GLOUTH LINK · API DOCS

一个 Base URL,
两种协议,所有模型。

Glouth Link 是双协议兼容的 LLM 网关。同一个 Base URL 同时接受 Chat Completions / Responses 和 Messages 两种协议;同一把 API Key 可以驱动主流的对话 SDK、消息 SDK、以及各类 AI 命令行工具等任何兼容客户端。

快速上手

三步:创建 Key → 设置 Base URL → 用熟悉的 SDK 调用。下面四种写法任选其一:

curl https://www.glouth.com/link/v1/chat/completions \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4-mini",
    "messages": [{"role":"user","content":"Hello"}]
  }'

认证

支持两种认证方式,任选其一:

方式一:Bearer Token(Authorization 头)

Authorization: Bearer sk-glouth-XXXXXXXXXXXXXXXXXX

方式二:x-api-key 头

x-api-key: sk-glouth-XXXXXXXXXXXXXXXXXX

注:Glouth Link 自动识别两种 header。若同时传,优先 Authorization。

Base URL 与协议

所有端点的 Base URL 是同一个:

https://www.glouth.com/link/v1

协议自动判定

• 请求带 anthropic-version 或 x-api-key 头 → Messages 协议
• 请求 body 含 system 字段、或 content block 数组 → Messages 协议
• 其他情况 → Chat Completions 协议

所有端点

下面列表是当前支持的端点。点击跳详细参数。

• POST /chat/completions — 对话生成
• POST /responses — Responses API
• POST /messages — Anthropic Messages API
• POST /images/generations — 图片生成
• POST /images/edits — 图片编辑
• POST /embeddings — Embeddings
• POST /rerank — Rerank
• GET /models — 模型列表
• POST /codex/chat/completions — Codex 专线
• GET /dashboard/billing/usage — 账单与用量

模型一览

实时价格 → /link#pricing(从 DB 实时读)。
可用模型 → GET /models(按 Key 权限筛选)。

OpenAI Chat Completions 兼容端点,接收聊天历史返回模型回复。支持流式、视觉、推理强度、函数调用。

参数

请求示例

curl https://www.glouth.com/link/v1/chat/completions \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "reasoning_effort": "medium",
    "messages": [
      {"role":"system","content":"You are a precise assistant."},
      {"role":"user","content":"What is 23 * 17?"}
    ]
  }'

响应示例

{
  "id": "chatcmpl_...",
  "object": "chat.completion",
  "model": "gpt-5.4",
  "choices": [{
    "index": 0,
    "message": {"role":"assistant","content":"391"},
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 34,
    "completion_tokens": 24,
    "total_tokens": 58,
    "prompt_tokens_details": {"cached_tokens": 0},
    "completion_tokens_details": {"reasoning_tokens": 17}
  }
}

OpenAI 新版 Responses 端点,通过 input 字段提交内容,适合长上下文与会话状态托管类用法。所有 chat/completions 的特性同样适用。

参数

请求示例

curl https://www.glouth.com/link/v1/responses \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "input": "用一句话介绍 Glouth Link。"
  }'

响应示例

{
  "id": "resp_...",
  "object": "response",
  "output": [{"type":"message","content":[{"type":"output_text","text":"..."}]}],
  "usage": {"input_tokens": 12, "output_tokens": 21}
}

完全兼容 Anthropic /v1/messages 接口形态。请求头带 anthropic-version 或 x-api-key,或 body 含 system/content block 时,主站自动走 Anthropic 协议路径。

参数

请求示例

curl https://www.glouth.com/link/v1/messages \
  -H "x-api-key: $GLOUTH_LINK_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-3-5-sonnet-20241022",
    "max_tokens": 1024,
    "thinking": {"type":"enabled","budget_tokens":4000},
    "messages": [
      {"role":"user","content":"What is 25*4?"}
    ]
  }'

