Self-Hosted Guide for Azure

[rachaelsingleton]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit report (我已阅读并同意 Language Policy).
• Please do not modify this template :) and fill in all the required fields.

Provide a description of requested docs changes

Looking for a guide to self-host in Azure.

[dosubot[bot]]

To self-host the Dify application in Azure, you need to configure various environment variables and settings. Here are the key steps and configurations you need to set up:

1. Common Variables: Set the URLs for different services.

   export CONSOLE_API_URL="https://api.console.dify.ai"
   export CONSOLE_WEB_URL="https://console.dify.ai"
   export SERVICE_API_URL="https://api.dify.ai"
   export APP_API_URL="https://app.dify.ai"
   export APP_WEB_URL="https://api.app.dify.ai"
   export FILES_URL="your-files-url"

2. Server Configuration: Set the server mode, debug options, and other server-related settings.

   export MODE="api"  # or "worker"
   export DEBUG=false
   export FLASK_DEBUG=false
   export SECRET_KEY=$(openssl rand -base64 42)
   export DEPLOY_ENV="PRODUCTION"
   export LOG_LEVEL="ERROR"
   export MIGRATION_ENABLED=true
   export CHECK_UPDATE_URL=false
   export OPENAI_API_BASE="https://api.openai.com/v1"

3. Container Startup Configuration: If using Docker, set the following:

   export DIFY_BIND_ADDRESS="0.0.0.0"
   export DIFY_PORT=5001
   export SERVER_WORKER_AMOUNT=3  # Example: number of CPU cores x 2 + 1
   export SERVER_WORKER_CLASS="gevent"
   export GUNICORN_TIMEOUT=360
   export CELERY_WORKER_CLASS="gevent"
   export CELERY_WORKER_AMOUNT=1

4. Database Configuration: Configure PostgreSQL settings.

   export DB_USERNAME="your-db-username"
   export DB_PASSWORD="your-db-password"
   export DB_HOST="your-db-host"
   export DB_PORT=5432
   export DB_DATABASE="your-db-name"
   export SQLALCHEMY_POOL_SIZE=30
   export SQLALCHEMY_POOL_RECYCLE=3600
   export SQLALCHEMY_ECHO=false

5. Redis Configuration: Set up Redis for caching and pub/sub.

   export REDIS_HOST="your-redis-host"
   export REDIS_PORT=6379
   export REDIS_DB=0
   export REDIS_USERNAME=""
   export REDIS_PASSWORD="your-redis-password"
   export REDIS_USE_SSL=false

6. Celery Configuration: Configure Celery for task queues.

   export CELERY_BROKER_URL="redis://:your-redis-password@your-redis-host:6379/1"
   export BROKER_USE_SSL=false

7. CORS Configuration: Set CORS policies.

   export CONSOLE_CORS_ALLOW_ORIGINS="*"
   export WEB_API_CORS_ALLOW_ORIGINS="*"

8. File Storage Configuration: Configure storage type and paths.

   export STORAGE_TYPE="azure-blob"
   export AZURE_BLOB_ACCOUNT_NAME="your-account-name"
   export AZURE_BLOB_ACCOUNT_KEY="your-account-key"
   export AZURE_BLOB_CONTAINER_NAME="your-container-name"
   export AZURE_BLOB_ACCOUNT_URL="https://your_account_name.blob.core.windows.net"

9. Vector Database Configuration: Set up the vector database.

   export VECTOR_STORE="weaviate"
   export WEAVIATE_ENDPOINT="http://weaviate:8080"
   export WEAVIATE_API_KEY="your-weaviate-api-key"
   export WEAVIATE_BATCH_SIZE=100
   export WEAVIATE_GRPC_ENABLED=true

10. Knowledge Configuration: Configure file upload limits and ETL type.

    export UPLOAD_FILE_SIZE_LIMIT=15M
    export UPLOAD_FILE_BATCH_LIMIT=5
    export ETL_TYPE="dify"

11. Multi-modal Configuration: Set multi-modal settings.

    export MULTIMODAL_SEND_IMAGE_FORMAT="base64"
    export UPLOAD_IMAGE_FILE_SIZE_LIMIT=10M

12. Sentry Configuration: Set up Sentry for monitoring.

    export SENTRY_DSN="your-sentry-dsn"
    export SENTRY_TRACES_SAMPLE_RATE=0.01
    export SENTRY_PROFILES_SAMPLE_RATE=0.01

