `sqlite` | Local SQLite file (`modernc.org/sqlite`). |
| `postgres` | PostgreSQL and compatible services (`pgx/v5`). |

### `database_url`

Connection string passed to the `database` adapter, in that adapter's format. For `database = sqlite`, a path or a `file:` URI; `pool_max_conns` sets how many connections may be open (by default the greater of 4 and the CPU count, as for PostgreSQL) and `pool_max_idle_conns` how many idle ones are kept (by default all of them; none is closed by time).

- Environment variable: `MY_DATABASE_URL` or `MY_DATABASE_URL_FILE`
- Default: none
- Required: yes

### `database_acquire_timeout`

How long a request or a job waits for a free connection of the database's pool before giving up, as a Go duration; `0` waits without a limit. It bounds only the wait, never the query or the transaction that follows. A request that gives up gets a `503` with `Retry-After`, and a job runs again later without using an attempt. Only for `database = postgres`: SQLite has no limit on connections to wait for, and setting it fails there. Each process has its own value, so processes of different roles can differ.

- Environment variable: `MY_DATABASE_ACQUIRE_TIMEOUT`
- Default: `60s`
- Required: no

### `database_dev_url`

Connection string of the dev database that `my schema apply` builds the desired schema in, for `database = postgres`, in the format of `database_url`. Each run creates a temporary schema there and drops it at the end. Set it when the role of `database_url` may not create a schema in its database; the role of this URL needs that privilege. The server itself never uses it.

- Environment variable: `MY_DATABASE_DEV_URL` or `MY_DATABASE_DEV_URL_FILE`
- Default: a temporary schema in the database of `database_url`
- Required: no

### `queue`

Job queue adapter.

- Environment variable: `MY_QUEUE`
- Default: `db`
- Required: no

| Value | Description |
| --- | --- |
| `db` | Queue table in the database. |
| `inline` | Runs jobs synchronously in the enqueuing process; for tests and development. Requires `role = all`. |
| `cloudtasks` | The queue table in the database (`database = postgres`), with Google Cloud Tasks carrying a wake-up signal for each job to `POST /_internal/jobs` (`queue_cloudtasks_` settings). Cloud Scheduler calls `POST /_internal/tick` every minute for the periodic jobs instead of a scheduler in the process, and no process polls the table. Requires `kv = db` or `valkey`. Application Default Credentials ([running the server](https://mycroblog.dev/next/operations/running/)). |

### `queue_cloudtasks_routes`

Comma-separated routes of `queue = cloudtasks`, each `type=queue resource name` with an optional `@receiver URL`, that send the wake-up signals of those job types to their own queue, and to a separate service when the URL is given (`queue_cloudtasks_url` otherwise). A type is a name such as `notification.email` or a prefix such as `federation.*`; a name wins over a prefix, and a longer prefix over a shorter one. The others go to `queue_cloudtasks_queue`. Give every process, `role = web` included, the same value: a process without it sends its signals to the default queue, which loses no job but undoes the split without an error. A type that matches no job type fails at startup.

- Environment variable: `MY_QUEUE_CLOUDTASKS_ROUTES`
- Default: none
- Required: no

### `queue_cloudtasks_queue`

Resource name of the Cloud Tasks queue of `queue = cloudtasks`, `projects/{project}/locations/{location}/queues/{queue}`, which the deployment template creates.

- Environment variable: `MY_QUEUE_CLOUDTASKS_QUEUE`
- Default: none
- Required: when `queue = cloudtasks`

### `queue_cloudtasks_url`

Absolute `https` URL of the server that Cloud Tasks and Cloud Scheduler call with `queue = cloudtasks`, without a path: the tasks go to `{url}/_internal/jobs`, and the Cloud Scheduler job must call `POST {url}/_internal/tick`. Set it to the URL of the Cloud Run service of the processes that run the jobs (`role = all` or `worker`).

- Environment variable: `MY_QUEUE_CLOUDTASKS_URL`
- Default: `https://{host}`
- Required: no

### `queue_cloudtasks_service_account`

Email of the service account whose OIDC ID token each task of `queue = cloudtasks` carries, and that the Cloud Scheduler job of the tick uses too. The server's own service account needs `roles/iam.serviceAccountUser` on it to create the tasks. `/_internal/jobs` and `/_internal/tick` accept only a Google-signed token of this account with a verified email.

- Environment variable: `MY_QUEUE_CLOUDTASKS_SERVICE_ACCOUNT`
- Default: none
- Required: when `queue = cloudtasks`

### `queue_cloudtasks_audience`

Audience of the OIDC ID tokens of `queue = cloudtasks`, which the tasks ask for and the endpoints require. Empty leaves the default of Cloud Tasks and Cloud Scheduler, the URL of the endpoint (`{queue_cloudtasks_url}/_internal/jobs`, `{queue_cloudtasks_url}/_internal/tick`); set the same value as the audience of the Cloud Scheduler job when you set it.

- Environment variable: `MY_QUEUE_CLOUDTASKS_AUDIENCE`
- Default: the URL of the endpoint called
- Required: no

### `queue_cloudtasks_dispatch_deadline`

How long Cloud Tasks waits for one delivery of `queue = cloudtasks` to answer, from `15s` to `30m`. It must be longer than the 5-minute lease of a job run; set the request timeout of the Cloud Run service to at least this value.

- Environment variable: `MY_QUEUE_CLOUDTASKS_DISPATCH_DEADLINE`
- Default: `10m`
- Required: no

### `blob`

Blob storage adapter for media.

- Environment variable: `MY_BLOB`
- Default: `fs`
- Required: no

| Value | Description |
| --- | --- |
| `fs` | Local file system. |
| `s3` | S3-compatible object storage (Amazon S3, Cloudflare R2, the XML API of Google Cloud Storage, MinIO, Garage): `blob_s3_bucket` and the other `blob_s3_` settings. |
| `gcs` | Google Cloud Storage through its JSON API, with Application Default Credentials (the service account of Cloud Run or of the instance), so that no key is kept: `blob_gcs_bucket`. |

### `media_image`

Processor of images: their variants, without metadata.

- Environment variable: `MY_MEDIA_IMAGE`
- Default: `native`
- Required: no

| Value | Description |
| --- | --- |
| `native` | In the server, in pure Go. |
| `plugin` | A plugin of yours (`plugin_media_image_address`) re-encodes images into their variants, so that heavy conversion runs on another machine; the server still checks uploads and every output itself, and removing metadata is left to the plugin ([plugins](https://mycroblog.dev/next/operations/plugins/)). |

### `media_video`

Processor of videos: writing their streams again without metadata.

- Environment variable: `MY_MEDIA_VIDEO`
- Default: `native`
- Required: no

| Value | Description |
| --- | --- |
| `native` | In the server, in pure Go. |
| `plugin` | A plugin of yours (`plugin_media_video_address`) remuxes videos, so that heavy conversion runs on another machine; the server still checks uploads and every output itself, and removing metadata is left to the plugin ([plugins](https://mycroblog.dev/next/operations/plugins/)). |

### `media_audio`

Processor of audio-only files: writing their streams again without metadata.

- Environment variable: `MY_MEDIA_AUDIO`
- Default: `native`
- Required: no

| Value | Description |
| --- | --- |
| `native` | In the server, in pure Go. |
| `plugin` | A plugin of yours (`plugin_media_audio_address`) remuxes audio files, so that heavy conversion runs on another machine; the server still checks uploads and every output itself, and removing metadata is left to the plugin ([plugins](https://mycroblog.dev/next/operations/plugins/)). |

### `blob_fs_dir`

Directory of the `fs` blob store, relative to the working directory or absolute. It holds the uploads as received (private) and the processed media, which the server serves under `/media/`. Share it between the web and worker processes, and back it up with the database.

- Environment variable: `MY_BLOB_FS_DIR`
- Default: `blobs`
- Required: no

### `blob_s3_bucket`

Bucket of the `s3` blob store. It holds the uploads as received (private) and the processed media, which the server reads from the bucket and serves under `/media/`, so the bucket stays private. Share it between the web and worker processes, and back it up with the database ([backup and restore](https://mycroblog.dev/next/operations/backup/)).

- Environment variable: `MY_BLOB_S3_BUCKET`
- Default: none
- Required: when `blob = s3`

### `blob_s3_endpoint`

Origin of the S3-compatible API of `blob = s3`, such as `https://<account>.r2.cloudflarestorage.com` (Cloudflare R2), `https://storage.googleapis.com` (Google Cloud Storage with an HMAC key), or `http://minio:9000`. Empty uses Amazon S3's endpoint of `blob_s3_region`. Use `https` unless the server is on a private network that you trust: the requests carry the blobs.

- Environment variable: `MY_BLOB_S3_ENDPOINT`
- Default: none
- Required: no

### `blob_s3_region`

Region of the bucket of `blob = s3`, such as `eu-west-1`; `auto` for Cloudflare R2 and Google Cloud Storage. Empty takes the region of the AWS SDK's default chain (`AWS_REGION`), and with `blob_s3_endpoint` set and no region there, `us-east-1`, which MinIO and most S3-compatible servers expect.

- Environment variable: `MY_BLOB_S3_REGION`
- Default: none
- Required: no

### `blob_s3_path_style`

How `blob = s3` addresses the bucket.

- Environment variable: `MY_BLOB_S3_PATH_STYLE`
- Default: `false`
- Required: no

| Value | Description |
| --- | --- |
| `false` | In the host name (`https://<bucket>.<endpoint>/<key>`), as Amazon S3, R2, and Cloud Storage take it. |
| `true` | In the path (`https://<endpoint>/<bucket>/<key>`), for MinIO and other servers without a host name per bucket. |

### `blob_s3_access_key_id`

Access key ID of `blob = s3`, with `blob_s3_secret_access_key`. Empty takes the credentials of the AWS SDK's default chain: the environment (`AWS_ACCESS_KEY_ID`), the shared files, or the role of the instance or container, which needs no key in the configuration on AWS.

- Environment variable: `MY_BLOB_S3_ACCESS_KEY_ID`
- Default: none
- Required: no

### `blob_s3_secret_access_key`

Secret access key of `blob_s3_access_key_id`.

- Environment variable: `MY_BLOB_S3_SECRET_ACCESS_KEY` or `MY_BLOB_S3_SECRET_ACCESS_KEY_FILE`
- Default: none
- Required: when `blob_s3_access_key_id` is set

### `blob_gcs_bucket`

Bucket of the `gcs` blob store. It holds the uploads as received (private) and the processed media, which the server reads from the bucket and serves under `/media/`, so the bucket stays private. The credentials are Application Default Credentials: the service account of Cloud Run or of the instance, or the key file of `GOOGLE_APPLICATION_CREDENTIALS`; the service account needs `roles/storage.objectUser` on the bucket. Signed URLs are signed with the private key of a key file, or without one with the IAM `signBlob` method as the service account itself, which needs `roles/iam.serviceAccountTokenCreator` on that service account. Share it between the web and worker processes, and back it up with the database ([backup and restore](https://mycroblog.dev/next/operations/backup/)).

- Environment variable: `MY_BLOB_GCS_BUCKET`
- Default: none
- Required: when `blob = gcs`

### `media_origin`

Origin (`scheme://host[:port]`, no path) at which the media files are served, for a CDN in front of the server's `/media/`. It is the origin of the URLs of the local media files (`/media/...`) and of the remote media proxy (`/media/proxy/...`) in the client API, ActivityPub and link previews, and the web client's Content-Security-Policy allows images and media from it. `https` is required; `http` is accepted only for a loopback host (`localhost`, `127.0.0.1`, `[::1]`). Unset, the media is served from the server's own origin. The URLs are built when they are read, so a change applies to the media that already exists. ActivityPub ids never use it ([serving media through a CDN](https://mycroblog.dev/next/operations/cdn/)).

- Environment variable: `MY_MEDIA_ORIGIN`
- Default: the server's own origin (`https://{host}`)
- Required: no

### `media_max_image_bytes`

Largest accepted image upload in bytes.

- Environment variable: `MY_MEDIA_MAX_IMAGE_BYTES`
- Default: `16777216`
- Required: no

### `media_max_image_side`

Largest accepted image width and height in pixels, read from the file's header before decoding.

- Environment variable: `MY_MEDIA_MAX_IMAGE_SIDE`
- Default: `8192`
- Required: no

### `media_max_image_pixels`

Largest accepted image area (width × height, and for an animated GIF the frames together), read from the file's header before decoding. The default is 33 MP (7680 × 4320).

