certifai/README.md

CERTifAI

This project is a SaaS application dashboard for administation of self-hosted private GenAI (generative AI) toolbox for companies and individuals. The purpose of the dashboard is to manage LLMs, Agents, MCP Servers and other GenAI related features.

The purpose of CERTifAIis to provide self-hosted or GDPR-Conform GenAI infrastructure to companies who do not wish to subscribe to non-EU cloud providers to protect their intellectual property from being used as training data.

Overview

The SaaS application dashboard is the landing page for the company admin to view, edit and manage the company internal GenAI tools. The following tasks can be performed by the administrator:

• User management: Can add, remove, set roles, permissions and add restrictions for other users.
• SSO/Oauth/LDAP: Can connect to company internal SSO/LDAP or other identity provider to load users and their respective permissions.
• Turn features on/off: Turn off/on different GenAI features
• Billing: View the current seats being used and token usage per seat for any given billing cycle
• Request support: Request support or new features using feedback form
• GenAI: View currently running LLMs, Agents, MCP Servers. Modify or add more resources, switch to a different model, launch tools like Langchain + Langfuse for creating new agents,tavily for internet search or more complex tools for use with GenAI. View endpoints and generate API Keys for integrations in other applications.

Development environment

This project is written in dioxus with fullstack and router features. MongoDB is used as a database for maintaining user state. Keycloak is used as identity provider for user management.

Features management

All features are detailed and described under the features folder in clear markdown instructions which are valid for both human and AI code developers.

Clean architecture

For the backend development, clean architecture is preferred. SOLID principles MUST be strictly followed. Clearly defined types, traits and their implementations MUST be used when generating new code. Individual files MUST be created if a file is exceeding more than 160 lines of code excluding any tests. The folder structure for clean architecure SHOULD BE as:

• service1/
  • Infrastructure/
  • Domain/
  • Application/
  • Presentation/
• service2/
  • Infrastructure/
  • Domain/
  • Application/
  • Presentation/ With each major service split in separate folders.

Git Workflow

We follow feature branch workflow for Git and bringing in new features. The main branch is the default and protected branch. Conventional commits MUST be used for writing commit messages. We follow semantic versioning as per SemVer

CI

The CI is run on gitea actions with runner tags docker.