Synology AI Console 1.2
In April, Synology announced and launched its highly anticipated AI solution for DSM-compatible NAS models. Although it was released later than expected, this delay now appears intentional, as Synology aimed to develop a centralized AI management platform rather than integrating third-party LLM support into individual applications.
Luka Manestar
This is the reason behind the creation of the Synology AI Console. With its initial integration into MailPlus and subsequent inclusion in the Office package, future updates—such as version 1.2 and later 2.0—will feature more productivity applications that will support both third-party and self-hosted LLMs as part of their development roadmap.
Luka Manestar
Luka Manestar
Synology AI Console 1.2 new features
With the announcement of version 1.2 in August, it is now available through the beta channel, introducing several new features. These include more flexible AI deployment, allowing the use of any OpenAI-compatible LLM via its API, along with custom de-identification capabilities.
Configuring a new API integration remains consistent with previous versions, with the key difference that users can now select any OpenAI-compatible provider, whether publicly hosted or locally deployed.
By entering the URL and optional API key, users can select any configured and accessible model offered by the provider. This enables multiple integrations from the same provider with different models, which can then be mapped to individual Synology applications as needed.
As before, once a provider is configured for a specific application, you can further customize which user groups are permitted or restricted from accessing the AI capabilities within that application. Additionally, this version introduces a second key feature: custom data de-identification.
Quoting the Synology definition, this means:
"Data de-identification allows the system to anonymize selected sensitive data, ensuring it is not transmitted to the AI providers and protecting user privacy."
In addition to prepopulated sensitivity settings, which can be classified accordingly, there is also the option to create custom sensitivity types tailored specifically to our needs.
Users will be able to specify the confidence level and define patterns with their primary and supporting elements, regardless of the language in use.
The primary detection element to be masked can be configured using Regex, keywords, or a function—such as a predefined Regex pattern that may include a validator. Additionally, keywords can be imported via a TXT or CSV file containing up to 2,048 individual words.
Once sensitive types are configured and selected, the new playground area can be used to test detection and de-identification results.
Each detected sensitive keyword will be replaced before being sent to the LLM and then restored to its original value once the query is executed.
The AI Console integration has so far been limited to MailPlus, Office, and the upcoming ChatPlus and Meet platforms. Although Drive has undergone its fourth update, it still lacks integration with the AI Console. This feature is planned for the next 4.1 version of Drive, offering some interesting options currently in development and testing. Further details will be shared in upcoming articles in Q1 2026.