- Environment variable: `MY_MEDIA_MAX_IMAGE_PIXELS`
- Default: `33177600`
- Required: no

### `media_unattached_retention`

How long an upload may stay unattached to any post or profile before the hourly cleanup deletes it, as a Go duration such as `24h`.

- Environment variable: `MY_MEDIA_UNATTACHED_RETENTION`
- Default: `24h`
- Required: no

### `media_replaced_retention`

How long the files of an image's replaced version stay after a newer version replaces them, as a Go duration such as `720h`. Pages that are open, caches, and other servers still hold their URLs, which answer `404` once the hourly cleanup deletes them.

- Environment variable: `MY_MEDIA_REPLACED_RETENTION`
- Default: `720h`
- Required: no

### `media_max_av_bytes`

Largest accepted video or audio upload in bytes. The file is held in memory while it is checked and remuxed, so this also bounds the memory of one upload.

- Environment variable: `MY_MEDIA_MAX_AV_BYTES`
- Default: `104857600`
- Required: no

### `media_max_video_duration`

Longest accepted video upload, as a Go duration such as `5m`.

- Environment variable: `MY_MEDIA_MAX_VIDEO_DURATION`
- Default: `5m`
- Required: no

### `media_max_audio_duration`

Longest accepted audio upload, as a Go duration such as `10m`.

- Environment variable: `MY_MEDIA_MAX_AUDIO_DURATION`
- Default: `10m`
- Required: no

### `custom_emoji_max_bytes`

Largest image in bytes that a local custom emoji may be registered with; the image must also be a PNG, a GIF, or a WebP of at most 1,024 pixels on a side. The default is Mastodon's limit. GoToSocial does not fetch remote emoji larger than 100 KiB by default.

- Environment variable: `MY_CUSTOM_EMOJI_MAX_BYTES`
- Default: `262144`
- Required: no

### `remote_custom_emoji_max_bytes`

Largest image in bytes of a remote custom emoji that the media proxy serves; the proxy answers 404 for a larger one, and for one that is not a PNG, a GIF, or a WebP.

- Environment variable: `MY_REMOTE_CUSTOM_EMOJI_MAX_BYTES`
- Default: `1048576`
- Required: no

### `kv`

Key-value store adapter for caches and rate limits.

- Environment variable: `MY_KV`
- Default: `db`
- Required: no

| Value | Description |
| --- | --- |
| `db` | Table in the database. |
| `memory` | In-process memory; limits are per process. |
| `valkey` | Valkey or another server that speaks the Redis protocol (`valkey_url`); entries are shared by every process on the server. |

### `valkey_url`

Valkey server of the adapters that use one (`kv = valkey`, `stream = valkey`):`valkey://[user:password@]host[:port][/db]` (`redis://` is the same), `valkeys://` or `rediss://` for TLS, or `unix:///path/to/socket`. The path selects the database number. A cluster is found by itself.

- Environment variable: `MY_VALKEY_URL` or `MY_VALKEY_URL_FILE`
- Default: none
- Required: when `kv = valkey` or `stream = valkey`

### `valkey_tls_ca_file`

PEM file of the certificate authorities that the TLS certificate of `valkey_url` chains to, for a managed service with a private CA. Empty uses the system roots; it needs a `valkeys://` or `rediss://` URL.

- Environment variable: `MY_VALKEY_TLS_CA_FILE`
- Default: none
- Required: no

### `valkey_key_prefix`

Prefix of every key that this server writes to `valkey_url`, so that several instances or other applications can share the server.

- Environment variable: `MY_VALKEY_KEY_PREFIX`
- Default: `my:`
- Required: no

### `stream`

Broker that carries the signals of the real-time updates (`GET /api/v1/streaming`) between processes. It is never inferred from the other settings. With `hub`, a stream receives only the changes made in its own process, so run one process with `role = all`; a process with another role warns at start. The other brokers reach every process of the instance, and a process connects to them only while it has open streams.

- Environment variable: `MY_STREAM`
- Default: `hub`
- Required: no

| Value | Description |
| --- | --- |
| `hub` | In-process only; for one process with `role = all` (SQLite, or one PostgreSQL server process). |
| `postgres` | PostgreSQL `LISTEN` / `NOTIFY` on the database of `database_url` (or `stream_postgres_url`). Requires `database = postgres`. |
| `valkey` | Valkey Pub/Sub on `valkey_url`, with the channel `{valkey_key_prefix}stream`. |
| `sns` | An Amazon SNS standard topic (`stream_sns_topic_arn`) that fans out to one SQS queue per listening process (`stream_sns_queue_prefix`); credentials and settings from the AWS SDK's default chain. |
| `pubsub` | A Google Cloud Pub/Sub topic (`stream_pubsub_topic`) with one subscription per listening process (`stream_pubsub_subscription_prefix`); Application Default Credentials, or the emulator at `PUBSUB_EMULATOR_HOST`. |

### `stream_postgres_url`

Connection string of the `LISTEN` connections of `stream = postgres`, in the format of `database_url`. Set it when `database_url` goes through a pooler in transaction mode (PgBouncer's transaction pooling, Neon's pooled connection string, Supabase's port 6543), where `LISTEN` does not work: it must reach the same database directly or through a pooler in session mode.

- Environment variable: `MY_STREAM_POSTGRES_URL` or `MY_STREAM_POSTGRES_URL_FILE`
- Default: value of `database_url`
- Required: no

### `stream_sns_topic_arn`

ARN of the Amazon SNS standard topic of `stream = sns`, which the deployment template creates. The queues are created in its account and region.

- Environment variable: `MY_STREAM_SNS_TOPIC_ARN`
- Default: none
- Required: when `stream = sns`

### `stream_sns_queue_prefix`

Start of the names of the per-process SQS queues of `stream = sns`, `{prefix}-{process id}-{n}`: at most 40 letters, digits, hyphens, and underscores. The `stream.sweep` job deletes the queues under it that ended processes left behind, so it belongs to one instance.

- Environment variable: `MY_STREAM_SNS_QUEUE_PREFIX`
- Default: none
- Required: when `stream = sns`

### `stream_pubsub_topic`

Full name of the Google Cloud Pub/Sub topic of `stream = pubsub`, `projects/{project}/topics/{topic}`, which the deployment template creates. The subscriptions are created in its project.

- Environment variable: `MY_STREAM_PUBSUB_TOPIC`
- Default: none
- Required: when `stream = pubsub`

### `stream_pubsub_subscription_prefix`

Start of the IDs of the per-process subscriptions of `stream = pubsub`, `{prefix}-{process id}-{n}`: a letter, then letters, digits, and `-_.~+%`, not starting with `goog`. The `stream.sweep` job deletes the subscriptions under it that ended processes left behind, so it belongs to one instance.

- Environment variable: `MY_STREAM_PUBSUB_SUBSCRIPTION_PREFIX`
- Default: none
- Required: when `stream = pubsub`

### `key_manager`

Key manager adapter that encrypts actor private keys.

- Environment variable: `MY_KEY_MANAGER`
- Default: `local`
- Required: no

| Value | Description |
| --- | --- |
| `local` | Uses the key-encryption key from `kek`. With `plugin_key_manager_address` set, the plugin only decrypts the secrets it still holds, while they move to `kek`. |
| `gcpkms` | Google Cloud KMS: the symmetric CryptoKey of `kms_key` encrypts the data-encryption keys, with Application Default Credentials ([key management](https://mycroblog.dev/next/operations/key-management/)). |
| `awskms` | AWS KMS: the symmetric key of `kms_key` encrypts the data-encryption keys, with the credentials and region of the AWS SDK's default chain ([key management](https://mycroblog.dev/next/operations/key-management/)). |
| `plugin` | A plugin of yours at `plugin_key_manager_address`, such as one before Vault or an HSM. It sees the data-encryption keys, so it is in the same trust boundary as the key-encryption key. The server does not start without its answer or when it fails a self-test, and nothing falls back to another key ([key management](https://mycroblog.dev/next/operations/key-management/)). |

### `kek`

Key-encryption key for actor private keys and the other secrets in the database: 32 random bytes in standard base64 (for example `openssl rand -base64 32`). Required when `key_manager = local`; with another `key_manager` it only decrypts, while the secrets move to the KMS or plugin. Back it up separately from the database; losing it loses every actor key ([key management](https://mycroblog.dev/next/operations/key-management/)).

- Environment variable: `MY_KEK` or `MY_KEK_FILE`
- Default: none
- Required: when `key_manager = local`

### `kek_previous`

Previous key-encryption keys, comma-separated, each in the format of `kek`. They only decrypt; new secrets are always encrypted with `kek`, or with the KMS or plugin when `key_manager` is not `local`. Set the replaced key here while rotating `kek` ([key management](https://mycroblog.dev/next/operations/key-management/)).

- Environment variable: `MY_KEK_PREVIOUS` or `MY_KEK_PREVIOUS_FILE`
- Default: none
- Required: no

### `kms_key`

Key of the KMS that encrypts the secrets when `key_manager` is `gcpkms` or `awskms`: the resource name of a Cloud KMS CryptoKey (`projects/{project}/locations/{location}/keyRings/{ring}/cryptoKeys/{key}`, without a version) or the ARN, an alias (`alias/...`), or the key ID of an AWS KMS key, which the server resolves to the key's ARN at start. At most 248 bytes. With `key_manager = local`, it only decrypts, like `kms_key_previous`.

- Environment variable: `MY_KMS_KEY`
- Default: none
- Required: when `key_manager` is `gcpkms` or `awskms`

### `kms_key_previous`

Previous keys of the same KMS, comma-separated, each in the format of `kms_key`. They only decrypt; set the replaced key here while moving to another `kms_key`. With `key_manager = local`, `kms_key` and these keys only decrypt the secrets still encrypted with the KMS, while they move to `kek` ([key management](https://mycroblog.dev/next/operations/key-management/)).

- Environment variable: `MY_KMS_KEY_PREVIOUS`
- Default: none
- Required: no

### `kms_dek_cache_ttl`

How long the server keeps a data-encryption key that the KMS or plugin decrypted, as a Go duration, so that a secret in use reaches it about once per this period and process. `0` disables the cache. Not used when `key_manager = local`.

- Environment variable: `MY_KMS_DEK_CACHE_TTL`
- Default: `10m`
- Required: no

### `kms_dek_cache_entries`

Most data-encryption keys that the cache of `kms_dek_cache_ttl` keeps, the least recently used going first. `0` disables the cache.

- Environment variable: `MY_KMS_DEK_CACHE_ENTRIES`
- Default: `1024`
- Required: no

### `mail`

Mailer adapter for verification emails. With `none`, no email reaches users, so an unverified address does not keep a user out and sign-ups in the `open` registration mode wait for a moderator's approval instead.

- Environment variable: `MY_MAIL`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | Sends nothing and logs each message's category; for instances without email. With `profile = development` it logs the whole message, including its links. |
| `smtp` | SMTP submission with mandatory STARTTLS (`mail_smtp_host`, `mail_smtp_port`, `mail_smtp_username`, `mail_smtp_password`). |
| `ses` | The Amazon SES v2 API (`mail_ses_region`, `mail_ses_configuration_set`); credentials from the AWS SDK's default chain, such as the execution role, so no SMTP password. |
| `plugin` | A plugin of yours (`plugin_mailer_address`), for a mail service that offers only an HTTP API. It gets the parts of each message and adds `Message-ID`, `Date`, and `Auto-Submitted` itself ([sending email](https://mycroblog.dev/next/operations/email/)). |
| `fake` | Test builds only: a stand-in mailer for end-to-end tests and demonstrations that sends nothing and writes each message whole, with its links, to the server log, while the server treats it as a mailer that delivers: it asks users to confirm their addresses and mails password reset links. Only a binary built with the build tag `e2e_fakes` has it, and that binary refuses it unless `profile = development`; released binaries and images never have it ([contributing guide](https://github.com/mycroblog/my/blob/main/CONTRIBUTING.md#build-tags)). It needs `mail_from` alone. |

### `mail_from`

Sender address of emails, as a bare address such as `noreply@social.example`. Emails name the instance as the sender (`From: "<instance name>" <mail_from>`), take their `Message-ID` domain from it, and send replies to the instance setting `contact_email` when it is set.

