Compare commits

..

No commits in common. "add-relay" and "main" have entirely different histories.

2 changed files with 0 additions and 2 deletions

View File

@ -351,7 +351,6 @@ _Pain point: "My keys, my infra, no per-token middleman fee."_
- [OmniRoute](https://github.com/diegosouzapw/OmniRoute) <!--s:diegosouzapw/OmniRoute-->⭐ 80.5k<!--/s--> - MIT self-hosted TypeScript gateway: one endpoint to 231+ providers (50+ free), plugging Claude Code / Codex / Cursor / Cline / Copilot into free Claude/GPT/Gemini with stacked token compression (15–95% savings), 17 routing strategies, smart auto-fallback and MCP/A2A. A 2026 breakout of the coding-agent "token-saver" wave — genuine code (not a relay farm), but its free/OAuth coding-tier routing carries provider-ToS/account-ban risk.
- [Chat Nio (CoAI)](https://github.com/coaidev/coai) <!--s:coaidev/coai-->⭐ 9.3k<!--/s--> - Multi-tenant "one-stop" gateway with a built-in admin + credit/subscription billing panel over 200+ models / 35+ providers, priority-based load balancing and model caching — the same commercial-panel genre as the new-api / one-api / VoAPI entries here.
- [TensorZero](https://github.com/tensorzero/tensorzero) <!--s:tensorzero/tensorzero-->⭐ 11.7k<!--/s--> - **Archived June 2026** ⚠️ (company wound down; repo read-only, Apache-2.0 code + community forks remain). Rust gateway unified with observability, evals, experimentation and optimization.
- [Relay](https://github.com/caelum0x/railyard/tree/main/apps/relay) <!--s:caelum0x/railyard-->⭐ 0<!--/s--> - Apache-2.0 maintained fork of TensorZero (archived June 2026): same `gateway` binary, `tensorzero.toml` config and `/inference`/`/openai/v1/*` endpoints, plus per-key budgets, a price catalog, fallback routing, an operator UI and prepaid success-only billing with AITP-402 pay-per-call challenges.
- [Bifrost](https://github.com/maximhq/bifrost) <!--s:maximhq/bifrost-->⭐ 9.1k<!--/s--> - Go gateway from Maxim AI claiming ~50x LiteLLM throughput; adaptive load balancing, cluster mode, MCP support, a built-in web UI, and documented Ollama support for local models — zero-config start via `npx` or Docker.
- [Otari](https://github.com/mozilla-ai/otari) <!--s:mozilla-ai/otari-->⭐ 733<!--/s--> - **Mozilla AI's** self-hosted gateway (Apache-2.0, Python): one OpenAI- and Anthropic-compatible endpoint over 40+ providers via [any-llm](https://github.com/mozilla-ai/any-llm), with virtual keys, per-user/per-key budgets enforced _before_ a request runs, and `/v1/usage` spend tracking; single-Docker start, optional hosted platform (otari.ai). Launched April 2026 and one of July's fastest risers.
- [Traceloop Hub](https://github.com/traceloop/hub) <!--s:traceloop/hub-->⭐ 231<!--/s--> - High-scale gateway written in Rust from the [Traceloop](https://www.traceloop.com/) team (OpenLLMetry / OTel-for-LLMs); OpenTelemetry-native observability built in.

View File

@ -330,7 +330,6 @@ _生态里被问得最多的问题之一,而网上的答案大多已过期。
- [OmniRoute](https://github.com/diegosouzapw/OmniRoute) <!--s:diegosouzapw/OmniRoute-->⭐ 80.5k<!--/s--> — MIT 自托管 TypeScript 网关:一个端点接入 231+ 厂商(50+ 免费),把 Claude Code / Codex / Cursor / Cline / Copilot 接到免费的 Claude/GPT/Gemini,叠加 token 压缩(省 15–95%)、17 种路由策略、智能自动回退与 MCP/A2A。2026 编码 Agent「省 token」浪潮的黑马——是真实代码(非中转农场),但其免费/OAuth 编码档路由存在厂商 ToS/封号风险。
- [Chat Nio (CoAI)](https://github.com/coaidev/coai) <!--s:coaidev/coai-->⭐ 9.3k<!--/s--> — 多租户「一站式」网关,内置管理后台 + 积分/订阅计费面板,聚合 200+ 模型 / 35+ 厂商,带优先级负载均衡与模型缓存——与本清单已收录的 new-api / one-api / VoAPI 属同一商业面板品类。
- [TensorZero](https://github.com/tensorzero/tensorzero) <!--s:tensorzero/tensorzero-->⭐ 11.7k<!--/s--> — ⚠️ **2026 年 6 月已归档**(公司关停;仓库只读,Apache-2.0 代码与社区分支尚存)。Rust 网关 + 可观测 + 评测 + 实验优化一体。
- [Relay](https://github.com/caelum0x/railyard/tree/main/apps/relay) <!--s:caelum0x/railyard-->⭐ 0<!--/s--> — TensorZero(2026 年 6 月已归档)的 Apache-2.0 维护分支:保留相同的 `gateway` 二进制、`tensorzero.toml` 配置和 `/inference`、`/openai/v1/*` 端点,新增按密钥预算、价目表、故障回退路由、运维 UI,以及预付费、仅成功计费与 AITP-402 按次付费挑战。
- [Bifrost](https://github.com/maximhq/bifrost) <!--s:maximhq/bifrost-->⭐ 9.1k<!--/s--> — Maxim AI 出品的 Go 网关,号称比 LiteLLM 快约 50 倍;自适应负载均衡、集群模式、支持 MCP,内置 Web 管理界面,官方文档支持本地 Ollama——`npx` 或 Docker 零配置起步。
- [Otari](https://github.com/mozilla-ai/otari) <!--s:mozilla-ai/otari-->⭐ 733<!--/s--> — **Mozilla AI** 出品的自托管网关(Apache-2.0,Python):一个端点同时兼容 OpenAI 与 Anthropic 协议,经 [any-llm](https://github.com/mozilla-ai/any-llm) 覆盖 40+ 厂商;虚拟 Key、按用户/按 Key 的预算在请求执行**前**拦截、`/v1/usage` 消费查询;单 Docker 起步,可选托管平台(otari.ai)。2026 年 4 月发布,7 月增速最快的新网关之一。
- [Traceloop Hub](https://github.com/traceloop/hub) <!--s:traceloop/hub-->⭐ 231<!--/s--> — [Traceloop](https://www.traceloop.com/) 团队(OpenLLMetry / LLM 版 OTel)出品的 Rust 高并发网关,内置 OpenTelemetry 原生可观测性。