selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

blindai logo

Self-host blindai

v0.6.3

AI privacy solution with hardware-enforced secure enclaves

BetaOpen SourceApache-2.0
Updated2 years ago
LanguageRust
Setup difficultyAdvanced
Stars
513
CPU
Intel SGX+FLC capable CPU (SGX2 recommended)
Ports
9923 (unattested server port (HTTP) for initial client verification), 9924 (attested server port for verified enclave communication)
Requires
Docker, Intel PCCS (Provisioning Certification Caching Service), Intel SGX AESM service, Intel SGX drivers
Docker
mithrilsecuritysas/blindai-server
Install with Python
blindai screenshot 1

What is blindai?

BlindAI is an open-source AI privacy solution from Mithril Security that enables confidential AI inference using hardware-enforced Trusted Execution Environments (Intel SGX and AWS Nitro Enclaves). It has two components: BlindAI API, a Python client for querying popular models hosted by Mithril Security, and BlindAI Core, which lets you self-host a server to deploy your own models securely. The server runs inside an SGX enclave, ensuring that neither the hosting provider nor any third party can access user data or model weights. BlindAI Core can be deployed on Azure DCsv3 VMs or on-premise Intel SGX+FLC hardware via Docker. The project provides a Python client SDK (pip install blindai) and a Rust-based server. Note: the README warns that BlindAI is not actively maintained and should not be used for sensitive data without contacting the maintainers.

Should you self-host blindai?

blindai scores 44/100 on selfhost.directory (how scoring works) and ranks #80 of 125 in AI. It has 513 GitHub stars.

There was no stable release in the past 12 months, so check the repository's recent activity before you rely on it.

Setup is advanced: plan for several services and some ongoing maintenance. An official Docker image is available. It is Apache-2.0-licensed and written mostly in Rust.

Its closest alternative on selfhost.directory is ComfyUI (93/100, 143 714 stars). It scores higher overall, so compare the two before you pick. See blindai vs ComfyUI side by side.

Features

  • Confidential AI Inference
  • Independent Security AuditAudit does not cover client SDK, BlindAI API, or Nitro enclaves
  • Model Upload/Delete
  • Multiple Enclave Support
  • ONNX Model Support
  • Pre-hosted Models (BlindAI API)Project not actively maintained; hosted API may be unavailable
  • Python client SDK
  • Remote Attestation
  • Self-hosted Model Deployment

blindai README

How to install blindai

Install blindai with Python

pip install blindai

Client SDK for connecting to BlindAI API or self-hosted BlindAI Core

How to configure blindai

  • BLINDAI_AZURE_DCS3_PATCHEnable Azure DCsv3 VM patch for SGX compatibility

    Enable Azure DCsv3 VM patch for SGX compatibility

    values: 1

  • PCCS_API_KEYIntel Provisioning Certification Caching Service API key for SGX attestation

    Intel Provisioning Certification Caching Service API key for SGX attestation

blindai releases & version history

FAQ

Has BlindAI been security audited?

Yes, a beta version of BlindAI Core was independently audited by Quarkslab SAS, though the audit does not cover the client SDK, BlindAI API, or Nitro enclaves.

What ports does BlindAI use?

BlindAI uses port 9923 (unattested, HTTP) for initial client verification and port 9924 (attested) for verified enclave communication.

What is the difference between BlindAI API and BlindAI Core?

BlindAI API is a Python client for querying popular models hosted by Mithril Security. BlindAI Core lets you self-host a server to deploy your own models with full privacy guarantees.

More in Privacy

Rate this project

Click to rate — no account needed.

0 comments

Posting as guest
Loading comments…