- Environment variable: `MY_MAIL_FROM`
- Default: none
- Required: when `mail` is not `none`

### `mail_smtp_host`

Host name of the SMTP server. The server must offer STARTTLS with a certificate valid for this name; the connection is refused otherwise.

- Environment variable: `MY_MAIL_SMTP_HOST`
- Default: none
- Required: when `mail = smtp`

### `mail_smtp_port`

TCP port of the SMTP server (submission).

- Environment variable: `MY_MAIL_SMTP_PORT`
- Default: `587`
- Required: no

### `mail_smtp_username`

User name for SMTP AUTH PLAIN, sent only after STARTTLS. Leave it empty for a server that needs no authentication.

- Environment variable: `MY_MAIL_SMTP_USERNAME`
- Default: none
- Required: no

### `mail_smtp_password`

Password for SMTP AUTH PLAIN.

- Environment variable: `MY_MAIL_SMTP_PASSWORD` or `MY_MAIL_SMTP_PASSWORD_FILE`
- Default: none
- Required: no

### `mail_ses_region`

AWS region of the SES endpoint of `mail = ses`, such as `eu-west-1`: the region where the identity of `mail_from` is verified. Empty takes the region of the AWS SDK's default chain (`AWS_REGION`).

- Environment variable: `MY_MAIL_SES_REGION`
- Default: none
- Required: no

### `mail_ses_configuration_set`

Name of the SES configuration set that `mail = ses` applies to every message, for its event destinations such as the SNS topic of the bounces and complaints ([sending email](https://mycroblog.dev/next/operations/email/)). Empty sends without one.

- Environment variable: `MY_MAIL_SES_CONFIGURATION_SET`
- Default: none
- Required: no

### `mail_ses_sns_topic_arn`

ARN of the SNS topic that SES publishes the bounces and complaints to, `arn:aws:sns:<region>:<account>:<topic>`, with the topic attribute `SignatureVersion` set to `2` ([sending email](https://mycroblog.dev/next/operations/email/)). Setting it opens `POST /mail/events/ses` for the topic's HTTPS subscription, which records the permanent bounces and the complaints as suppressions; empty answers `404` there.

- Environment variable: `MY_MAIL_SES_SNS_TOPIC_ARN`
- Default: none
- Required: no

### `captcha`

CAPTCHA service that the sign-in and registration forms check. The service receives the visitor's IP address, browser details, and cookies, and the two pages load its script; say so in the instance's privacy policy. Only `none` keeps every request on this server.

- Environment variable: `MY_CAPTCHA`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | No CAPTCHA; nothing is sent to a third party. |
| `turnstile` | Cloudflare Turnstile. |
| `hcaptcha` | hCaptcha. |
| `recaptcha` | Google reCAPTCHA v2 with the checkbox (not v3 or Enterprise). |

### `captcha_site_key`

Public site key of the `captcha` service, which the widget on the forms carries.

- Environment variable: `MY_CAPTCHA_SITE_KEY`
- Default: none
- Required: when `captcha` is not `none`

### `captcha_secret_key`

Secret key of the `captcha` service, which the server sends with each token to the service's siteverify API. The server does not try the key at start: a key that the service refuses is logged as an error on the first check, and until it is fixed registrations are refused and sign-ins skip the CAPTCHA.

- Environment variable: `MY_CAPTCHA_SECRET_KEY` or `MY_CAPTCHA_SECRET_KEY_FILE`
- Default: none
- Required: when `captcha` is not `none`

### `vapid_private_key`

VAPID private key of the server's Web Push notifications (RFC 8292): a P-256 private key, 32 bytes in base64url. Generate one with `my config vapid-key`. Without it, the server sends no Web Push and clients do not offer it. Changing it ends every subscription, since browsers subscribe with its public key ([key management](https://mycroblog.dev/next/operations/key-management/)).

- Environment variable: `MY_VAPID_PRIVATE_KEY` or `MY_VAPID_PRIVATE_KEY_FILE`
- Default: none
- Required: no

### `profile`

Whether this is a production instance. Settings that weaken a security default for development, such as `http_allow_private_networks`, are refused in the `production` profile.

- Environment variable: `MY_PROFILE`
- Default: `production`
- Required: no

| Value | Description |
| --- | --- |
| `production` | A production instance. |
| `development` | A development or test instance. |

### `http_allow_private_networks`

Lets outbound HTTP (federation, media, link previews) reach loopback, private, and other non-public addresses and host names, over plain `http` and on any port. For development and tests against local servers only: the server logs a warning at start, and the `production` profile refuses it.

- Environment variable: `MY_HTTP_ALLOW_PRIVATE_NETWORKS`
- Default: `false`
- Required: no

| Value | Description |
| --- | --- |
| `false` | Only public `https` destinations on ports 443 and 80. |
| `true` | Every destination; requires `profile = development`. |

### `webhook_allowed_destinations`

Destinations on loopback or a private network that the instance's webhooks may reach, such as an automation tool on the same host or in the same VPC: comma-separated host names, IP addresses, and CIDRs such as `n8n.internal, 10.0.0.0/8`. The SSRF guard of outbound HTTP refuses them otherwise; for these destinations alone, and for the webhooks alone, it also allows plain `http` and any port. A host name is allowed whatever it resolves to; an address range allows the URLs that name an address in it and the host names that resolve into it only when they are listed too. Only the operator can change it: the administrators of the web client cannot reach an internal network on their own. Empty allows none.

- Environment variable: `MY_WEBHOOK_ALLOWED_DESTINATIONS`
- Default: none
- Required: no

### `login_provider_allowed_destinations`

Destinations on loopback or a private network that the external sign-in providers may be on, such as an OpenID Connect provider in the same VPC: comma-separated host names, IP addresses, and CIDRs such as `keycloak.internal, 10.0.0.0/8`. The SSRF guard of outbound HTTP refuses them otherwise; for these destinations alone, and for the requests to sign-in providers alone, it also allows plain `http` and any port. A host name is allowed whatever it resolves to; an address range allows the URLs that name an address in it and the host names that resolve into it only when they are listed too. Only the operator can change it: the administrators of the web client cannot reach an internal network on their own. Empty allows none.

- Environment variable: `MY_LOGIN_PROVIDER_ALLOWED_DESTINATIONS`
- Default: none
- Required: no

### `push_allowed_destinations`

For development only: destinations on loopback or a private network that the Web Push sends may reach, such as a fake push service on the same host: comma-separated host names, IP addresses, and CIDRs such as `localhost, 127.0.0.0/8`. The SSRF guard of outbound HTTP refuses them otherwise; for these destinations alone, and for the Web Push sends alone, it also allows any port, still over `https` and without following a redirect. A host name is allowed whatever it resolves to; an address range allows the URLs that name an address in it and the host names that resolve into it only when they are listed too. Every other outbound request keeps the guard. The `production` profile refuses it, and the server logs a warning at start when it is set. Empty allows none.

- Environment variable: `MY_PUSH_ALLOWED_DESTINATIONS`
- Default: none
- Required: no

### `relay_max_activities_per_hour`

How many activities each relay may bring in per hour. The ones over it are dropped before their posts are fetched, and counted in `my.relay.activities` and on the relays screen. It bounds the fetches, rows, and jobs that a relay costs.

- Environment variable: `MY_RELAY_MAX_ACTIVITIES_PER_HOUR`
- Default: `1000`
- Required: no

### `billing_sources`

Cloud providers whose usage charges are imported into the instance's costs every 24 hours, comma-separated: `aws` (AWS Cost Explorer, about $0.01 per billed request), `gcp` (the Cloud Billing export to BigQuery, or the Pub/Sub notifications of a budget: `billing_gcp_method`), `cloudflare` (the usage charges of a pay-as-you-go Cloudflare account by product, without plans, taxes, or credits: `billing_cloudflare_account_id`), and `plugin` (a plugin of yours that aggregates any other providers: `plugin_billing_address`). Each import records the charges of the current billing month and of the previous one (through the 15th, or until it is final) by service and charge type, as estimates until the provider closes the month, then final, then corrections; `GET /api/v1/admin/billing_imports` shows the state of the imports. A provider whose adapter is not in this build is refused at start. Credentials come from the SDK's default chain of AWS and Google Cloud, and from `billing_cloudflare_api_token` for Cloudflare ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)). Empty imports nothing.

- Environment variable: `MY_BILLING_SOURCES`
- Default: none
- Required: no

### `billing_aws_max_requests_per_day`

How many billed AWS Cost Explorer requests (one per page, $0.01 each) the imports of `billing_sources = aws` may make per UTC day, counting the scheduled imports, their retries, and the imports that an administrator starts. An import with none left calls nothing and is recorded as `request_cap`; the next day's import tries again.

- Environment variable: `MY_BILLING_AWS_MAX_REQUESTS_PER_DAY`
- Default: `4`
- Required: no

### `billing_aws_max_requests_per_month`

How many billed AWS Cost Explorer requests the imports may make per UTC calendar month, the bound of the cost of the imports (the default is $1.20 a month: four imports a day for thirty days). An import with none left calls nothing and is recorded as `request_cap` until the next month.

- Environment variable: `MY_BILLING_AWS_MAX_REQUESTS_PER_MONTH`
- Default: `120`
- Required: no

### `billing_aws_metric`

Cost Explorer metric that the imports of `billing_sources = aws` record.

- Environment variable: `MY_BILLING_AWS_METRIC`
- Default: `NetUnblendedCost`
- Required: no

| Value | Description |
| --- | --- |
| `NetUnblendedCost` | The cost after the discounts (Savings Plans, Reserved Instances, private pricing), the closest to the invoice. |
| `UnblendedCost` | The cost before the discounts. |

### `billing_aws_filter`

Which costs of the AWS account the imports of `billing_sources = aws` take. The most accurate is an AWS account that serves this instance alone, with `none`: the taxes and the support fees carry no tag and may fall out of a tag filter ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)).

- Environment variable: `MY_BILLING_AWS_FILTER`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | Every cost that the credentials can see: the account's own, or, from the management account of an organization, every member account's. |
| `linked_account` | The costs of the member accounts of `billing_aws_linked_accounts` (the `LINKED_ACCOUNT` dimension). |
| `tag` | The costs of the resources that carry the cost allocation tag `billing_aws_tag`, which must be activated in the Billing console first. |

### `billing_aws_tag`

The activated cost allocation tag of `billing_aws_filter = tag`, `key=value` such as `app=my`. A tag shows in the billing data from its activation on, up to 24 hours after it is activated; older costs need a backfill in the Billing console ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)).

- Environment variable: `MY_BILLING_AWS_TAG`
- Default: none
- Required: when `billing_aws_filter = tag`

### `billing_aws_linked_accounts`

The 12-digit IDs of the member accounts of `billing_aws_filter = linked_account`, comma-separated. Reading them from a member account needs Linked Account Access, which the management account turns on in the Cost Explorer preferences.

- Environment variable: `MY_BILLING_AWS_LINKED_ACCOUNTS`
- Default: none
- Required: when `billing_aws_filter = linked_account`

### `billing_aws_role_arn`

ARN of an IAM role that the imports of `billing_sources = aws` assume with the Google ID token of the instance's service account (`AssumeRoleWithWebIdentity`, audience: this ARN), for a server on Google Cloud (Cloud Run) without AWS keys. The role trusts the identity provider `accounts.google.com` for the service account ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)). Empty uses the AWS SDK's default chain (the execution role, `AWS_ACCESS_KEY_ID`, the shared files).

- Environment variable: `MY_BILLING_AWS_ROLE_ARN`
- Default: none
- Required: no

### `billing_gcp_method`

How the imports of `billing_sources = gcp` read the Google Cloud usage charges. The billing months are the invoice months of Cloud Billing, US Pacific calendar months; a month is final from the 5th of the next month on (Pacific time).

- Environment variable: `MY_BILLING_GCP_METHOD`
- Default: `bigquery`
- Required: no

| Value | Description |
| --- | --- |
| `bigquery` | Sum the standard usage cost export of Cloud Billing to BigQuery (`billing_gcp_table`) with one query a day, within the free tier of BigQuery ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)). |
| `budget` | Read the notifications of a monthly Cloud Billing budget from a Pub/Sub pull subscription (`billing_gcp_budget_subscription`) once a day: the budget's estimated total of each month, one line named after the budget, without the services, the charge types, or the export's days ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)). A simple way for an operator without the BigQuery export; free within the free tier of Pub/Sub. |

### `billing_gcp_table`

The table of the standard usage cost export of Cloud Billing to BigQuery, `project.dataset.table` such as `my-project.billing_export.gcp_billing_export_v1_012345_6789AB_CDEF01`. The service account of the server needs `roles/bigquery.dataViewer` on its dataset ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)).

