Synology AI Console
DSM Version
7.4 7.3 7.2

Synology AI Console

Features

  • Expands the functionalities in supported Synology packages with third-party AI API integration
  • Provides de-identification options to protect data privacy
  • Offers a centralized management console for administrators to enable AI for packages, monitor API integration status and usage, and manage permissions for groups and users

Specifications

Requirements and compatibility
GPU-powered models
  • Available only on supported Synology storage systems with a GPU
  • Model control:
    • Embedding model: Enables AI-powered search in Synology Drive by understanding document context and intent beyond keyword matching
    • OCR (Optical Character Recognition) model: Extracts text from images, scanned documents, and video frames to make their content searchable in Synology Drive
      • Supported languages: Simplified Chinese, Traditional Chinese, Czech, Danish, English, French, German, Hungarian, Italian, Japanese, Korean, Dutch, Norwegian, Polish, Portuguese (Brazil), Portuguese (Portugal), Russian, Spanish, Swedish, Thai, Turkish
      • Selecting more languages increases memory usage
    • STT (Speech-to-Text) model: Converts speech in audio and video files into searchable text
    • Image captioning model: Generates descriptive captions for images or video frames based on visual content to enable search by description
    • Refer to the estimated memory usage for each model below. Because the system and other services also consume memory, we recommend reserving at least 16 GB for the system memory
      • Embedding model: 2 GB
      • OCR (Optical Character Recognition) model: 2 GB
      • STT (Speech-to-Text) model: 3 GB
      • Image captioning model: 3 GB
API integration management
  • Supported AI providers: Amazon Bedrock, Azure OpenAI, Baidu AI Cloud, Google AI Studio, Google Vertex AI, OpenAI, and other providers with OpenAI-compatible APIs
  • Set up base URL and rate limits for each AI API integration
Permissions
  • Configure AI access by packages
  • Configure AI access to exclude specific groups or users
De-identification
  • Supported language models for name entities:
    • Requires installing Container Manager on the Synology NAS
    • Supported only on devices with 8 GB or more memory. The system checks the device's available memory size before enabling this feature
    • When multiple languages are detected in a prompt, the AI selects the dominant language and uses its corresponding de-identification model
    • Supported languages:
      • Chinese, Danish, Dutch, English, French, German, Italian, Japanese, Korean, Norwegian, Polish, Portuguese, Russian, Spanish, Swedish
      • Each language pack uses approximately 1 GB of memory
    • Supported predefined types:
      • No package dependencies or memory requirements
      • Global: Credit card number, Crypto wallet address, Date, Email address, International Bank Account Number (IBAN), IP address, URL, account token, and more
      • Region-specific: Passport number, Tax identification number, Driver’s license number, Medicare number, and more
        • Supported regions: Australia, Austria, Belgium, Bulgaria, Chile, China, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, India, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Norway, Poland, Portugal, Romania, Singapore, Slovakia, Slovenia, Spain, Sweden, Taiwan, United Kingdom, United States
      • Info types may vary by Synology AI Console version
Auditing
  • Transaction logs record details about AI requests, including send time, IP address, user, applied packages, applied API integration, AI provider, used model, action, input/output tokens, total tokens used, status
  • Admin logs track configuration changes made by administrators, including time, user, IP address, type, event
  • Log management enables administrators to maintain further control over logs:
    • Supports log retention policies for efficient storage management
    • Provides the option to log detailed inputs and outputs of AI requests for enhanced data control

Limitations

  • De-identification relies on AI models or regular expressions and cannot guarantee full masking of sensitive data
  • Accuracy of AI-generated information decreases as more words are de-identified