Transparent breakdown of enterprise data flows, model inference isolation, Indian data localization, sub-processors, and Technical & Organizational Measures (TOMs).
Primary database clusters, vector embeddings, and tenant assets are hosted within the sovereign territory of India, aligned with MeitY and DPDP guidelines.
Customer prompts, conversation history, and uploaded knowledge base documents are strictly segregated and NEVER used to train or refine public baseline AI models.
Multi-tenant architecture enforces PostgreSQL Row-Level Security (RLS) dynamically locked to app.current_tenant on every single database transaction.
For enterprise workspaces, the Customer acts as the Data Fiduciary determining the purposes and lawful bases of its team members' data. Xarwiz Technologies LLP acts as the Data Processor executing operations strictly on documented instructions from the Customer.
| Sub-Processor | Purpose | Location | Data Transferred |
|---|---|---|---|
| Google Cloud Platform (GCP) | Cloud VM & Database Hosting | India (asia-south2) | Encrypted application data & database volumes |
| Ollama Local Inference | On-Premise / Local LLM Inference | India (Self-Hosted) | Ephemeral prompt inputs (zero persistent storage) |
| Razorpay / Cashfree / PhonePe | Payment Processing & Invoicing | India | Billing customer name, email, order amount |
| Resend | Transactional Email Delivery | Cloud (Global) | Recipient email, verification & privacy alert links |