- Environment variable: `MY_BILLING_GCP_TABLE`
- Default: none
- Required: when `billing_sources` includes `gcp` and `billing_gcp_method = bigquery`

### `billing_gcp_query_project`

The Google Cloud project that runs the BigQuery queries of `billing_gcp_method = bigquery` and is billed for their bytes; the service account of the server needs `roles/bigquery.jobUser` on it. Empty uses the project of `billing_gcp_table`.

- Environment variable: `MY_BILLING_GCP_QUERY_PROJECT`
- Default: none
- Required: no

### `billing_gcp_projects`

Google Cloud project IDs, comma-separated, whose costs the imports of `billing_gcp_method = bigquery` take (`project.id` of the export). Empty takes every cost of the billing account. The taxes and the other charges of the invoice that carry no project may fall out of this filter ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)).

- Environment variable: `MY_BILLING_GCP_PROJECTS`
- Default: none
- Required: no

### `billing_gcp_label`

A label `key=value`, such as `app=my`, whose resources' costs the imports of `billing_gcp_method = bigquery` take (`labels` of the export). Empty takes the costs with or without labels. The taxes and the costs of resources without the label fall out of this filter.

- Environment variable: `MY_BILLING_GCP_LABEL`
- Default: none
- Required: no

### `billing_gcp_max_bytes_billed`

The `maximumBytesBilled` of each BigQuery query of `billing_gcp_method = bigquery`, in bytes: a query that would bill more fails as `bytes_cap` and bills nothing. The default, 1 GiB, keeps a query a day well within the free tier of 1 TiB a month, which the billing account shares with its other uses; BigQuery bills at least 10 MB a query.

- Environment variable: `MY_BILLING_GCP_MAX_BYTES_BILLED`
- Default: `1073741824`
- Required: no

### `billing_gcp_budget_subscription`

The pull subscription of the Pub/Sub topic of a monthly Cloud Billing budget that `billing_gcp_method = budget` reads, `projects/{project}/subscriptions/{subscription}`. The imports acknowledge a notification once a later one of its budget and month replaces it or once its month is final and no longer imported, and keep the latest of each month imported, so the subscription needs a message retention of at least 14 days (31 recommended) and must not be read by anything else. The service account of the server needs `roles/pubsub.subscriber` on it ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)).

- Environment variable: `MY_BILLING_GCP_BUDGET_SUBSCRIPTION`
- Default: none
- Required: when `billing_sources` includes `gcp` and `billing_gcp_method = budget`

### `billing_cloudflare_account_id`

The Cloudflare account whose usage charges `billing_sources = cloudflare` imports: its account ID, 32 hexadecimal digits, on the account's overview in the dashboard. Only a pay-as-you-go account has them; an Enterprise contract account fails as `not_enabled` ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)).

- Environment variable: `MY_BILLING_CLOUDFLARE_ACCOUNT_ID`
- Default: none
- Required: when `billing_sources` includes `cloudflare`

### `billing_cloudflare_api_token`

An API token of the account of `billing_cloudflare_account_id` that the Cloudflare imports read with. Give it the account permission Billing Read and nothing else: the imports only read. Use a token of its own rather than one that also edits DNS or Workers; the server never reads `CLOUDFLARE_API_TOKEN` ([importing cloud costs](https://mycroblog.dev/next/operations/billing-import/)).

- Environment variable: `MY_BILLING_CLOUDFLARE_API_TOKEN` or `MY_BILLING_CLOUDFLARE_API_TOKEN_FILE`
- Default: none
- Required: when `billing_sources` includes `cloudflare`

### `payments`

The payment provider of the supporters: local users subscribe to paid plans on its hosted checkout (PAY.JP: the server's own checkout page), and the roles of the plans follow the state of their payments. Only one provider is active on a server; switching to another one needs the subscribers to subscribe again. The provider calls `POST /payments/webhooks/{provider}` (signed or carrying the account's token, at most 1 MiB, recorded once per event), and a reconciliation reads its lists every `payments_reconcile_interval`. Card details never reach the server. An administrator still enables the supporters and sets up the plans on the admin screen.

- Environment variable: `MY_PAYMENTS`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | No payments: the webhook endpoint and the reconciliation do not run. |
| `stripe` | Stripe Billing: Checkout in subscription mode and the customer portal, with `payments_stripe_secret_key` and `payments_stripe_webhook_secret` ([payments with Stripe](https://mycroblog.dev/next/operations/payments-stripe/)). |
| `payjp` | PAY.JP on its API v1: plans and subscriptions of PAY.JP, with the server's own checkout and card pages that load payjp.js, with `payments_payjp_secret_key`, `payments_payjp_public_key`, and `payments_payjp_webhook_token` ([payments with PAY.JP](https://mycroblog.dev/next/operations/payments-payjp/)). JPY only. |
| `fake` | Test builds only: a stand-in provider for end-to-end tests and demonstrations that keeps its side in the server's memory and serves its own checkout page, where paying records the subscription as paid without any money moving. Only a binary built with the build tag `e2e_fakes` has it, and that binary refuses it unless `profile = development`; released binaries and images never have it ([contributing guide](https://github.com/mycroblog/my/blob/main/CONTRIBUTING.md#build-tags)). It needs no other setting. |

### `payments_stripe_secret_key`

The API key of the Stripe account of `payments = stripe`: a restricted key (`rk_test_` / `rk_live_`) with the permissions of [payments with Stripe](https://mycroblog.dev/next/operations/payments-stripe/), or a secret key (`sk_test_` / `sk_live_`). Its prefix sets the mode: a test key charges nothing, and the webhook endpoint refuses live events with it (and test events with a live key).

- Environment variable: `MY_PAYMENTS_STRIPE_SECRET_KEY` or `MY_PAYMENTS_STRIPE_SECRET_KEY_FILE`
- Default: none
- Required: when `payments = stripe`

### `payments_stripe_webhook_secret`

The signing secret (`whsec_`) of the webhook endpoint `https://<domain>/payments/webhooks/stripe` that the operator registers on Stripe's dashboard, or of `stripe listen` for local tests. While a secret is rolled, Stripe signs with the old and the new one, so one is enough.

- Environment variable: `MY_PAYMENTS_STRIPE_WEBHOOK_SECRET` or `MY_PAYMENTS_STRIPE_WEBHOOK_SECRET_FILE`
- Default: none
- Required: when `payments = stripe`

### `payments_payjp_secret_key`

The secret key (`sk_test_` / `sk_live_`) of the PAY.JP account of `payments = payjp`. Its prefix sets the mode: a test key charges nothing, and the webhook endpoint refuses live events with it (and test events with a live key). The server also signs the URLs of its checkout and card pages with a key derived from it, so a new secret key ends the pages open at the time.

- Environment variable: `MY_PAYMENTS_PAYJP_SECRET_KEY` or `MY_PAYMENTS_PAYJP_SECRET_KEY_FILE`
- Default: none
- Required: when `payments = payjp`

### `payments_payjp_public_key`

The public key (`pk_test_` / `pk_live_`) of the PAY.JP account of `payments = payjp`, of the mode of `payments_payjp_secret_key`, which the server's checkout and card pages give payjp.js.

- Environment variable: `MY_PAYMENTS_PAYJP_PUBLIC_KEY`
- Default: none
- Required: when `payments = payjp`

### `payments_payjp_webhook_token`

The webhook token (`whook_`) of the PAY.JP account in the mode of `payments_payjp_secret_key`, which PAY.JP sends in `X-Payjp-Webhook-Token` to `https://<domain>/payments/webhooks/payjp`. Test and live mode have different tokens.

- Environment variable: `MY_PAYMENTS_PAYJP_WEBHOOK_TOKEN` or `MY_PAYMENTS_PAYJP_WEBHOOK_TOKEN_FILE`
- Default: none
- Required: when `payments = payjp`

### `payments_reconcile_interval`

How often the reconciliation with `payments` runs: it reads the provider's changes since the previous run, the checkouts in progress, the money movements, and a sweep of every subscription, and corrects what the webhooks missed. A Go duration from `15m` to `168h`; raise it, or `payments_reconcile_max_requests`, when the runs often stop at the cap.

- Environment variable: `MY_PAYMENTS_RECONCILE_INTERVAL`
- Default: `6h`
- Required: no

### `payments_reconcile_max_requests`

How many requests one reconciliation sends to `payments`, a page of a list counting as one. A run that reaches it records where it stopped and a continuation picks up the rest a minute later. The provider's adapter also keeps the requests within its rate limits.

- Environment variable: `MY_PAYMENTS_RECONCILE_MAX_REQUESTS`
- Default: `1000`
- Required: no

### `translator_libretranslate_url`

The base URL of the LibreTranslate server of the translation provider `libretranslate`, such as `https://libretranslate.com` or a server you run on a private address, such as `http://libretranslate:5000`; `http:` is accepted on any host. Setting it registers the provider, which the instance setting `translation_providers` then enables.

- Environment variable: `MY_TRANSLATOR_LIBRETRANSLATE_URL`
- Default: none
- Required: no

### `translator_libretranslate_api_key`

The API key of the LibreTranslate server of `translator_libretranslate_url`, when the server requires one, as `libretranslate.com` does. The server sends it with each translation and never shows it to clients.

- Environment variable: `MY_TRANSLATOR_LIBRETRANSLATE_API_KEY` or `MY_TRANSLATOR_LIBRETRANSLATE_API_KEY_FILE`
- Default: none
- Required: no

### `translator_deepl_api_key`

The authentication key of DeepL's API, which makes DeepL a translation provider that `translation_providers` can enable as `deepl` ([translation](https://mycroblog.dev/next/operations/translation/#deepl)). A key of the free plan ends in `:fx`; DeepL may keep the text that the free plan translates.

- Environment variable: `MY_TRANSLATOR_DEEPL_API_KEY` or `MY_TRANSLATOR_DEEPL_API_KEY_FILE`
- Default: none
- Required: no

### `translator_deepl_url`

The `https:` URL of DeepL's API that the server calls, such as a proxy in front of it; `http:` is accepted for localhost only.

- Environment variable: `MY_TRANSLATOR_DEEPL_URL`
- Default: `https://api-free.deepl.com` for a key of the free plan, `https://api.deepl.com` otherwise
- Required: no

### `translator_google_project`

The Google Cloud project that the translations of the provider `google`, Google Cloud Translation, are billed to; the Cloud Translation API must be enabled in it, and the service account of the server needs `roles/cloudtranslate.user` on it. Setting it registers the provider, which the instance setting `translation_providers` enables; credentials come from Application Default Credentials ([translation providers](https://mycroblog.dev/next/operations/translation/#google-cloud-translation)). Empty registers no such provider.

- Environment variable: `MY_TRANSLATOR_GOOGLE_PROJECT`
- Default: none
- Required: no

### `translator_google_location`

The location of the requests of the provider `google`: `global`, or a location of Cloud Translation such as `us-central1` (`projects/{project}/locations/{location}`). Not every location serves the general translation model; see Google's list of the locations of Cloud Translation.

- Environment variable: `MY_TRANSLATOR_GOOGLE_LOCATION`
- Default: `global`
- Required: no

### `translator_aws_region`

The AWS region of Amazon Translate, such as `us-east-1`. Setting it registers Amazon Translate as the translation provider `aws`, which the instance setting `translation_providers` enables ([translation providers](https://mycroblog.dev/next/operations/translation/#amazon-translate)); the credentials come from the AWS SDK's default chain, such as the execution role. Empty registers no such provider.

- Environment variable: `MY_TRANSLATOR_AWS_REGION`
- Default: none
- Required: no

### `translator_fake`

Test builds only: registers the stand-in translation provider `fake`, which the instance setting `translation_providers` then enables, for end-to-end tests and demonstrations. It calls no service: it puts the target language before each text, as `[ja] Hello`, and fails as a provider that is down when a text contains `[translation-fails]`.

