Janus.Guru

Security starts with data boundaries and responsibility

Before a pilot, define where documents will be stored, which fragments a model may receive, who can access them, how long logs are kept, and who approves answers. Confirm those terms for the specific deployment.

What the product controls

  • a response uses one selected collection
  • document-based answers show their retrieved sources
  • API keys can be limited by collections and request quotas
  • API-key creation and revocation can be recorded in the audit log

What each deployment fixes

  • document and index storage
  • cloud, local, or hybrid model routing
  • data allowed into model context
  • retention and deletion rules
  • roles, access, and personal-data processing

What remains human

  • legal assessment
  • editorial publication
  • an official response
  • access and retention policy approval
  • primary-source review

Frequently asked questions

  • Does Janus.Guru send every document to an external model?
    Not necessarily. Retrieval selects relevant fragments; the model route and context policy are deployment decisions.
  • Does Janus.Guru guarantee that data is stored in Kazakhstan?
    No. That depends on the chosen infrastructure, providers, and deployment architecture. Confirm it for each project.
  • Is the public demo suitable for personal or restricted documents?
    No. The public demo uses open demonstration collections. Sensitive data need a separate deployment with agreed access and processing rules.