13. Notion Integration Configuration: Configure Notion integration.

    export NOTION_INTEGRATION_TYPE="internal"
    export NOTION_CLIENT_SECRET="your-notion-client-secret"
    export NOTION_CLIENT_ID="your-notion-client-id"
    export NOTION_INTERNAL_SECRET="your-notion-internal-secret"

14. Mail Configuration: Set up mail settings.

    export MAIL_TYPE="smtp"
    export SMTP_SERVER="your-smtp-server"
    export SMTP_PORT=587
    export SMTP_USERNAME="your-smtp-username"
    export SMTP_PASSWORD="your-smtp-password"
    export SMTP_USE_TLS=true
    export MAIL_DEFAULT_SEND_FROM="no-reply@dify.ai"

15. Other Configurations: Set additional configurations.

    export INVITE_EXPIRY_HOURS=72

Additionally, you need to configure the hosting provider settings in your application. For Azure OpenAI, you can enable and configure it as follows:

class HostingConfiguration:
    provider_map: dict[str, HostingProvider] = {}
    moderation_config: HostedModerationConfig = None

    def init_app(self, app: Flask) -> None:
        config = app.config

        if config.get('EDITION') != 'CLOUD':
            return

        self.provider_map["azure_openai"] = self.init_azure_openai(config)
        self.provider_map["openai"] = self.init_openai(config)
        self.provider_map["anthropic"] = self.init_anthropic(config)
        self.provider_map["minimax"] = self.init_minimax(config)
        self.provider_map["spark"] = self.init_spark(config)
        self.provider_map["zhipuai"] = self.init_zhipuai(config)

        self.moderation_config = self.init_moderation_config(config)

    def init_azure_openai(self, app_config: Config) -> HostingProvider:
        quota_unit = QuotaUnit.TIMES
        if app_config.get("HOSTED_AZURE_OPENAI_ENABLED"):
            credentials = {
                "openai_api_key": app_config.get("HOSTED_AZURE_OPENAI_API_KEY"),
                "openai_api_base": app_config.get("HOSTED_AZURE_OPENAI_API_BASE"),
                "base_model_name": "gpt-35-turbo"
            }

            quotas = []
            hosted_quota_limit = int(app_config.get("HOSTED_AZURE_OPENAI_QUOTA_LIMIT", "1000"))
            trial_quota = TrialHostingQuota(
                quota_limit=hosted_quota_limit,
                restrict_models=[
                    RestrictModel(model="gpt-4", base_model_name="gpt-4", model_type=ModelType.LLM),
                    RestrictModel(model="gpt-4-32k", base_model_name="gpt-4-32k", model_type=ModelType.LLM),
                    RestrictModel(model="gpt-4-1106-preview", base_model_name="gpt-4-1106-preview", model_type=ModelType.LLM),
                    RestrictModel(model="gpt-4-vision-preview", base_model_name="gpt-4-vision-preview", model_type=ModelType.LLM),
                    RestrictModel(model="gpt-35-turbo", base_model_name="gpt-35-turbo", model_type=ModelType.LLM),
                    RestrictModel(model="gpt-35-turbo-1106", base_model_name="gpt-35-turbo-1106", model_type=ModelType.LLM),
                    RestrictModel(model="gpt-35-turbo-instruct", base_model_name="gpt-35-turbo-instruct", model_type=ModelType.LLM),
                    RestrictModel(model="gpt-35-turbo-16k", base_model_name="gpt-35-turbo-16k", model type=ModelType.LLM),
                    RestrictModel(model="text-davinci-003", base_model_name="text-davinci-003", model type=ModelType.LLM),
                    RestrictModel(model="text-embedding-ada-002", base_model_name="text-embedding-ada-002", model type=ModelType.TEXT_EMBEDDING),
                    RestrictModel(model="text-embedding-3-small", base_model_name="text-embedding-3-small", model type=ModelType.TEXT_EMBEDDING),
                    RestrictModel(model="text-embedding-3-large", base_model_name="text-embedding-3-large", model type=ModelType.TEXT_EMBEDDING),
                ]
            )
            quotas.append(trial_quota)

            return HostingProvider(
                enabled=True,
                credentials=credentials,
                quota_unit=quota_unit,
                quotas=quotas
            )

        return HostingProvider(
            enabled=False,
            quota_unit=quota_unit,
        )

After setting these environment variables and configuring the hosting provider, you can deploy your application to Azure using Docker or other deployment methods supported by Azure ^[1][2].

To continue talking to Dosu, mention @dosu.

[Sarmingsteiner]

I am struggling to make Dify run on my Synology NAS, too.
@dosu Which Container exactly did you configure like that?