- Environment variable: `MY_TRANSLATOR_FAKE`
- Default: `false`
- Required: no

| Value | Description |
| --- | --- |
| `false` | No such provider. |
| `true` | The provider `fake`. Only a binary built with the build tag `e2e_fakes` has it, and that binary refuses it unless `profile = development`; released binaries and images never have it ([contributing guide](https://github.com/mycroblog/my/blob/main/CONTRIBUTING.md#build-tags)). |

### `search`

The search engine of the full-text search of statuses. The search of accounts and hashtags uses the database and does not depend on this setting.

- Environment variable: `MY_SEARCH`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | No full-text search: the server connects to no engine and enqueues no search job, and `GET /api/v1/instance` tells clients that the search is not available. |
| `fake` | Test builds only: a stand-in index in the server's memory, for end-to-end tests and demonstrations. Only a binary built with the build tag `e2e_fakes` has it, and that binary refuses it unless `profile = development`; released binaries and images never have it ([contributing guide](https://github.com/mycroblog/my/blob/main/CONTRIBUTING.md#build-tags)). It needs no other setting. |
| `elasticsearch` | An Elasticsearch cluster, version 8 or later, at `search_elasticsearch_url`, with `search_elasticsearch_api_key` or `search_elasticsearch_username` and `search_elasticsearch_password` when it has security on. The server creates and upgrades the index itself ([search engines](https://mycroblog.dev/next/operations/search-engines/#elasticsearch)). |
| `opensearch` | An OpenSearch cluster, or a domain of Amazon OpenSearch Service, at `search_opensearch_url` ([search engines](https://mycroblog.dev/next/operations/search-engines/#opensearch)). The server creates its index at start. |
| `meilisearch` | A Meilisearch server (self-hosted or Meilisearch Cloud) at `search_meilisearch_url`, with `search_meilisearch_api_key`. The server creates the index `search_meilisearch_index` and its settings. |
| `plugin` | A plugin of yours (`plugin_search_address`), for an engine that the server has no built-in support for, such as Groonga or an index built with Sudachi. The plugin receives the text, author, visibility, and language of the posts that the index holds and the queries of signed-in people, and returns post IDs only ([plugins](https://mycroblog.dev/next/operations/plugins/#searching-posts), [search protocol](https://mycroblog.dev/next/plugins/search/)). |

### `search_elasticsearch_url`

The URL of the REST API of the Elasticsearch cluster of `search = elasticsearch`, such as `https://search.internal.example:9200`, without credentials, query, or fragment; a path is kept, for a proxy that serves the API under one. Use `https:` whenever the server sends credentials over a network that others share.

- Environment variable: `MY_SEARCH_ELASTICSEARCH_URL`
- Default: none
- Required: when `search = elasticsearch`

### `search_elasticsearch_api_key`

The API key of `search = elasticsearch`, the `encoded` value that Elasticsearch returns when it creates the key, sent as `Authorization: ApiKey <key>`. Set it or `search_elasticsearch_username` and `search_elasticsearch_password`, not both; with neither, the server sends no credentials.

- Environment variable: `MY_SEARCH_ELASTICSEARCH_API_KEY` or `MY_SEARCH_ELASTICSEARCH_API_KEY_FILE`
- Default: none
- Required: no

### `search_elasticsearch_username`

The user of `search = elasticsearch` for basic authentication, with `search_elasticsearch_password`, when `search_elasticsearch_api_key` is not set.

- Environment variable: `MY_SEARCH_ELASTICSEARCH_USERNAME`
- Default: none
- Required: no

### `search_elasticsearch_password`

The password of `search_elasticsearch_username`.

- Environment variable: `MY_SEARCH_ELASTICSEARCH_PASSWORD` or `MY_SEARCH_ELASTICSEARCH_PASSWORD_FILE`
- Default: none
- Required: when `search_elasticsearch_username` is set

### `search_elasticsearch_index`

The name of the alias that the index of `search = elasticsearch` lives behind; the server names the index behind it from this name with a suffix, such as `my-statuses-v1`. Up to 200 lowercase letters, digits, `-`, `_`, or `.`, starting with a letter or digit. Give each server that shares a cluster its own name.

- Environment variable: `MY_SEARCH_ELASTICSEARCH_INDEX`
- Default: `my-statuses`
- Required: no

### `search_elasticsearch_index_settings_file`

JSON file of the settings of each index that the server creates for `search = elasticsearch`, one object such as `{"number_of_shards": 1, "number_of_replicas": 0, "analysis": {...}}`: the analysis, the shards, the replicas, and any other index setting. It must define the analyzer `my_text`, which analyzes the posts, and may define `my_text_search`, which then analyzes the words of the queries. Empty uses the built-in settings, which need no plugin. The server reads it at start; after a change, it creates a new empty index and `my search rebuild` must fill it ([search engines](https://mycroblog.dev/next/operations/search-engines/#tuning-the-analysis-of-elasticsearch)).

- Environment variable: `MY_SEARCH_ELASTICSEARCH_INDEX_SETTINGS_FILE`
- Default: none
- Required: no

### `search_elasticsearch_ca_file`

PEM file of the certificate authorities that the TLS certificate of `search_elasticsearch_url` chains to, such as the `http_ca.crt` that Elasticsearch generates when it starts with security on. Empty uses the system roots; it needs an `https:` URL.

- Environment variable: `MY_SEARCH_ELASTICSEARCH_CA_FILE`
- Default: none
- Required: no

### `search_opensearch_url`

The URL of the REST API of the OpenSearch cluster of `search = opensearch`, such as `https://search.example.org:9200`, with no credentials, query, or fragment; a path is kept, for a cluster behind a reverse proxy.

- Environment variable: `MY_SEARCH_OPENSEARCH_URL`
- Default: none
- Required: when `search = opensearch`

### `search_opensearch_index`

The name of the alias of the index of `search = opensearch`. The server keeps the documents in the index named after it with the version of its mapping, such as `my-statuses-v1`, and points the alias at it. Lowercase letters, digits, `-`, and `_`, starting with a letter or a digit.

- Environment variable: `MY_SEARCH_OPENSEARCH_INDEX`
- Default: `my-statuses`
- Required: no

### `search_opensearch_index_settings_file`

The path of a JSON file whose object the server uses as the `settings` of every index of `search = opensearch` that it creates: the analysis, the shards, the replicas, and any other index setting. It must define the analyzers `my_text`, of the text of the posts, and `my_text_search`, of the queries ([tuning the analysis](https://mycroblog.dev/next/operations/search-engines/#tuning-the-analysis-of-opensearch)). Empty uses the built-in settings. A change of the file takes effect at the next start, in a new empty index, so it needs a rebuild of the index.

- Environment variable: `MY_SEARCH_OPENSEARCH_INDEX_SETTINGS_FILE`
- Default: none
- Required: no

### `search_opensearch_username`

The user of the HTTP Basic authentication of `search = opensearch`, such as a user of the OpenSearch security plugin or the master user of an Amazon OpenSearch Service domain with fine-grained access control. Empty sends no credentials. It excludes `search_opensearch_aws_signing`.

- Environment variable: `MY_SEARCH_OPENSEARCH_USERNAME`
- Default: none
- Required: no

### `search_opensearch_password`

The password of `search_opensearch_username`.

- Environment variable: `MY_SEARCH_OPENSEARCH_PASSWORD` or `MY_SEARCH_OPENSEARCH_PASSWORD_FILE`
- Default: none
- Required: when `search_opensearch_username` is set

### `search_opensearch_aws_signing`

Whether the server signs the requests of `search = opensearch` with AWS Signature Version 4, as a domain of Amazon OpenSearch Service with an IAM access policy needs. The credentials come from the AWS SDK's default chain: the environment, the shared files, the execution role.

- Environment variable: `MY_SEARCH_OPENSEARCH_AWS_SIGNING`
- Default: `off`
- Required: no

| Value | Description |
| --- | --- |
| `off` | No signing. |
| `es` | Sign for a domain of Amazon OpenSearch Service (the service name `es`) in `search_opensearch_aws_region`. OpenSearch Serverless (`aoss`) is not supported. |

### `search_opensearch_aws_region`

The AWS region of the domain of `search_opensearch_aws_signing = es`, such as `us-east-1`.

- Environment variable: `MY_SEARCH_OPENSEARCH_AWS_REGION`
- Default: the region of the AWS SDK's default chain (`AWS_REGION`)
- Required: no

### `search_meilisearch_url`

The base URL of the Meilisearch server of `search = meilisearch`, such as `https://meilisearch.example.org` or `http://meilisearch:7700` on a private network. It holds no credentials: the key goes in `search_meilisearch_api_key`.

- Environment variable: `MY_SEARCH_MEILISEARCH_URL`
- Default: none
- Required: when `search = meilisearch`

### `search_meilisearch_api_key`

The master key of the Meilisearch server, or an API key with the actions `search`, `documents.add`, `documents.get`, `documents.delete`, `indexes.create`, `indexes.get`, `indexes.delete`, `settings.update`, and `tasks.get` on `search_meilisearch_index`. Empty sends no key, for a server without a master key.

- Environment variable: `MY_SEARCH_MEILISEARCH_API_KEY` or `MY_SEARCH_MEILISEARCH_API_KEY_FILE`
- Default: none
- Required: no

### `search_meilisearch_index`

The uid of the Meilisearch index of the posts, which the server creates when it does not exist: letters, digits, hyphens, and underscores. Servers that share a Meilisearch server need different ones.

- Environment variable: `MY_SEARCH_MEILISEARCH_INDEX`
- Default: `my_statuses`
- Required: no

### `search_meilisearch_settings_file`

The path of a JSON object of Meilisearch index settings, to tune the engine: `rankingRules` (which must start with `sort`), `typoTolerance`, `stopWords`, `synonyms`, `dictionary`, `separatorTokens`, `nonSeparatorTokens`, `localizedAttributes`, `proximityPrecision`, `pagination`, and the like. The server sets `searchableAttributes`, `displayedAttributes`, `filterableAttributes`, and `sortableAttributes` itself and refuses a file that sets them. A setting the file leaves out takes the server's default. The server reads the file at start and sends the settings when they differ from what the index has. Empty uses the server's defaults.

- Environment variable: `MY_SEARCH_MEILISEARCH_SETTINGS_FILE`
- Default: none
- Required: no

### `voice`

How the voice chat rooms connect their participants. With `off`, the voice chat API answers 404, the web client shows no voice chat, and the other `voice_*` settings are not checked at start.

- Environment variable: `MY_VOICE`
- Default: `off`
- Required: no

| Value | Description |
| --- | --- |
| `off` | No voice chat. |
| `p2p` | The browsers connect to each other, with the STUN servers of `voice_stun_urls`; the server only relays what they send to set up the connections. The server gives the browsers the TURN servers of `voice_turn` (coturn or Cloudflare) with short-lived credentials. A room holds at most `voice_p2p_max_participants` people. Without STUN and TURN, which the server warns about at start, only browsers on the same network connect. |
| `livekit` | A LiveKit server (self-hosted or LiveKit Cloud) carries the audio, with `voice_livekit_url`, `voice_livekit_api_key`, and `voice_livekit_api_secret`. Its own STUN and TURN servers serve the browsers. LiveKit sends its events to `https://<domain>/voice/webhooks/livekit`. |
| `sora` | A Sora server (self-hosted or Sora Cloud) carries the audio, with `voice_sora_mode`, `voice_sora_signaling_urls`, `voice_sora_join_secret`, and the settings of the mode. Its own STUN and TURN servers serve the browsers. Sora calls `https://<domain>/voice/webhooks/sora/auth` (its auth webhook) before every connection and sends its event and session webhooks to `https://<domain>/voice/webhooks/sora/events`. The Sora transport is tested only against a stand-in that follows Sora's documents, not against a real Sora server. |

### `voice_p2p_max_participants`

How many people a room of `voice = p2p` holds, from 2 to 50; one more is refused. Every speaker sends their audio to every other participant, so a larger room needs more upload bandwidth of each speaker.

- Environment variable: `MY_VOICE_P2P_MAX_PARTICIPANTS`
- Default: `10`
- Required: no

### `voice_sfu_max_participants`

How many people a room of `voice = livekit` or `voice = sora` holds; one more is refused. `0` sets no limit.

- Environment variable: `MY_VOICE_SFU_MAX_PARTICIPANTS`
- Default: `10`
- Required: no

### `voice_max_room_duration`

The longest a room stays open, a Go duration such as `4h`; a room that reaches it is ended. `0` sets no limit.