响应示例

{
  "id": "msg_...",
  "type": "message",
  "role": "assistant",
  "model": "claude-3-5-sonnet-20241022",
  "content": [{"type":"text","text":"100"}],
  "stop_reason": "end_turn",
  "usage": {"input_tokens": 26, "output_tokens": 18}
}

调用 gpt-image-2 模型从文本提示生成图片。响应中返回 b64_json 或 url。按请求计固定单价。

参数

请求示例

curl https://www.glouth.com/link/v1/images/generations \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A minimalist green leaf on white background",
    "size": "1024x1024",
    "n": 1
  }'

响应示例

{
  "created": 1778561900,
  "data": [{
    "b64_json": "iVBORw0KGgoAAAA..."
  }]
}

基于现有图片做编辑/重绘。输入图片以 base64 形式放入 JSON body(无需 multipart),与 prompt 一起提交。

参数

请求示例

curl https://www.glouth.com/link/v1/images/edits \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "Add soft snow falling in front of the scene",
    "image_data_url": "data:image/png;base64,iVBORw0KGgo...",
    "size": "1024x1024"
  }'

响应示例

{"created": ..., "data": [{"b64_json": "..."}]}

将文本转为向量,用于检索 / 相似度匹配。OpenAI 兼容格式。

参数

请求示例

curl https://www.glouth.com/link/v1/embeddings \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"your-embedding-model","input":"Glouth Link"}'

响应示例

{
  "object": "list",
  "data": [{"object":"embedding","embedding":[0.012, -0.041, ...],"index":0}],
  "model": "your-embedding-model",
  "usage": {"prompt_tokens":3, "total_tokens":3}
}

对文档列表按 query 相关度重排,常用于搜索结果优化。

参数

请求示例

curl https://www.glouth.com/link/v1/rerank \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"your-rerank-model","query":"glouth","documents":["a","b","c"],"top_n":2}'

响应示例

{"results":[{"index":2,"relevance_score":0.81},...]}

返回当前 Key 可见的模型集合。若 Key 设置了 allowed_models 白名单,响应自动收窄。

请求示例

curl https://www.glouth.com/link/v1/models \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY"

响应示例

{
  "object": "list",
  "data": [
    {"id":"gpt-5.4","object":"model","created":1778000000,"owned_by":"glouth"},
    {"id":"gpt-5.4-mini","object":"model","created":1778000000,"owned_by":"glouth"},
    ...
  ]
}

Codex 专线是同等能力的低价模型组(codex-* 前缀,价格约低 30%),适合代码生成、批量任务、长流水线场景。响应结构跟 /chat/completions 完全一致。

参数

请求示例

curl https://www.glouth.com/link/v1/codex/chat/completions \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"codex-gpt-5.5","messages":[{"role":"user","content":"hi"}]}'

响应示例

{… 结构同 /chat/completions …}

查询当前 Key 关联的余额、硬额度、今日 / 本月消耗、近 30 天模型占比。响应字段稳定。别名 GET /key/usage。

请求示例

curl https://www.glouth.com/link/v1/dashboard/billing/usage \
  -H "Authorization: Bearer $GLOUTH_LINK_KEY"

响应示例

{
  "balance_cents": 12345,
  "hard_limit_cents": 100000,
  "used_cents": {"today": 482, "this_month": 4152, "this_30d": 5811},
  "request_count": 1234,
  "by_model_30d": [{"model":"gpt-5.4-mini","requests":612,"cost_cents":2103}, ...]
}

推理程度调节(reasoning_effort)

更深推理 → 答案更稳但更慢更贵。Glouth 支持 4 档:low / medium / high / xhigh(后者为 Glouth 扩展)。

Chat Completions 协议

{
  "model": "gpt-5.4",
  "reasoning_effort": "high",
  "messages": [...]
}

Messages 协议

通过 thinking.budget_tokens 按预算自动映射:

{
  "model": "gpt-5.5",
  "thinking": {"type": "enabled", "budget_tokens": 4000},
  "messages": [...]
}