Container Startup Configuration: If using Docker, set the following:

export DIFY_BIND_ADDRESS="0.0.0.0"
export DIFY_PORT=5001
export SERVER_WORKER_AMOUNT=3 # Example: number of CPU cores x 2 + 1
export SERVER_WORKER_CLASS="gevent"
export GUNICORN_TIMEOUT=360
export CELERY_WORKER_CLASS="gevent"
export CELERY_WORKER_AMOUNT=1

[crazywoola]

https://docs.dify.ai/getting-started/install-self-hosted/docker-compose @rachaelsingleton

It can be easily started by docker. Just install the docker in azure machines, and follow the docs above.

[rachaelsingleton]

Sure, followed the steps but there's issue after issue that keeps coming up.

I've set up Postgres and Redis in azure and updated the connection strings/env variables as required. I cannot get passed the install screen, it keeps cycling between signing up and logging in.

There are no errors in the logs, and the only thing I have to go on is a 403 - 2024-06-03 10:36:10 nginx-1 | 172.23.0.1 - - [03/Jun/2024:14:36:10 +0000] "POST /console/api/setup HTTP/1.1" 403 154 "http://localhost/install" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/125.0.0.0 Safari/537.36 Edg/125.0.0.0" "-"

I've tried to pull out each service individually and I believe it's when the postgres DB is set up in azure.

[crazywoola]

The circle keeps spinning typically means you are trying to access backend from a.xxx.com but the client send request to localhost.
You could check out the logs from the chrome console.

dify/docker/docker-compose.yaml

Lines 16 to 37 in a325a29

	# The base URL of console application web frontend, refers to the Console base URL of WEB service if console domain is
	# different from api or web app domain.
	# example: http://cloud.dify.ai
	CONSOLE_WEB_URL: ''
	# Password for admin user initialization.
	# If left unset, admin user will not be prompted for a password when creating the initial admin account.
	INIT_PASSWORD: ''
	# The base URL of console application api server, refers to the Console base URL of WEB service if console domain is
	# different from api or web app domain.
	# example: http://cloud.dify.ai
	CONSOLE_API_URL: ''
	# The URL prefix for Service API endpoints, refers to the base URL of the current API service if api domain is
	# different from console domain.
	# example: http://api.dify.ai
	SERVICE_API_URL: ''
	# The URL prefix for Web APP frontend, refers to the Web App base URL of WEB service if web app domain is different from
	# console or api domain.
	# example: http://udify.app
	APP_WEB_URL: ''
	# File preview or download Url prefix.
	# used to display File preview or download Url to the front-end or as Multi-model inputs;
	# Url is signed and has expiration time.

In this case you need to make sure you've set those up correctly.

[rachaelsingleton]

It's not spinning. I am prompted with the install page, I fill out the info get a success message, I'm redirected to the login page. When I try to login, I'm redirected to the install page and so on. I've started to piece this out in hopes to find the error, so at this point everything but the postrgres DB is spun up via the docker compose. I've updated the environment variables in the docker compose to point to my db hosted in azure. That's the only change I've made, updating environment variables. I get no connection error to the DB.. I've messed up the password and username on purpose, and when I do this I get an connection error.

[crazywoola]

I guess you need to check out the accounts table if you have setup your account correctly.

And since you are using cloud db, and the configuration of db itself is no longer needed.

Only these will take effect

      # The configurations of postgres database connection.
      # It is consistent with the configuration in the 'db' service below.
      DB_USERNAME: postgres
      DB_PASSWORD: difyai123456
      DB_HOST: db
      DB_PORT: 5432
      DB_DATABASE: dify

[crazywoola]

And is this public available? I can take a look at frontend api requests to see what's going on.

[gijigae]

@rachaelsingleton I'm currently running a dify instance with Azure Database for PostgreSQL flexible server without any issue on Azure VM. Have you checked the logs of api or worker containers to see what errors they returned? When I first started the instance, I received the following error in the api container and had to add the missing extenstion, uuid-ossp. After that I was able to create an admin account and log in. Hope it helps.

api-1  | sqlalchemy.exc.NotSupportedError: (psycopg2.errors.FeatureNotSupported) extension "uuid-ossp" is not allow-listed for "azure_pg_admin" users in Azure Database for PostgreSQL
api-1  | HINT:  to see the full list of allowed extensions, please run: "show azure.extensions;"
api-1  |
api-1  | [SQL: CREATE EXTENSION IF NOT EXISTS "uuid-ossp";]
api-1  | (Background on this error at: https://sqlalche.me/e/20/tw8g)
api-1  | Running migrations