- Environment variable: `MY_VOICE_MAX_ROOM_DURATION`
- Default: `4h`
- Required: no

### `voice_max_open_rooms`

How many rooms may be open on the server at the same time; opening one more is refused. `0` sets no limit.

- Environment variable: `MY_VOICE_MAX_OPEN_ROOMS`
- Default: `10`
- Required: no

### `voice_stun_urls`

The STUN servers with which the browsers of `voice = p2p` find their public addresses, comma-separated `stun:` or `stuns:` URLs such as `stun:stun.example.org:3478`. Empty uses none. `voice = livekit` and `voice = sora` do not use it and warn at start when it is set.

- Environment variable: `MY_VOICE_STUN_URLS`
- Default: none
- Required: no

### `voice_turn`

The TURN server that relays the audio of `voice = p2p` between browsers that cannot reach each other directly, with short-lived credentials for each participant. The server checks the setting at start and gives the browsers the TURN servers of coturn or Cloudflare. `voice = livekit` and `voice = sora` do not use it and warn at start when it is not `none`.

- Environment variable: `MY_VOICE_TURN`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | No TURN server: browsers behind strict NATs or firewalls may not connect. |
| `coturn` | A coturn server at `voice_turn_coturn_urls`, with its `static-auth-secret` in `voice_turn_coturn_secret`. |
| `cloudflare` | Cloudflare's TURN service, with `voice_turn_cloudflare_key_id` and `voice_turn_cloudflare_api_token`; the server asks it for the credentials of each participant. |

### `voice_turn_coturn_urls`

The URLs of the coturn server of `voice_turn = coturn`, comma-separated `turn:` or `turns:` URLs such as `turn:turn.example.org:3478?transport=udp, turns:turn.example.org:5349`.

- Environment variable: `MY_VOICE_TURN_COTURN_URLS`
- Default: none
- Required: when `voice = p2p` and `voice_turn = coturn`

### `voice_turn_coturn_secret`

The `static-auth-secret` of the coturn server of `voice_turn = coturn`, which signs the short-lived credentials that the server gives the participants.

- Environment variable: `MY_VOICE_TURN_COTURN_SECRET` or `MY_VOICE_TURN_COTURN_SECRET_FILE`
- Default: none
- Required: when `voice = p2p` and `voice_turn = coturn`

### `voice_turn_cloudflare_key_id`

The ID of the TURN key of `voice_turn = cloudflare`, from Cloudflare's dashboard.

- Environment variable: `MY_VOICE_TURN_CLOUDFLARE_KEY_ID`
- Default: none
- Required: when `voice = p2p` and `voice_turn = cloudflare`

### `voice_turn_cloudflare_api_token`

The API token of the TURN key of `voice_turn = cloudflare`, with which the server asks Cloudflare for the credentials of each participant.

- Environment variable: `MY_VOICE_TURN_CLOUDFLARE_API_TOKEN` or `MY_VOICE_TURN_CLOUDFLARE_API_TOKEN_FILE`
- Default: none
- Required: when `voice = p2p` and `voice_turn = cloudflare`

### `voice_turn_credential_ttl`

How long the TURN credentials of a participant stay valid, a Go duration; the web client asks for new ones before they expire. With `voice_turn = cloudflare` it may be at most `48h`.

- Environment variable: `MY_VOICE_TURN_CREDENTIAL_TTL`
- Default: `6h`
- Required: no

### `voice_ice_transport_policy`

Which connections the browsers of `voice = p2p` make. `voice = livekit` and `voice = sora` do not use it and warn at start when it is `relay`.

- Environment variable: `MY_VOICE_ICE_TRANSPORT_POLICY`
- Default: `all`
- Required: no

| Value | Description |
| --- | --- |
| `all` | Direct connections where possible and the TURN server otherwise; the participants of a room may learn each other's IP addresses. |
| `relay` | Only through the TURN server, which hides the participants' IP addresses from each other; needs `voice_turn`. |

### `voice_livekit_url`

The `wss:` URL of the LiveKit server of `voice = livekit` that the browsers connect to, such as `wss://livekit.example.org`; `ws:` is accepted for localhost only.

- Environment variable: `MY_VOICE_LIVEKIT_URL`
- Default: none
- Required: when `voice = livekit`

### `voice_livekit_api_url`

The `https:` URL of the API of the LiveKit server of `voice = livekit` that the server calls, when it differs from `voice_livekit_url`; `http:` is accepted for localhost only.

- Environment variable: `MY_VOICE_LIVEKIT_API_URL`
- Default: `voice_livekit_url` with `https:` (`http:` for `ws:`)
- Required: no

### `voice_livekit_api_key`

The API key of the LiveKit server of `voice = livekit`.

- Environment variable: `MY_VOICE_LIVEKIT_API_KEY`
- Default: none
- Required: when `voice = livekit`

### `voice_livekit_api_secret`

The API secret of `voice_livekit_api_key`, which signs the access tokens of the participants and verifies the webhooks of the LiveKit server.

- Environment variable: `MY_VOICE_LIVEKIT_API_SECRET` or `MY_VOICE_LIVEKIT_API_SECRET_FILE`
- Default: none
- Required: when `voice = livekit`

### `voice_sora_mode`

Where the Sora server of `voice = sora` runs.

- Environment variable: `MY_VOICE_SORA_MODE`
- Default: none
- Required: when `voice = sora`

| Value | Description |
| --- | --- |
| `self_hosted` | A Sora server of the operator: its API behind a reverse proxy with authentication at `voice_sora_api_url`, and its webhooks with Basic authentication (`voice_sora_webhook_username`, `voice_sora_webhook_password`). |
| `cloud` | Sora Cloud, with `voice_sora_cloud_project_id` and `voice_sora_cloud_api_key`. |

### `voice_sora_signaling_urls`

The signaling URLs of the Sora server of `voice = sora` that the browsers connect to, comma-separated `wss:` URLs; `ws:` is accepted for localhost only.

- Environment variable: `MY_VOICE_SORA_SIGNALING_URLS`
- Default: none
- Required: when `voice = sora`

### `voice_sora_api_url`

The `https:` URL of the reverse proxy in front of the API of the self-hosted Sora server, which the server calls; the API of Sora itself has no authentication and must not be reachable from outside. `http:` is accepted for localhost only.

- Environment variable: `MY_VOICE_SORA_API_URL`
- Default: none
- Required: when `voice = sora` and `voice_sora_mode = self_hosted`

### `voice_sora_api_header_name`

The name of the header with which the reverse proxy of `voice_sora_api_url` authenticates the server, such as `Authorization`; set it together with `voice_sora_api_header_value`. Empty sends none.

- Environment variable: `MY_VOICE_SORA_API_HEADER_NAME`
- Default: none
- Required: no

### `voice_sora_api_header_value`

The value of the header `voice_sora_api_header_name`, such as `Basic dXNlcjpwYXNz`.

- Environment variable: `MY_VOICE_SORA_API_HEADER_VALUE` or `MY_VOICE_SORA_API_HEADER_VALUE_FILE`
- Default: none
- Required: when `voice = sora` and `voice_sora_api_header_name` is set

### `voice_sora_webhook_username`

The Basic authentication user name that the self-hosted Sora server sends with its webhooks.

- Environment variable: `MY_VOICE_SORA_WEBHOOK_USERNAME`
- Default: none
- Required: when `voice = sora` and `voice_sora_mode = self_hosted`

### `voice_sora_webhook_password`

The Basic authentication password of `voice_sora_webhook_username`.

- Environment variable: `MY_VOICE_SORA_WEBHOOK_PASSWORD` or `MY_VOICE_SORA_WEBHOOK_PASSWORD_FILE`
- Default: none
- Required: when `voice = sora` and `voice_sora_mode = self_hosted`

### `voice_sora_cloud_project_id`

The project ID of Sora Cloud of `voice_sora_mode = cloud`.

- Environment variable: `MY_VOICE_SORA_CLOUD_PROJECT_ID`
- Default: none
- Required: when `voice = sora` and `voice_sora_mode = cloud`

### `voice_sora_cloud_api_key`

The API key of the Sora Cloud project, which signs the access tokens of the participants, authenticates the calls of its API, and verifies the signatures of its webhooks.

- Environment variable: `MY_VOICE_SORA_CLOUD_API_KEY` or `MY_VOICE_SORA_CLOUD_API_KEY_FILE`
- Default: none
- Required: when `voice = sora` and `voice_sora_mode = cloud`

### `voice_sora_join_secret`

The secret that signs the values with which the participants of `voice = sora` join a room, which the Sora server sends back for the server to authorize each connection. A random string of at least 32 bytes, such as the output of `openssl rand -base64 32`.

- Environment variable: `MY_VOICE_SORA_JOIN_SECRET` or `MY_VOICE_SORA_JOIN_SECRET_FILE`
- Default: none
- Required: when `voice = sora`

### `dead_job_retention`

How long a dead job (a job that failed for good or used up its attempts) is kept after it failed, for an operator to inspect and requeue it; the hourly purge then deletes it. A Go duration such as `720h` (30 days).

- Environment variable: `MY_DEAD_JOB_RETENTION`
- Default: `720h`
- Required: no

### `job_attempt_retention`

How long the record of a job's run that did not succeed (a retry, or the run that made it a dead job) is kept after the run, for the queue dashboard of the admin screen; the hourly purge then deletes it. Successful runs are not recorded. A Go duration such as `720h` (30 days).

- Environment variable: `MY_JOB_ATTEMPT_RETENTION`
- Default: `720h`
- Required: no

### `dynamic_client_unredeemed_retention`

How long a dynamically registered OAuth client (RFC 7591, `POST /oauth/register`) may wait for the exchange of its first authorization code: the hourly clean-up deletes a client that is older than this and had no code exchanged. From 30 days on, `dynamic_client_idle_retention` decides instead, because exchanged codes are purged by then. A Go duration such as `24h`.

- Environment variable: `MY_DYNAMIC_CLIENT_UNREDEEMED_RETENTION`
- Default: `24h`
- Required: no

### `dynamic_client_idle_retention`

How long a dynamically registered OAuth client without a usable token may go unused before the hourly clean-up deletes it. Until the last use of a token is recorded, the time the client's newest token was issued stands for it. A Go duration such as `2160h` (90 days).

- Environment variable: `MY_DYNAMIC_CLIENT_IDLE_RETENTION`
- Default: `2160h`
- Required: no

### `metrics_exporter`

Where the service metrics go: the request, federation, delivery, job, and stream metrics, named after the OpenTelemetry semantic conventions or with the prefix `my.`. Labels never carry a remote domain, an account, a URL, or a client IP.

- Environment variable: `MY_METRICS_EXPORTER`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | No metrics exporter; measuring costs next to nothing. The job `metrics.snapshot` still writes the instance-wide gauges every 15 minutes as one `metrics snapshot` log line. On Cloud Run the platform's metrics and log-based metrics made from the log lines stand in for them. |
| `prometheus` | The Prometheus text format at `GET /metrics` on `metrics_listen`, a listener of its own, for a Prometheus-compatible server (and KEDA) to scrape; for self-hosted instances. |
| `otlp` | OTLP over HTTP (protobuf) to an OpenTelemetry Collector, configured by the standard environment variables `OTEL_EXPORTER_OTLP_ENDPOINT` (default `http://localhost:4318`), `OTEL_EXPORTER_OTLP_HEADERS`, and `OTEL_METRIC_EXPORT_INTERVAL` (default 60 seconds). |
| `emf` | CloudWatch Embedded Metric Format lines, which CloudWatch Logs turns into metrics; for AWS (ECS). The log line that ends each request and job carries its duration metrics, and the other measurements are written on standard output, one line per label set, at the end of each request and job. Requests in progress and open streams are not written. |

### `metrics_listen`

TCP address, as `host:port`, of the listener that serves `GET /metrics` when `metrics_exporter = prometheus`, apart from `listen`. It has no authentication: keep it on loopback or a private network, or protect it with the network or a proxy.

- Environment variable: `MY_METRICS_LISTEN`
- Default: `127.0.0.1:9464`
- Required: no

### `metrics_emf_namespace`

CloudWatch namespace of the metrics of `metrics_exporter = emf`.

- Environment variable: `MY_METRICS_EMF_NAMESPACE`
- Default: `my`
- Required: no

### `metrics_emf_dimensions`

Labels that become CloudWatch dimensions with `metrics_exporter = emf`, comma-separated, such as `my.surface, my.role`; the other labels stay fields of the log line. CloudWatch bills every combination of dimension values as a metric, so the deployment template chooses them. Empty makes every metric one series per namespace.

