Compare commits

...

1 Commits

Author SHA1 Message Date
caelum0x
2f84b3efa9 feat: add Umpirix to Runtimes & Platforms
Self-hostable, Apache-2.0 typed-decision server (choice/score/noul with
probabilities). Wire-compatible with TypeSafe's System One API: the
official typesafe-sdk and @typesafe-ai/sdk work unmodified against it.
Runs on CPU with no API key or database to boot; ships its own MCP
server and a no-server CLI classifier. Formerly OpenJev.

Added to both README.md and README_zh.md, same section and position
(after Jev Agent), per CONTRIBUTING.md. Verified with
`python3 scripts/build_site.py --strict` (0 sync problems).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-11-12 01:08:10 +03:00
2 changed files with 2 additions and 0 deletions

View File

@ -131,6 +131,7 @@ Servers, gateways, and native runtimes for running decision models.
- [OpenRouter decision models](https://openrouter.ai/models?output_modalities=decisions) - Decision models from several publishers behind OpenRouter's alpha Decisions API, which chat-completions SDKs cannot call. Usage guide as an agent skill: [openrouter-decisions](https://openrouter.ai/skills/openrouter-decisions).
- [Vercel AI Gateway](https://vercel.com/ai-gateway/models/jev) - Hosts Jev as `typesafe-ai/jev`
- [Jev Agent](https://jev-agent.com/api-access) - Unofficial hosted gateway for the Jev model API and 12 other decision models: `/api/v1/systemone` mirrors TypeSafe's request shape, `/api/v1/decisions` takes OpenAI's Decisions API request, and `/api/v1/classify` runs eight published Jev rulesets; also an MCP server and a no-key playground, with 5 free credits a month. Independent site, not affiliated with TypeSafe AI.
- [Umpirix](https://github.com/caelum0x/umpirix) - Self-hostable, Apache-2.0 typed-decision server with its own in-process engine (CPU-only int8 ONNX embedding classifier with a lexical fallback, no API key or database to boot) behind `/v1/systemone`. Wire-compatible with TypeSafe's System One API: the official `typesafe-sdk` and `@typesafe-ai/sdk` work unmodified against it via `TYPESAFE_BASE_URL`. Also ships an MCP server (choice/score/noul/decide) and a no-server CLI classifier. Formerly named OpenJev; independent, not affiliated with TypeSafe AI.
- [stuntd](https://github.com/bladedevoff/stuntd) - Local proxy on the open Laya model that speaks the Jev System One API, records the app's Choice, Score and Noul answers from a Jev upstream, trains a per-question head, and serves it with a calibrated confidence threshold and fallback to the upstream
- [Bud Decision Studio](https://github.com/BudEcosystem/Bud-Decision-Studio) - Open-source, cross-platform desktop runtime and playground from Bud Ecosystem for running, evaluating, managing, and training Jev-like decision models locally, with a `/v1/systemone` API. Supports 10+ decision models.

View File

@ -131,6 +131,7 @@
- [OpenRouter 决策模型](https://openrouter.ai/models?output_modalities=decisions) - OpenRouter alpha 版 Decisions API 上多家发布方的决策模型;chat completions SDK 无法调用该 API。使用指南以 Agent skill 形式提供:[openrouter-decisions](https://openrouter.ai/skills/openrouter-decisions)。
- [Vercel AI Gateway](https://vercel.com/ai-gateway/models/jev) - 以 `typesafe-ai/jev` 托管 Jev
- [Jev Agent](https://jev-agent.com/api-access) - 非官方托管网关,提供 Jev 模型 API 及另外 12 个决策模型:`/api/v1/systemone` 与 TypeSafe 官方请求格式一致,`/api/v1/decisions` 接受 OpenAI Decisions API 请求,`/api/v1/classify` 运行 8 套公开的 Jev 规则集;另有 MCP 服务器和免密钥 playground,每月 5 个免费额度。独立网站,与 TypeSafe AI 无关联。
- [Umpirix](https://github.com/caelum0x/umpirix) - 可自托管的 Apache-2.0 类型化决策服务端,内置自有推理引擎(CPU 专用 int8 ONNX embedding 分类器,附词法回退,启动无需 API key 或数据库),对外提供 `/v1/systemone`。与 TypeSafe 的 System One API 线级兼容:官方 `typesafe-sdk` 和 `@typesafe-ai/sdk` 只需设置 `TYPESAFE_BASE_URL` 即可无修改连接。另提供 MCP 服务器(choice/score/noul/decide)和无需服务端的 CLI 分类器。曾用名 OpenJev;独立项目,与 TypeSafe AI 无关联。
- [stuntd](https://github.com/bladedevoff/stuntd) - 基于开放 Laya 模型的本地代理,实现 Jev System One API;记录来自 Jev 上游的 Choice、Score 和 Noul 答案,为每个问题训练一个决策头,并以校准过的置信度阈值提供服务,低于阈值时回退到上游
- [Bud Decision Studio](https://github.com/BudEcosystem/Bud-Decision-Studio) - 来自 Bud Ecosystem 的开源跨平台桌面运行时与 Playground,可在本地运行、评测、管理和训练 Jev 风格决策模型,并提供 `/v1/systemone` API。支持 10+ 个决策模型。