Skip to main content
Suiri is designed as an inference-native platform with privacy and data minimization as first-class principles.

​
What Suiri Does Not Store

  • No prompt storage: User prompts are not persisted after the inference request is completed
  • No response storage: Model outputs are returned directly to the caller and are not retained
  • No training on customer data: Customer prompts and responses are never used to train or fine-tune models
All inference requests are processed ephemerally in memory for the sole purpose of generating a response.

​
What Is Logged

To operate and secure the platform, Suiri retains minimal, non-content metadata, including:
  • Request timestamp
  • Model ID
  • Token counts (input, output, total)
  • API key identifier
  • Region and routing metadata
  • Error codes (if applicable)
This metadata is used exclusively for:
  • Billing and usage reporting
  • Platform reliability and debugging
  • Abuse prevention and rate limiting
Prompt text and model responses are never included in logs.

​
Data Residency & Routing

Suiri routes inference requests to the closest available edge datacenter that hosts the requested model, minimizing latency and reducing cross-region data movement. Enterprise deployments may request:
  • Region pinning
  • Dedicated capacity
  • Custom routing policies
Contact Suiri support to discuss enterprise data residency requirements.

​
Security Practices

  • All API traffic is encrypted in transit using TLS
  • API keys are required for all inference requests
  • Billing is enforced as a security control to prevent abuse