- Environment variable: `MY_METRICS_EMF_DIMENSIONS`
- Default: none
- Required: no

### `traces_exporter`

Where the distributed traces go: the spans of the HTTP requests, the job runs (each the root of a new trace linked to the request that enqueued it), the transactions and queries, the deliveries, and the outbound requests. Span attributes never carry an account, a URL or a path, a client IP, a User-Agent, an SQL text, or an error message; the outbound requests carry the destination's host name. No outbound request carries a `traceparent`, and an incoming `traceparent` is taken only from `trusted_proxies`.

- Environment variable: `MY_TRACES_EXPORTER`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | No spans; tracing costs next to nothing. The log lines still carry `trace_id` and `span_id` of a `traceparent` from `trusted_proxies`, which correlates them with the platform's traces (Cloud Run). |
| `otlp` | OTLP over HTTP (protobuf, gzip) to an OpenTelemetry Collector or another receiver, configured by the standard environment variables `OTEL_EXPORTER_OTLP_TRACES_ENDPOINT` or `OTEL_EXPORTER_OTLP_ENDPOINT` (default `http://localhost:4318`) and `OTEL_EXPORTER_OTLP_HEADERS`. The server does not authenticate to Cloud Trace or X-Ray itself: send to a collector next to it, which does ([running the server](https://mycroblog.dev/next/operations/running/)). |

### `traces_sample_ratio`

Share of the traces recorded with `traces_exporter = otlp`, from `0` to `1`, among the requests and jobs without a sampled parent: a request follows the sampling of a `traceparent` from `trusted_proxies`, and a job is recorded when the request that enqueued it was. The default records about 1,000 traces a day of an instance that serves 100,000 requests a day.

- Environment variable: `MY_TRACES_SAMPLE_RATIO`
- Default: `0.01`
- Required: no

### `traces_flush`

When the spans of `traces_exporter = otlp` are sent.

- Environment variable: `MY_TRACES_FLUSH`
- Default: `batch`
- Required: no

| Value | Description |
| --- | --- |
| `batch` | In batches, every few seconds, in the background; for processes that keep running (self-hosted). |
| `unit` | Also at the end of every recorded HTTP request and job, before it returns, to a collector on the same host; for the platforms that throttle or freeze an idle instance (Cloud Run with request-based billing). A request that is not recorded sends nothing. |

### `errors_exporter`

Where the error reports also go: the panics, the errors behind a 5xx response or a job that used up its retries, the startup failures, and the uncaught errors of the web client, each with its stack and a fingerprint that groups its repeats. The log line of each report and the error groups in the database are kept whatever this setting.

- Environment variable: `MY_ERRORS_EXPORTER`
- Default: `none`
- Required: no

| Value | Description |
| --- | --- |
| `none` | Only the log line and the error groups in the database. |
| `sentry` | Also to a Sentry-compatible service at `errors_sentry_dsn`: Sentry (hosted or self-hosted), GlitchTip, or Datadog Error Tracking. |
| `otlp` | Also as OpenTelemetry log records of exceptions (event name `exception`) over OTLP/HTTP (protobuf, gzip), to the endpoint of the standard environment variables `OTEL_EXPORTER_OTLP_LOGS_ENDPOINT` or `OTEL_EXPORTER_OTLP_ENDPOINT` (default `http://localhost:4318`) with `OTEL_EXPORTER_OTLP_HEADERS`. Whether the receiving backend groups them as errors depends on the backend; for Sentry, GlitchTip, and Datadog use `sentry`. |

### `errors_sentry_dsn`

DSN of the project that receives the error reports of `errors_exporter = sentry`, as the service shows it, such as `https://<key>@o0.ingest.sentry.io/<project>`.

- Environment variable: `MY_ERRORS_SENTRY_DSN` or `MY_ERRORS_SENTRY_DSN_FILE`
- Default: none
- Required: when `errors_exporter` is `sentry`

### `errors_environment`

Name of the environment that every error report carries, such as `staging`, to tell the reports of several deployments apart: up to 64 letters, digits, dots, hyphens, and underscores.

- Environment variable: `MY_ERRORS_ENVIRONMENT`
- Default: the value of `profile`
- Required: no

### `pprof_listen`

TCP address, as `host:port`, of a listener that serves the Go runtime's profiles (heap, allocations, goroutines, CPU, execution trace) under `/debug/pprof/`, for `go tool pprof`, apart from `listen`. Empty opens no such listener. It has no authentication, and a profile shows the internals of the process and costs CPU while it is taken: keep it on loopback, and reach it through a tunnel.

- Environment variable: `MY_PPROF_LISTEN`
- Default: none
- Required: no

### `plugin_translator_address`

Where the `translator` plugin listens: `unix:///path/to.sock` or `https://host[:port]`; `http://` is refused, on loopback too. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_TRANSLATOR_ADDRESS`
- Default: none
- Required: no

### `plugin_translator_protocol`

What the server sends to the `translator` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_TRANSLATOR_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_translator_token`

Bearer token that the server sends to the `translator` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_TRANSLATOR_TOKEN` or `MY_PLUGIN_TRANSLATOR_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_translator_tls_ca_file`

File of the CA that verifies the `translator` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_TRANSLATOR_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_translator_tls_cert_file`

File of the client certificate that the server shows the `translator` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_TRANSLATOR_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_translator_tls_key_file`

File of the key of that client certificate of the `translator` plugin.

- Environment variable: `MY_PLUGIN_TRANSLATOR_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_translator_timeout`

Deadline of one call of the `translator` plugin, less than `30s`. The server stops a call after `20s` whatever this says.

- Environment variable: `MY_PLUGIN_TRANSLATOR_TIMEOUT`
- Default: `10s`
- Required: no

### `plugin_translator_name`

The name that readers see next to a translation by the `translator` plugin and that the admin screen lists, such as the service behind your plugin, at most 100 characters. Clients get this name and never the address.

- Environment variable: `MY_PLUGIN_TRANSLATOR_NAME`
- Default: none
- Required: when `plugin_translator_address` is set

### `plugin_search_address`

Where the `search` plugin listens: `unix:///path/to.sock` or `https://host[:port]`; `http://` is refused, on loopback too. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_SEARCH_ADDRESS`
- Default: none
- Required: when `search = plugin`

### `plugin_search_protocol`

What the server sends to the `search` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_SEARCH_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_search_token`

Bearer token that the server sends to the `search` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_SEARCH_TOKEN` or `MY_PLUGIN_SEARCH_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_search_tls_ca_file`

File of the CA that verifies the `search` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_SEARCH_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_search_tls_cert_file`

File of the client certificate that the server shows the `search` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_SEARCH_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_search_tls_key_file`

File of the key of that client certificate of the `search` plugin.

- Environment variable: `MY_PLUGIN_SEARCH_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_search_timeout`

Deadline of one call of the `search` plugin, less than `30s`. The server stops a call after `10s` whatever this says.

- Environment variable: `MY_PLUGIN_SEARCH_TIMEOUT`
- Default: `10s`
- Required: no

### `plugin_billing_address`

Where the `billing` plugin listens: `unix:///path/to.sock`, `https://host[:port]`, or `http://host[:port]`, which is plain TCP: nothing encrypts or authenticates it, and the server warns at start. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_BILLING_ADDRESS`
- Default: none
- Required: when `billing_sources` includes `plugin`

### `plugin_billing_protocol`

What the server sends to the `billing` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_BILLING_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_billing_token`

Bearer token that the server sends to the `billing` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_BILLING_TOKEN` or `MY_PLUGIN_BILLING_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_billing_tls_ca_file`

File of the CA that verifies the `billing` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_BILLING_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_billing_tls_cert_file`

File of the client certificate that the server shows the `billing` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_BILLING_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_billing_tls_key_file`

File of the key of that client certificate of the `billing` plugin.

- Environment variable: `MY_PLUGIN_BILLING_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_billing_timeout`

Deadline of one call of the `billing` plugin, less than `5m0s`.

- Environment variable: `MY_PLUGIN_BILLING_TIMEOUT`
- Default: `2m0s`
- Required: no

### `plugin_billing_location`

IANA time zone that cuts the billing months of `billing_sources = plugin`, such as `UTC` or `Asia/Tokyo`. The imports ask the plugin for the months of this zone.

- Environment variable: `MY_PLUGIN_BILLING_LOCATION`
- Default: none
- Required: when `billing_sources` includes `plugin`

### `plugin_billing_provider`

The provider's name that the totals, the costs page, and the audit log show for `billing_sources = plugin`, such as `Hetzner`, at most 100 characters. All the plugin's costs are under this one name, and the services of its lines tell them apart.

- Environment variable: `MY_PLUGIN_BILLING_PROVIDER`
- Default: none
- Required: when `billing_sources` includes `plugin`

### `plugin_mailer_address`

Where the `mailer` plugin listens: `unix:///path/to.sock` or `https://host[:port]`; `http://` is refused, on loopback too. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_MAILER_ADDRESS`
- Default: none
- Required: when `mail = plugin`

### `plugin_mailer_protocol`

What the server sends to the `mailer` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_MAILER_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_mailer_token`

Bearer token that the server sends to the `mailer` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_MAILER_TOKEN` or `MY_PLUGIN_MAILER_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_mailer_tls_ca_file`

File of the CA that verifies the `mailer` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_MAILER_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_mailer_tls_cert_file`

File of the client certificate that the server shows the `mailer` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_MAILER_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_mailer_tls_key_file`

File of the key of that client certificate of the `mailer` plugin.

- Environment variable: `MY_PLUGIN_MAILER_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_mailer_timeout`

Deadline of one call of the `mailer` plugin, less than `5m0s`.

- Environment variable: `MY_PLUGIN_MAILER_TIMEOUT`
- Default: `30s`
- Required: no

### `plugin_media_image_address`

Where the `media_image` plugin listens: `unix:///path/to.sock`, `https://host[:port]`, or `http://host[:port]`, which is plain TCP: nothing encrypts or authenticates it, and the server warns at start. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_MEDIA_IMAGE_ADDRESS`
- Default: none
- Required: when `media_image = plugin`

### `plugin_media_image_protocol`

What the server sends to the `media_image` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_MEDIA_IMAGE_PROTOCOL`
- Default: `proto`
- Required: no

| Value | Description |
| --- | --- |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_media_image_token`

Bearer token that the server sends to the `media_image` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_MEDIA_IMAGE_TOKEN` or `MY_PLUGIN_MEDIA_IMAGE_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_media_image_tls_ca_file`

File of the CA that verifies the `media_image` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_MEDIA_IMAGE_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_media_image_tls_cert_file`

File of the client certificate that the server shows the `media_image` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_MEDIA_IMAGE_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_media_image_tls_key_file`

File of the key of that client certificate of the `media_image` plugin.

- Environment variable: `MY_PLUGIN_MEDIA_IMAGE_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_media_image_timeout`

Deadline of one call of the `media_image` plugin, less than `5m0s`.

- Environment variable: `MY_PLUGIN_MEDIA_IMAGE_TIMEOUT`
- Default: `4m0s`
- Required: no

### `plugin_media_video_address`

Where the `media_video` plugin listens: `unix:///path/to.sock`, `https://host[:port]`, or `http://host[:port]`, which is plain TCP: nothing encrypts or authenticates it, and the server warns at start. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_MEDIA_VIDEO_ADDRESS`
- Default: none
- Required: when `media_video = plugin`

### `plugin_media_video_protocol`

What the server sends to the `media_video` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_MEDIA_VIDEO_PROTOCOL`
- Default: `proto`
- Required: no

| Value | Description |
| --- | --- |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_media_video_token`

Bearer token that the server sends to the `media_video` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_MEDIA_VIDEO_TOKEN` or `MY_PLUGIN_MEDIA_VIDEO_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_media_video_tls_ca_file`

File of the CA that verifies the `media_video` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_MEDIA_VIDEO_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_media_video_tls_cert_file`

File of the client certificate that the server shows the `media_video` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_MEDIA_VIDEO_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_media_video_tls_key_file`

File of the key of that client certificate of the `media_video` plugin.

- Environment variable: `MY_PLUGIN_MEDIA_VIDEO_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_media_video_timeout`

Deadline of one call of the `media_video` plugin, less than `5m0s`.

