Swifbit documentation

This documentation covers everything you need to install, configure, deploy, and administer Swifbit. If you are evaluating for the first time, start with the Quickstart.

What Swifbit does

• Ask — chat with your organizational memory. Every answer cites its sources and carries a confidence rating. No evidence means no invented answer.
• Assist — paste an incoming message and get a reply drafted only from approved knowledge, with markers for anything that still needs confirmation.
• Knowledge — every imported document is visible, editable, versioned, and gated by review states.
• Connectors — versioned packages that sync helpdesks, wikis, drives, and email into the knowledge layer.
• Administration — teams and access control, member lifecycle, audit trails, and an organization-wide control center.

Design principles

• Evidence before eloquence — retrieval is limited to approved, accessible knowledge; missing evidence becomes a visible gap rather than a plausible invention.
• Permissions inside retrieval — team-based source access is enforced in search, citations, and AI context, not just hidden in the UI.
• Self-hosted and model-agnostic — your infrastructure, your models, your data. Runs with local Ollama, any OpenAI-compatible endpoint, or Vertex AI.

Licensing & access

Swifbit is commercial software licensed per organization. The source repository and release bundles are private and provided to licensed customers and evaluation teams, and deployments are activated with a license key issued at purchase. To start an evaluation, request a demo; for purchasing and repository access, use the contact form or email [email protected]. Full details in Access & licensing.