Umrani

Umrani is a local multi-provider AI relay, model speed tester, automatic failover router, and analytics dashboard. It exposes an OpenAI-compatible local endpoint while routing requests to configured OpenAI-compatible, Anthropic, Gemini, Cohere, or Azure endpoints.

Requirements

• Node.js 18 or newer
• Python 3.10 or newer
• Python requests package

Install the Python dependency once:

python -m pip install "requests>=2.31,<3"

Run from npm

npm install --global umrani-ai-router
umrani

Or run without a global installation:

npx umrani-ai-router

Umrani opens the local dashboard at http://127.0.0.1:8080. The OpenAI-compatible relay uses http://127.0.0.1:8787/v1 by default.

Options:

umrani --no-browser
umrani --port 8090

The dashboard binds to loopback by default. A non-loopback bind is rejected unless --allow-remote is supplied explicitly. Remote exposure is not recommended because the control dashboard has no authentication layer.

Privacy and configuration

Provider API keys remain on the local machine and are protected by the existing per-user ConfigStore. Browser API responses mask saved keys. Never include .env, secret.key, config.json, or other credentials in a published npm package.

Instant failover limits

Umrani has no relay retry backoff. After a failure is detected, hot in-process route selection and handoff are designed to complete in under 1 ms. DNS, TCP/TLS setup, and the time an upstream provider takes to return an error are network-bound and cannot be guaranteed at 1 ms.

Publishing

Before publishing, verify the package name and contents:

npm login
npm whoami
npm view umrani-ai-router
npm run check
npm pack --dry-run
npm publish --access public

If the package name is already taken, change name in package.json to a scoped name such as @your-username/umrani, then publish with npm publish --access public.