- Environment variable: `MY_PLUGIN_MEDIA_VIDEO_TIMEOUT`
- Default: `4m0s`
- Required: no

### `plugin_media_audio_address`

Where the `media_audio` plugin listens: `unix:///path/to.sock`, `https://host[:port]`, or `http://host[:port]`, which is plain TCP: nothing encrypts or authenticates it, and the server warns at start. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_MEDIA_AUDIO_ADDRESS`
- Default: none
- Required: when `media_audio = plugin`

### `plugin_media_audio_protocol`

What the server sends to the `media_audio` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_MEDIA_AUDIO_PROTOCOL`
- Default: `proto`
- Required: no

| Value | Description |
| --- | --- |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_media_audio_token`

Bearer token that the server sends to the `media_audio` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_MEDIA_AUDIO_TOKEN` or `MY_PLUGIN_MEDIA_AUDIO_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_media_audio_tls_ca_file`

File of the CA that verifies the `media_audio` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_MEDIA_AUDIO_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_media_audio_tls_cert_file`

File of the client certificate that the server shows the `media_audio` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_MEDIA_AUDIO_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_media_audio_tls_key_file`

File of the key of that client certificate of the `media_audio` plugin.

- Environment variable: `MY_PLUGIN_MEDIA_AUDIO_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_media_audio_timeout`

Deadline of one call of the `media_audio` plugin, less than `5m0s`.

- Environment variable: `MY_PLUGIN_MEDIA_AUDIO_TIMEOUT`
- Default: `4m0s`
- Required: no

### `plugin_key_manager_address`

Where the `key_manager` plugin listens: `unix:///path/to.sock` or `https://host[:port]`; `http://` is refused, on loopback too. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_KEY_MANAGER_ADDRESS`
- Default: none
- Required: when `key_manager = plugin`

### `plugin_key_manager_protocol`

What the server sends to the `key_manager` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_KEY_MANAGER_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_key_manager_token`

Bearer token that the server sends to the `key_manager` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_KEY_MANAGER_TOKEN` or `MY_PLUGIN_KEY_MANAGER_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_key_manager_tls_ca_file`

File of the CA that verifies the `key_manager` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_KEY_MANAGER_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_key_manager_tls_cert_file`

File of the client certificate that the server shows the `key_manager` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_KEY_MANAGER_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_key_manager_tls_key_file`

File of the key of that client certificate of the `key_manager` plugin.

- Environment variable: `MY_PLUGIN_KEY_MANAGER_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_key_manager_timeout`

Deadline of one call of the `key_manager` plugin, less than `30s`.

- Environment variable: `MY_PLUGIN_KEY_MANAGER_TIMEOUT`
- Default: `5s`
- Required: no

### `plugin_inbound_status_address`

Where the `inbound.status` plugin listens: `unix:///path/to.sock` or `https://host[:port]`; `http://` is refused, on loopback too. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_INBOUND_STATUS_ADDRESS`
- Default: none
- Required: no

### `plugin_inbound_status_protocol`

What the server sends to the `inbound.status` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_INBOUND_STATUS_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_inbound_status_token`

Bearer token that the server sends to the `inbound.status` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_INBOUND_STATUS_TOKEN` or `MY_PLUGIN_INBOUND_STATUS_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_inbound_status_tls_ca_file`

File of the CA that verifies the `inbound.status` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_INBOUND_STATUS_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_inbound_status_tls_cert_file`

File of the client certificate that the server shows the `inbound.status` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_INBOUND_STATUS_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_inbound_status_tls_key_file`

File of the key of that client certificate of the `inbound.status` plugin.

- Environment variable: `MY_PLUGIN_INBOUND_STATUS_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_inbound_status_timeout`

Deadline of one call of the `inbound.status` plugin, less than `5m0s`.

- Environment variable: `MY_PLUGIN_INBOUND_STATUS_TIMEOUT`
- Default: `3s`
- Required: no

### `plugin_inbound_follow_address`

Where the `inbound.follow` plugin listens: `unix:///path/to.sock`, `https://host[:port]`, or `http://host[:port]`, which is plain TCP: nothing encrypts or authenticates it, and the server warns at start. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_INBOUND_FOLLOW_ADDRESS`
- Default: none
- Required: no

### `plugin_inbound_follow_protocol`

What the server sends to the `inbound.follow` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_INBOUND_FOLLOW_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_inbound_follow_token`

Bearer token that the server sends to the `inbound.follow` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_INBOUND_FOLLOW_TOKEN` or `MY_PLUGIN_INBOUND_FOLLOW_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_inbound_follow_tls_ca_file`

File of the CA that verifies the `inbound.follow` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_INBOUND_FOLLOW_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_inbound_follow_tls_cert_file`

File of the client certificate that the server shows the `inbound.follow` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_INBOUND_FOLLOW_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_inbound_follow_tls_key_file`

File of the key of that client certificate of the `inbound.follow` plugin.

- Environment variable: `MY_PLUGIN_INBOUND_FOLLOW_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_inbound_follow_timeout`

Deadline of one call of the `inbound.follow` plugin, less than `5m0s`.

- Environment variable: `MY_PLUGIN_INBOUND_FOLLOW_TIMEOUT`
- Default: `3s`
- Required: no

### `plugin_signup_check_address`

Where the `signup.check` plugin listens: `unix:///path/to.sock` or `https://host[:port]`; `http://` is refused, on loopback too. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_SIGNUP_CHECK_ADDRESS`
- Default: none
- Required: no

### `plugin_signup_check_protocol`

What the server sends to the `signup.check` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_SIGNUP_CHECK_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_signup_check_token`

Bearer token that the server sends to the `signup.check` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_SIGNUP_CHECK_TOKEN` or `MY_PLUGIN_SIGNUP_CHECK_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_signup_check_tls_ca_file`

File of the CA that verifies the `signup.check` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_SIGNUP_CHECK_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_signup_check_tls_cert_file`

File of the client certificate that the server shows the `signup.check` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_SIGNUP_CHECK_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_signup_check_tls_key_file`

File of the key of that client certificate of the `signup.check` plugin.

- Environment variable: `MY_PLUGIN_SIGNUP_CHECK_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_signup_check_timeout`

Deadline of one call of the `signup.check` plugin, less than `30s`.

- Environment variable: `MY_PLUGIN_SIGNUP_CHECK_TIMEOUT`
- Default: `2s`
- Required: no

### `plugin_media_check_address`

Where the `media.check` plugin listens: `unix:///path/to.sock`, `https://host[:port]`, or `http://host[:port]`, which is plain TCP: nothing encrypts or authenticates it, and the server warns at start. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_MEDIA_CHECK_ADDRESS`
- Default: none
- Required: no

### `plugin_media_check_protocol`

What the server sends to the `media.check` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_MEDIA_CHECK_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_media_check_token`

Bearer token that the server sends to the `media.check` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_MEDIA_CHECK_TOKEN` or `MY_PLUGIN_MEDIA_CHECK_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_media_check_tls_ca_file`

File of the CA that verifies the `media.check` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_MEDIA_CHECK_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_media_check_tls_cert_file`

File of the client certificate that the server shows the `media.check` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_MEDIA_CHECK_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_media_check_tls_key_file`

File of the key of that client certificate of the `media.check` plugin.

- Environment variable: `MY_PLUGIN_MEDIA_CHECK_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_media_check_timeout`

Deadline of one call of the `media.check` plugin, less than `30s`.

- Environment variable: `MY_PLUGIN_MEDIA_CHECK_TIMEOUT`
- Default: `10s`
- Required: no

### `plugin_local_status_address`

Where the `local.status` plugin listens: `unix:///path/to.sock` or `https://host[:port]`; `http://` is refused, on loopback too. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_LOCAL_STATUS_ADDRESS`
- Default: none
- Required: no

### `plugin_local_status_protocol`

What the server sends to the `local.status` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_LOCAL_STATUS_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_local_status_token`

Bearer token that the server sends to the `local.status` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_LOCAL_STATUS_TOKEN` or `MY_PLUGIN_LOCAL_STATUS_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_local_status_tls_ca_file`

File of the CA that verifies the `local.status` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_LOCAL_STATUS_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_local_status_tls_cert_file`

File of the client certificate that the server shows the `local.status` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_LOCAL_STATUS_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_local_status_tls_key_file`

File of the key of that client certificate of the `local.status` plugin.

- Environment variable: `MY_PLUGIN_LOCAL_STATUS_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_local_status_timeout`

Deadline of one call of the `local.status` plugin, less than `30s`.

- Environment variable: `MY_PLUGIN_LOCAL_STATUS_TIMEOUT`
- Default: `1s`
- Required: no

### `plugin_inbound_actor_address`

Where the `inbound.actor` plugin listens: `unix:///path/to.sock`, `https://host[:port]`, or `http://host[:port]`, which is plain TCP: nothing encrypts or authenticates it, and the server warns at start. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_INBOUND_ACTOR_ADDRESS`
- Default: none
- Required: no

### `plugin_inbound_actor_protocol`

What the server sends to the `inbound.actor` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_INBOUND_ACTOR_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_inbound_actor_token`

Bearer token that the server sends to the `inbound.actor` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_INBOUND_ACTOR_TOKEN` or `MY_PLUGIN_INBOUND_ACTOR_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_inbound_actor_tls_ca_file`

File of the CA that verifies the `inbound.actor` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_INBOUND_ACTOR_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_inbound_actor_tls_cert_file`

File of the client certificate that the server shows the `inbound.actor` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_INBOUND_ACTOR_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_inbound_actor_tls_key_file`

File of the key of that client certificate of the `inbound.actor` plugin.

- Environment variable: `MY_PLUGIN_INBOUND_ACTOR_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_inbound_actor_timeout`

Deadline of one call of the `inbound.actor` plugin, less than `30s`.

- Environment variable: `MY_PLUGIN_INBOUND_ACTOR_TIMEOUT`
- Default: `3s`
- Required: no

### `plugin_local_profile_address`

Where the `local.profile` plugin listens: `unix:///path/to.sock`, `https://host[:port]`, or `http://host[:port]`, which is plain TCP: nothing encrypts or authenticates it, and the server warns at start. Empty delegates nothing, and the server opens no connection. Host names are resolved per connection. Restart to change it; no instance setting or API can.

- Environment variable: `MY_PLUGIN_LOCAL_PROFILE_ADDRESS`
- Default: none
- Required: no

### `plugin_local_profile_protocol`

What the server sends to the `local.profile` plugin; it must be what the plugin accepts.

- Environment variable: `MY_PLUGIN_LOCAL_PROFILE_PROTOCOL`
- Default: `json`
- Required: no

| Value | Description |
| --- | --- |
| `json` | The Connect protocol with JSON, which a plugin of an HTTP server and a JSON library serves with no Connect library. |
| `proto` | The Connect protocol with binary protobuf, for a server on a Connect library. |
| `grpc` | gRPC, for a server that speaks only that. |

### `plugin_local_profile_token`

Bearer token that the server sends to the `local.profile` plugin, over `https://` or `unix://` only.

- Environment variable: `MY_PLUGIN_LOCAL_PROFILE_TOKEN` or `MY_PLUGIN_LOCAL_PROFILE_TOKEN_FILE`
- Default: none
- Required: no

### `plugin_local_profile_tls_ca_file`

File of the CA that verifies the `local.profile` plugin at an `https://` address.

- Environment variable: `MY_PLUGIN_LOCAL_PROFILE_TLS_CA_FILE`
- Default: the system's roots
- Required: no

### `plugin_local_profile_tls_cert_file`

File of the client certificate that the server shows the `local.profile` plugin at an `https://` address (mTLS), with its key file.

- Environment variable: `MY_PLUGIN_LOCAL_PROFILE_TLS_CERT_FILE`
- Default: none
- Required: no

### `plugin_local_profile_tls_key_file`

File of the key of that client certificate of the `local.profile` plugin.

- Environment variable: `MY_PLUGIN_LOCAL_PROFILE_TLS_KEY_FILE`
- Default: none
- Required: no

### `plugin_local_profile_timeout`

Deadline of one call of the `local.profile` plugin, less than `30s`.

- Environment variable: `MY_PLUGIN_LOCAL_PROFILE_TIMEOUT`
- Default: `1s`
- Required: no

## Renamed and removed settings

Using one of these keys, in the config file or as its environment variable (or `<NAME>_FILE`), stops the server with an error.

None.
