OmniArk

AI Proxy Pool for Multiple Providers — Load balancing, credit tracking, and unified OpenAI-compatible routing for Codex, Qoder, and BYOK accounts.

OmniArk is a fork of the original priyo000/etteum-pool project. This repo re-brands and strips the codebase down to a pure Bun/TypeScript proxy pool — no Python auth bot, no browser automation, no VCC generator, and only the providers we actually use.

---

⚡ Quick Start

Prerequisite: Bun 1.x

git clone https://github.com/risunCode/OmniArk.git ~/omniark
cd ~/omniark
bun install.ts

Then start the server:

omniark start

Open the dashboard at http://localhost:12800.

---

Manual install

bun install
bun run build          # build dashboard
cp .env.example .env   # edit API_KEY / ENCRYPTION_KEY
bun src/index.ts       # start backend

---

What changed from the original?

Original project: priyo000/etteum-pool

This fork focuses on keeping the proxy pool core lean:

• Removed: Python dependencies, Playwright/Camoufox browser automation, the VCC generator, and the auth-bot orchestration.
• Removed providers: Canva, YouMind, CodeBuddy, CodeBuddy China, GitLab Duo.
• Kept providers: Codex, Qoder, plus BYOK (bring-your-own-key).
• Rebrand: all etteum / Etteum / poolprox3 / aiproxy references moved to OmniArk.

So if you’re looking for the original full-featured version with the auth bot and extra providers, head over to priyo000/etteum-pool. OmniArk is the trimmed-down, no-browser-fuss variant we run ourselves.

---

Environment

Variable	Default	Purpose
PORT	1930	Backend / OpenAI-compatible proxy port
DASHBOARD_PORT	1931	Dashboard static-server port
API_KEY	pool-proxy-secret-key	Bearer token for clients and dashboard
DATABASE_PATH	./data/omniark.db	SQLite database location
ENCRYPTION_KEY	random (installer)	32-char hex; encrypts saved tokens

The backend, dashboard, WebSocket endpoint, and OpenAI-compatible proxy are one service on the same PORT. There is no separate dashboard process or dashboard port in the deployment image. On Railway, use the public service domain without a port; Railway routes it to the service's runtime PORT.

Optional log-body toggles:

Variable	Default	Purpose
OMNIARK_LOG_BODY_ENABLED	true	Capture request/response bodies in logs
OMNIARK_LOG_BODY_FULL	true	Log full bodies instead of truncating
OMNIARK_LOG_BODY_REDACT	false	Redact sensitive fields in bodies
OMNIARK_LOG_BODY_MAX_BYTES	65536	Max body bytes to log

---

CLI commands

Command	Description
omniark start	Start backend + dashboard (dev)
omniark run	Alias for bun src/index.ts
omniark build	Build dashboard for production
omniark migrate	Run database migrations
omniark status	Check if the server is running
omniark stop	Stop the background instance
omniark restart	Stop + start
omniark logs	Tail the log file
omniark doctor	Print quick health diagnostics
omniark help	Show usage

---

Supported providers

• codex
• qoder
• byok (bring your own OpenAI-compatible key)

Add accounts through the dashboard at http://localhost:1931.

---

Client integration

The dashboard Integrations page can generate config snippets for popular AI coding clients (Codex CLI, Windsurf, Cline, etc.). The proxy exposes a standard OpenAI-compatible /v1/chat/completions endpoint.

---

Credits

• Original project: priyo000/etteum-pool
• Fork / rebrand: OmniArk — https://github.com/risunCode/OmniArk

License

MIT