Skip to content

deepsikha-dash/AegisAI

 
 

AegisAI

Open-source AI Governance, Risk & Compliance (AI-GRC) Platform

License: AGPL-3.0 Python FastAPI React PRs Welcome

Getting Started · Architecture · API Reference · Guard Module · RAG Module · Report a Bug


What is AegisAI?

Every company shipping AI in Europe now faces legal obligations under the EU AI Act (in force April 2026). Most compliance tools cost thousands per month and are closed-source.

AegisAI is the open-source alternative — a full-stack platform that combines three things into one:

Module What it does
Compliance Engine Register AI systems, classify EU AI Act risk (Minimal / Limited / High / Unacceptable), generate required documentation
LLM Guard Real-time prompt injection detection using regex + DistilBERT/DeBERTa ML classifier — protect your LLM APIs
RAG Intelligence Ask natural language questions about EU AI Act, GDPR, ISO 42001 — grounded answers from regulatory source docs

Tech Stack

Layer Technology
Frontend React 18, TypeScript, Vite, Tailwind CSS
Backend Python 3.11, FastAPI, SQLAlchemy, PostgreSQL
ML (Guard) PyTorch, HuggingFace Transformers (DeBERTa-v3), scikit-learn
RAG LangChain, FAISS, OpenAI Embeddings
MLOps MLflow, Prometheus metrics
Infra Docker, Kubernetes (HPA configs included)
Auth JWT, bcrypt
Payments Stripe (optional)

Quick Start

Option 1 — Docker (recommended)

git clone https://github.com/SdSarthak/AegisAI.git
cd AegisAI

cp backend/.env.example backend/.env
# Edit backend/.env — add your GEMINI_API_KEY and/or OPENAI_API_KEY

docker compose up -d

Option 2 — Manual

# Backend
cd backend
python -m venv venv && source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env   # fill in values
uvicorn app.main:app --reload

# Frontend (new terminal)
cd frontend
npm install
npm run dev

Project Structure

AegisAI/
├── backend/
│   ├── app/
│   │   ├── api/v1/          # REST endpoints (auth, ai_systems, guard, rag, ...)
│   │   ├── core/            # Config, DB, JWT security
│   │   ├── models/          # SQLAlchemy ORM models
│   │   ├── schemas/         # Pydantic request/response schemas
│   │   └── modules/
│   │       ├── guard/       # LLM Guard — regex + ML classifier + sanitizer
│   │       ├── rag/         # RAG — vector store, retrieval chain, MLflow
│   │       └── llm/         # LLM client (OpenAI-compatible)
│   ├── data/                # Training data for Guard classifier
│   └── tests/
├── frontend/                # React + TypeScript dashboard
├── infra/                   # Kubernetes deployment & HPA configs
├── notebooks/               # Jupyter — train Guard classifier on GPU (Colab-ready)
├── docs/                    # Architecture, API reference, module guides
└── docker-compose.yml

Roadmap

  • EU AI Act risk classification engine
  • AI system registry + compliance dashboard
  • Compliance document generation (Technical Docs, Risk Assessment, Conformity Declaration)
  • LLM Guard — regex filter + ML intent classifier + sanitizer
  • RAG query endpoint (plug in your regulatory documents)
  • Pre-loaded regulatory knowledge base (EU AI Act, GDPR, ISO 42001, NIST AI RMF)
  • Audit log for all Guard scan decisions
  • Stripe billing integration
  • OAuth2 / SSO support
  • Multi-regulation support (UK AI Bill, India DPDP)
  • Analytics dashboard (compliance score over time)
  • Slack / webhook notifications for compliance drift

These open items are great places to contribute — see CONTRIBUTING.md.


Contributing

We welcome contributions of all kinds — code, docs, tests, regulatory expertise.

See CONTRIBUTING.md for the full guide.

Not sure where to start? Browse issues labelled:


License

AegisAI is licensed under AGPL-3.0-only.

  • Free for open-source and self-hosted use.
  • If you run a modified version as a SaaS, you must release your source code.
  • For commercial licensing, contact the author.

Copyright (C) 2024 Sarthak Doshi (@SdSarthak)


Built with care. If AegisAI helps you, give it a star.

About

Open-source AI Governance, Risk & Compliance (AI-GRC) platform — EU AI Act compliance, LLM Guard, and RAG regulatory intelligence.

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • Python 53.4%
  • TypeScript 32.7%
  • Jupyter Notebook 13.1%
  • JavaScript 0.4%
  • Dockerfile 0.2%
  • HTML 0.2%