Feature request: openai_compatible API_TYPE so any OpenAI-compatible endpoint works with no service class

未关闭
#369 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
3/5
预计耗时
1-2 天
新手友好度
58/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
活跃
技术栈
python
领域
backend

调研方向

Start by reading the custom-model path in the agents.yaml config and the model factory where service classes are registered (ufo/config and ufo/llm or similar module). Trace how API_BASE, API_KEY, and API_MODEL flow into an existing OpenAI-style service class, then add an openai_compatible API_TYPE that reuses it with no code required. Done means HOST_AGENT and APP_AGENT can point at any OpenAI-compatible endpoint purely via config.

由索引模型根据 Issue 内容生成。

描述

Hi — I run DemonRoute (demonroute.com), an OpenAI-compatible API at https://api.demonroute.com/v1.

I will be direct about why I am here: we are talking to a lot of projects, because there is no large established provider for no-subscription, anonymous API keys for open-weight uncensored models. We are earning distribution one repo at a time.

UFO is a case where the current path is close but not quite open. Today the custom-model route in agents.yaml takes API_BASE, API_KEY, and API_MODEL, but using it still means writing a Python service class and registering it in the model factory. That is a real barrier for anyone who just wants to point HOST_AGENT or APP_AGENT at an OpenAI-compatible endpoint.

The feature ask: add an openai_compatible API_TYPE that sends standard chat-completions requests to API_BASE with API_KEY and API_MODEL, no code required. That would make any OpenAI-compatible provider work out of the box, DemonRoute included.

Why it matters for UFO users: every desktop step is an LLM call with screenshots attached, so cost adds up fast, and your own docs flag cost optimization. We carry Dolphin, MythoMax, Hermes, and abliterated Llama/Qwen builds that handle adversarial automation tasks without refusing, and we never train on user content. Nothing on a user's screen is ever used to train a model.

Who is behind it: during the NFT peak I became the largest NFT artist on the third-largest chain in the world, overnight, with zero funding. That turned into consulting for Fortune 500s, celebrities, and emerging tech companies. I run everything under NFT Demon Holdings LLC, and DemonRoute is my passion priority.

Reciprocal: any app that lists DemonRoute as a provider option gets listed on demonroute.com/integrations.

Jeremy, DemonRoute

主要语言
Python
星标
10.2k
派生
1.1k
平均合并
1 分钟
30 天内合并 PR
1

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

microsoft/UFO 的其他 Issue

查看 microsoft/UFO 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。