|
<< Click to Display Table of Contents >> Raynet One > 2026.2 > User Guide > First impressions > Recurring areas and elements > AI Chat Technical Architecture |
This section describes the internal technical architecture of the AI Chat feature, including its core components, communication flow, external service integrations, and security principles.
AI Chat uses a fully server-side storage model. All chats and messages are persisted in the backend database. This enables persistent context within each chat and supports multiple parallel chats per user. There is no dependency on local browser storage.
AI Chat employs WebSockets to deliver real-time, incremental responses. User messages are processed on the server, which may query internal data sources as needed.
When you send a message following workflow will be performed:
1.Your message is sent to the AI service.
2.The assistant processes your request and may query the database if needed.
3.You will see status updates showing what the assistant is working on.
4.The response streams back to you in real-time.
All communication with external AI providers is handled exclusively server-side through a dedicated backend layer. LiteLLM acts as an abstraction layer between Raynet One and external AI providers.
•Abstraction via LiteLLM: LiteLLM provides a uniform interface for different AI providers and models (e.g., OpenAI, Azure). It enables switching providers or models without changes to application code and centralizes provider communication.
•Secure Communication Flow: The frontend never communicates directly with any AI provider. All requests follow the strict path: Frontend > Backend > LiteLLM > AI Provider.
This design guarantees security, maintainability, and provider independence.
Voice capabilities are an optional server-side extension of the AI Chat feature. The voice layer integrates fully into the backend-controlled communication flow and follows the same security principles and permission boundaries as text-based interactions.
The layer consists of two independent capabilities:
•Speech-to-Text (STT): Converts spoken user input into text via an external speech service.
•Text-to-Speech (TTS): Converts assistant responses into spoken output via an external speech service.
All voice processing runs server-side. Voice handling does not rely on browser-native speech APIs, and a custom TTS provider is used instead of native browser APIs such as speechSynthesis. Voice interaction supports a controlled start-and-stop flow during use.
Voice features are inactive by default and become available only after the corresponding speech services have been configured.
This section describes the external services that provide AI-supported and enriched data processing capabilities within the License Management context. These integrations operate independently and complement the AI Chat feature.
SAMCloud
SAMCloud provides AI-enhanced document intelligence for License Management workflows. It parses uploaded files such as invoices and license contracts, extracts structured core information, and provides access to additional knowledge, enrichment services, and AI-supported analysis capabilities.
In the License Management context, AI Chat can use processed SAMCloud results to provide document-based license insights, such as license contract summaries, license balances, and optimization opportunities.
Technology Catalog
The Technology Catalog provides structured, normalized, and enriched product and technology metadata. It supplies enriched vendor, version, and category information used for consistent asset and license data handling.
In the License Management context, AI Chat can use Technology Catalog data for consistent product identification and higher-quality product- and license-related responses.
The Technology Catalog also serves as a data foundation for AI-supported analysis and enhances the quality and consistency of license- and product-related insights.
Dependency and Availability Model
In the License Management context, the AI Chat backend can query the Technology Catalog for normalized product data and SAMCloud for parsed document results. These services enrich the information available to AI Chat conversations. AI Chat continues to function for general natural language interaction even if one or both supporting services are unavailable. In License Management, these services work together to enrich the available data and analysis capabilities, but none of them is a direct technical prerequisite for the others.
Architectural Summary
For enriched AI-supported License Management capabilities:
•SAMCloud enables document intelligence and AI-enhanced parsing.
•The Technology Catalog provides structured product intelligence.
•AI Chat enables conversational interaction and AI-assisted analysis.
Together, these services form the AI supported License Management ecosystem.