Sovereign Cloud & Air-Gapped Deployment

MITHUNAI can be deployed as a completely self-contained appliance within your virtual private cloud (VPC), sovereign government cloud, or air-gapped datacenter without external dependencies. Every component—the API gateway, background ingestion workers, pgvector embedding storage, and Redis cache—runs on your infrastructure, ensuring zero telemetry or customer data leaves your security boundary.

Deployment Topologies

MITHUNAI supports three production deployment topologies:

Managed Multi-Tenant

Fully managed SaaS hosted in ISO 27001 and SOC 2 Type II certified regions with tenant isolation.

Dedicated Single-Tenant VPC

Isolated cloud instances in AWS, GCP, or Azure with customer-managed encryption keys (CMEK) and dedicated subnets.

Self-Hosted Air-Gapped

Containerized distribution running entirely offline on your hardware with no internet access required.

Docker Compose Quickstart

For developer evaluations and isolated single-node production environments, deploy using Docker Compose:

Start Services

# Clone the repository and configure sovereign environment variables
git clone https://github.com/mithunaihq/mithunai.git
cd mithunai/docker

# Copy environment template
cp env.example .env

# Start all core services in background
docker compose up -d

Core Service Ports

Service	Port	Description
`nginx`	`80` / `443`	Reverse proxy and SSL termination.
`api`	`5001`	Backend HTTP API service (`/arukz/api/v1`).
`db`	`5432`	PostgreSQL with `pgvector` extension enabled.
`redis`	`6379`	In-memory cache and Celery broker.
`web`	`3000`	Administrative console and analytics dashboard.

Air-Gapped Compliance & Zero Telemetry

MITHUNAI enforces strict offline compliance guarantees for regulated healthcare, financial, and defense installations:

Zero Unauthorized Phone-Home: The platform performs no external license verification pings, usage telemetry beacons, or third-party analytics calls.
Local Embedding Support: Connect local embedding servers (e.g., Hugging Face Text Embeddings Inference running BAAI/bge-large-en-v1.5 or nomic-embed-text) to eliminate external embedding APIs.
Local LLM Execution: Route inference queries to private local engines such as vLLM, Triton Inference Server, or Ollama hosting open-weights models (Llama-3.3-70B, DeepSeek-R1, Qwen-2.5-Coder).
Static Assets Bundled: All documentation, fonts, icons, and JavaScript widgets are self-contained without CDN links.

Get started

Knowledge

Assistants & Agents

Channels & Integration

Architecture & Solutions

Administration

Trust and support

Sovereign Cloud & Air-Gapped Deployment

Deployment Topologies

Managed Multi-Tenant

Dedicated Single-Tenant VPC

Self-Hosted Air-Gapped

Docker Compose Quickstart

Core Service Ports

Air-Gapped Compliance & Zero Telemetry

​Deployment Topologies

Managed Multi-Tenant

Dedicated Single-Tenant VPC

Self-Hosted Air-Gapped

​Docker Compose Quickstart

​Core Service Ports

​Air-Gapped Compliance & Zero Telemetry

Deployment Topologies

Docker Compose Quickstart

Core Service Ports

Air-Gapped Compliance & Zero Telemetry