Skip to content

Repository files navigation

Zenith Relay

Zenith Relay

Personal ChatGPT accounts and compatible API sources behind one private endpoint.

English documentation · Русская документация

Download latest release · Changelog · AGPL-3.0-only

Documentation

Choose a language first. Each overview links directly to a separate guide, so the Help Center and GitHub navigation do not depend on one long page.

Mode English Русский
Overview Read Открыть
This computer Guide Инструкция
Choose API Guide Инструкция
My server Guide Инструкция

What Is Shipped

  • Local-first Tauri desktop app with a React/Vite UI.
  • ChatGPT OAuth, existing-profile import, and compatible API sources.
  • Local personal pool with quota/health checks, model rules, proxies, routing, response affinity, and redacted usage history.
  • Provider-neutral source discovery with explicit protocol bindings, confirmed reasoning capabilities, API-reported model prices, and optional per-source price overrides.
  • Usage diagnostics that distinguish Relay, account, and provider failures without recording prompts, response bodies, or secrets.
  • Account views keep provider-reported quota windows separate from direct token-based API-equivalent and optional purchase-cost payback; Relay does not turn a quota percentage into a monetary entitlement.
  • Optional user-managed Relay Server with encrypted vault, SQLite state, management API, managed ChatGPT/Codex profile attachment, backup/restore, and append-only migrations.
  • Live model catalogs and reversible ChatGPT/Codex profile attachment with automatic snapshots.
  • Signed in-app updates, including in-place replacement and rollback for the Windows portable EXE.

Relay is a personal deployment. It is not Zenith customer billing, a public account marketplace, or the internal Zenith account pool.

Current Direction

The next work prioritizes reliable, provider-neutral operation over adding vendor-specific shortcuts: prove the existing personal-pool and server paths with real permitted accounts, measure user-visible latency, and keep model, price, and error behavior covered by regression tests. New account connectors, client integrations, and multi-server scale remain demand-gated. The exact acceptance gates and their order are in ROADMAP.md.

Screenshots

Overview Connections

Pool Usage

Development

cd src
bun install
bun run verify
bun run test:e2e

For the desktop bundle use bun run app:build. Shared runtime and server checks are listed in CONTRIBUTING.md.

Current implementation boundaries are in PLANNING.md; unfinished acceptance work is in ROADMAP.md. Release history is in CHANGELOG.md.

About

Personal ChatGPT accounts and compatible API sources behind one private endpoint.

Topics

Resources

Contributing

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages