Orin FAQs

Account data and model training¶

No. Your data, including metadata and aggregated data, is only available to answer your own requests to Orin.

Disabling AI features¶

Yes. Your system administrator can toggle the feature switch in Application Settings to disable Orin.

Admin visibility into AI usage¶

No. System administrators don't have visibility of AI usage.

Third-party models used¶

Orin uses Azure AI OpenAI. The models currently used are gpt-4o-mini, text-3-embedding-large, and text-embedding-ada-002, but these may change as testing, performance, and processing requirements dictate.

Where and how Orin processes data¶

Orin uses a Large Language Model (LLM) to understand your requests and generate responses.

When using the LLM, Orin sends your message plus instructions on how to process it. Depending on the message, Orin may also send some information about your system to provide accurate, context-aware answers.

Orin uses two vector databases:

• Apteco software information: contains documentation about Apteco software, prepared by Apteco during Orin's base training
• Your system metadata: contains metadata from your Apteco system to provide dataset-specific information

Where Orin stores your data¶

Your vector databases remain within your Apteco system environment (Apteco Cloud, Partner Data Centre, or on-premises). Orin uses the Apteco engine to prepare and manage metadata and aggregated results. This ensures Personally Identifiable Information (PII) stays within your Apteco environment and isn't transferred to the LLM.

What data Orin sends to the language model¶

LLM processing occurs in the EU Data Boundary through Azure OpenAI Data Zones.

System metadata¶

The LLM receives:

• Variable names and descriptions
• Table structure and relationships
• Aggregated data (counts, sums, averages)
• Dashboard configuration metadata

User-controlled content¶

Orin sends no PII to the LLM, unless you explicitly include it in:

• Your request text
• Audience names or dashboard tile titles
• Dashboard cube or chart tile dimensions (for example, using email addresses as chart labels)

It would be odd for a user to do this. But if you include PII in these ways, Orin sends that data to the LLM as part of your request.

Data flows between Orin and Microsoft Azure within the EU data boundary¶

%%{init: {'themeVariables': {'fontSize': '20px'}, 'flowchart': {'padding': 20}}}%%
flowchart LR

subgraph Client
    Browser["Web Browser\nOrin Chat UI"]
end

subgraph AptecoCloud["Apteco Cloud or Client Network"]
    subgraph "APIs"
        OrbitAPI["Orbit API"]
        ChatAPI["Orin Chat API"]
    end
    JobQueue["Job Queue"]
    subgraph "FastStats Service"
        MySystem["My system"]
    end
end

subgraph AzureAI["Microsoft Azure (EU Data Boundary)"]
    AIProxy["AI Proxy"]
    OpenAI["OpenAI LLM\n(EU Data Zone)"]
    Licensing["Licensing & AI info"]
end

Browser --> OrbitAPI
Browser --> ChatAPI
OrbitAPI --> ChatAPI
ChatAPI --> OrbitAPI
OrbitAPI --> JobQueue
JobQueue --> MySystem
ChatAPI <--> AIProxy
AIProxy --> OpenAI
OpenAI --> AIProxy
AIProxy --> Licensing

Processing by deployment type¶