Module architecture
Our AI Accelerator module aims to provide seamless AI interactions within the user interface while ensuring secure and efficient backend processing.
UI integration
The majority of the generative UI widgets are JS Field-based. This allows to extend the feature set based on our API layer.
Backend integration
The module contains a proxy layer that acts as a buffer between the backend and the AI service. It helps in maintaining separation of concerns and potentially offers caching and other optimizations.
Additionally, the layer ensures that calls to the AI service are correctly formulated and securely transmitted. Credentials to ensure authorized access are added on call from the frontend.
Credentials management
The proxy layer ensures that AI service credentials are securely managed, appended to AI calls when necessary, and regularly rotated for security purposes. See the configuration section for reference.
AI call flow
flowchart LR
A[UI] --->|User trigger | B[Backend]
B --->|Processes request| C(proxy) ---> |Appends creds| D[AI Service]
D --> E((Result))
E --->C--->B--->|Renders on frontend |A
- Initiation from UI
-
Users trigger AI operations through the Magnolia JS Field.
- Backend processing
-
The backend receives the request, processes it, and forwards it to the AI service through the proxy.
- Proxy interaction
-
The proxy appends necessary credentials and routes the call to the AI service.
- Response handling
-
Once the AI service returns a result, the data flows back through the proxy and backend, eventually rendering the result on the Magnolia JS Field UI.