How Delphina protects your data, controls access, and where it runs.Documentation Index
Fetch the complete documentation index at: https://docs.delphina.ai/llms.txt
Use this file to discover all available pages before exploring further.
Data handling
- Read-only access — the agent cannot modify your warehouse data.
- No persistent data storage — query results are stored in S3 within your deployment. The control plane stores only metadata.
- Credential security — warehouse credentials are encrypted at rest, decrypted only when establishing a connection.
- Workspace isolation — each workspace uses its own credentials. See Workspaces.
Access controls
- SSO authentication — Google Workspace, Okta, or other SAML/OIDC providers. No separate passwords.
- Role-based access — Explorer, Developer, Admin per workspace. See User Management.
- Organization boundaries — no cross-organization access.
Compliance
- SOC 2 Type II — audited controls for security, availability, and confidentiality.
- HIPAA — available for customers with healthcare data requirements.
- No model training — your data is never used to train AI models.
Deployment options
| Option | Description |
|---|---|
| Multi-tenant SaaS | Shared infrastructure with logical isolation. Most organizations. |
| Single-tenant | Dedicated AWS account + EKS cluster. Complete physical isolation. |
| VPC deployment | Runs within your own AWS account. Data never leaves your network. |
LLM usage
- Query context (schemas, metric definitions, results) is sent to the LLM to generate SQL and interpret results.
- No data is used for training — commercial API agreements prohibit it.
- Data is not persisted by LLM providers beyond the request lifetime.
- Available providers include Anthropic via AWS Bedrock.