# 自动映射:
# <2000  → low
# <8000  → medium
# <16000 → high
# ≥16000 → xhigh

快速模式

2 个降延迟的方法:

方案一:设 reasoning_effort: "low"

不改模型,只降推理深度。

方案二:透传 service_tier

{"model": "gpt-5.4", "service_tier": "flex", "messages": [...]}

图片输入(Vision)

Chat Completions 风格

{
  "model": "gpt-5.4",
  "messages": [{
    "role": "user",
    "content": [
      {"type": "text", "text": "这是什么?"},
      {"type": "image_url", "image_url": {"url": "https://.../cat.jpg"}}
    ]
  }]
}

Messages 风格

{
  "model": "gpt-5.5",
  "max_tokens": 512,
  "messages": [{
    "role": "user",
    "content": [
      {"type": "text", "text": "这是什么?"},
      {"type": "image", "source": {"type": "base64", "media_type": "image/jpeg", "data": "..."}}
    ]
  }]
}

流式 SSE

所有对话端点支持 stream:true。请求 Accept: text/event-stream,响应按 SSE 分帧。

Chat Completions 风格响应帧

data: {"id":"...","object":"chat.completion.chunk","choices":[{"delta":{"content":"H"}}]}

data: {"id":"...","object":"chat.completion.chunk","choices":[{"delta":{"content":"i"}}]}

data: {"id":"...","object":"chat.completion.chunk","choices":[{"finish_reason":"stop","delta":{}}],"usage":{...}}

data: [DONE]

Messages 风格响应帧

event: message_start
data: {"type":"message_start","message":{...}}

event: content_block_start
data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"H"}}

...

event: message_stop
data: {"type":"message_stop"}

提示缓存(Prompt Caching)

常用系统 prompt 自动被上游缓存,后续命中价格降至 input 的 10%。Glouth 不需要你做任何特殊设置 — 上游(OpenAI / Anthropic)自动判定。

响应 usage.prompt_tokens_details.cached_tokens 字段告诉你这次命中了多少 cache。

函数调用 / Tools

OpenAI 风格

{
  "model": "gpt-5.4",
  "messages": [...],
  "tools": [{
    "type": "function",
    "function": {
      "name": "get_weather",
      "description": "获取某地天气",
      "parameters": {
        "type": "object",
        "properties": {"city": {"type":"string"}}
      }
    }
  }],
  "tool_choice": "auto"
}

Anthropic 风格

{
  "model": "claude-3-5-sonnet-20241022",
  "max_tokens": 512,
  "messages": [...],
  "tools": [{
    "name": "get_weather",
    "description": "获取某地天气",
    "input_schema": {"type":"object","properties":{"city":{"type":"string"}}}
  }]
}

Claude Code 集成

用 Glouth Link 取代官方 Anthropic 端点,RMB 充值 + 国内直连。

建议配置

# 一键安装(macOS / Linux)
curl -fsSL https://www.glouth.com/link/claude/setup.sh | sh

# 或手动配置 ~/.claude/settings.json
{
  "apiKeyHelper": "echo \"$GLOUTH_LINK_KEY\"",
  "env": {
    "ANTHROPIC_API_KEY": "$GLOUTH_LINK_KEY",
    "ANTHROPIC_BASE_URL": "https://www.glouth.com/link/v1"
  }
}

OpenAI SDK

Node.js / TypeScript

import OpenAI from 'openai'
const client = new OpenAI({
  apiKey: process.env.GLOUTH_LINK_KEY,
  baseURL: 'https://www.glouth.com/link/v1',
})

Python

from openai import OpenAI
client = OpenAI(
    api_key=os.environ['GLOUTH_LINK_KEY'],
    base_url='https://www.glouth.com/link/v1',
)

Anthropic SDK

from anthropic import Anthropic
client = Anthropic(
    api_key=os.environ['GLOUTH_LINK_KEY'],
    base_url='https://www.glouth.com/link/v1',
)

错误码

提工单请把响应头里的 X-Glouth-Request-ID 一并发给客服。