Bay Run
Bay Run
Bay Run 是一个发布时免费、兼容 OpenAI 的 REST 和 MCP 服务,用于在官方源 https://run.huggingbay.xyz 上进行小模型推理。
免费发布:30 秒
发布层级是免费的。使用表单编码的 client_credentials 请求生成一个演示 bearer 令牌,然后调用 API。公布的护栏为 60 次请求/分钟和 1,500 次/天;演示 bearer 令牌在 24 小时后过期。
从在线、可复制粘贴的快速入门开始:https://run.huggingbay.xyz/quickstart。
该仓库的独立演示是一个小型 classify → 签名回执 → verify 流程:
python examples/bakeoff.pyimport json
from urllib.parse import urlencode
from urllib.request import Request, urlopen
BASE = "https://run.huggingbay.xyz"
def post(path, body, headers):
request = Request(BASE + path, data=json.dumps(body).encode(), headers=headers)
return json.load(urlopen(request))
token_request = Request(
BASE + "/oauth/token",
data=urlencode({"grant_type": "client_credentials"}).encode(),
headers={"Content-Type": "application/x-www-form-urlencoded"},
)
token = json.load(urlopen(token_request))["access_token"]
headers = {"Authorization": f"Bearer {token}", "Content-Type": "application/json"}
result = post("/v1/classify", {"input": "The setup was quick and clear.", "model": "auto"}, headers)
receipt = result["provenance_receipt"]
verified = post("/v1/provenance/verify", {"receipt": receipt}, headers)
assert verified.get("valid") is True or verified.get("receipt_valid") is True
print(result["labels"][0])
print("receipt verified")此演示使用真实的 /v1/classify 和来源验证接口;它不会虚构一个 bake-off 端点。
Related MCP server: universal-ai-agent-bridge
REST 和 SDK
REST 接口包括兼容 OpenAI 的 chat/completions(包括 SSE 流式传输和工具)、嵌入、重排序、分类、摘要、RAG、计算、JSON 验证和记忆。在构建/测试期间,轻量 SDK 位于此 checkout 旁边:
pip install -e ../bayrun-sdk/python
npm install ../bayrun-sdk/jsfrom bayrun import Client
c = Client()
answer = c.classify("A helpful result.")
stream = c.chat([{"role": "user", "content": "Say hello."}], stream=True)import { Client } from "bayrun";
const c = new Client();
const answer = await c.classify("A helpful result.");
const stream = await c.chat([{ role: "user", content: "Say hello." }], { stream: true });持久化的 remember/recall/forget 记忆由所有者委派,且未授权给共享演示 bearer 令牌。请使用特定用途的开发者密钥,或使用通过 token_endpoint_auth_method=client_secret_post 注册的客户端。
适用于 OpenAI Python、Vercel AI SDK、LangChain 和 LlamaIndex 的适配器代码片段位于 docs/integrations.md。
MCP:正好三个工具
将 MCP 客户端指向 https://run.huggingbay.xyz/mcp/。在线服务器恰好暴露以下工具:
get_task_quoterun_taskverify_result
令牌和请求格式的详细信息保留在在线快速入门中,以便复制粘贴的契约保持最新。
信任与数据政策
REST 结果带有签名来源回执,任务执行会返回一个签名回执,可通过验证接口进行检查。请阅读 https://run.huggingbay.xyz/quickstart 上的回执语义和在线验证请求。
在发送数据之前,请阅读官方政策:https://run.huggingbay.xyz/.well-known/data-policy.json。已发布的政策表明 Bay Run 不会使用输入进行训练。
Hugging Face 即插即用镜像目录
Hugging Bay 的镜像目录可在 https://huggingbay.xyz 获取。对于支持 HF_ENDPOINT 的 Hugging Face Hub 客户端,请将客户端指向镜像源:
export HF_ENDPOINT=https://huggingbay.xyz该目录目前列出 893+ 个镜像模型。在依赖特定模型或仓库路径之前,请检查镜像的当前服务状态。
This server cannot be deployed
Maintenance
Related MCP Connectors
One MCP endpoint for Claude, GPT & Gemini: 100+ tools + no-code connectors + agent workers.
MCP server for building and testing AI agents with multi-model experimentation and insights.
The OpenRouter for tools. One MCP connection gives any AI agent 254 hosted tools, pay per call.
Chat, image, video, and audio models from many vendors behind one remote MCP server and one key.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceMCP server that enables agents to dynamically switch between multiple AI models (OpenAI, Anthropic, Google, etc.) with unified protocol-driven configuration and capability discovery.Apache 2.0
- AlicenseAqualityBmaintenanceEnables any AI agent to connect to any OpenAI-compatible LLM provider through a unified MCP interface, with automatic failover, caching, cost tracking, and multi-model consensus.13MIT
- AlicenseNot gradedqualityAmaintenanceEnables MCP-compatible applications to manage, run, chat with, and generate embeddings from local or cloud LLM models, and to perform web search and content fetching via cloud APIs.897 npmAGPL 3.0
- AlicenseNot gradedqualityAmaintenanceEnables MCP-compatible applications to manage and use local or cloud-hosted models, including listing, pulling, generating, chatting, embeddings, and web search/fetch through a unified tool interface.897 npmAGPL 3.0