What: switch to 你/你的 tone; standardize Skills/Gateway网关/local loopback/私信 wording Why: align zh-CN docs with issue 6995 feedback + idiomatic tech style Tests: pnpm docs:build
126 lines
3.2 KiB
Markdown
126 lines
3.2 KiB
Markdown
---
|
||
read_when:
|
||
- 集成需要 OpenAI Chat Completions 的工具
|
||
summary: 从 Gateway网关暴露一个兼容 OpenAI 的 /v1/chat/completions HTTP 端点
|
||
title: OpenAI Chat Completions
|
||
x-i18n:
|
||
generated_at: "2026-02-01T20:35:14Z"
|
||
model: claude-opus-4-5
|
||
provider: pi
|
||
source_hash: 6f935777f489bff925a3bf18b1e4b7493f83ae7b1e581890092e5779af59b732
|
||
source_path: gateway/openai-http-api.md
|
||
workflow: 14
|
||
---
|
||
|
||
# OpenAI Chat Completions (HTTP)
|
||
|
||
OpenClaw 的 Gateway网关可以提供一个小型的兼容 OpenAI 的 Chat Completions 端点。
|
||
|
||
此端点**默认禁用**。请先在配置中启用它。
|
||
|
||
- `POST /v1/chat/completions`
|
||
- 与 Gateway网关使用相同端口(WS + HTTP 多路复用):`http://<gateway-host>:<port>/v1/chat/completions`
|
||
|
||
在底层,请求会作为普通的 Gateway网关智能体运行来执行(与 `openclaw agent` 相同的代码路径),因此路由/权限/配置与你的 Gateway网关保持一致。
|
||
|
||
## 认证
|
||
|
||
使用 Gateway网关的认证配置。发送 Bearer 令牌:
|
||
|
||
- `Authorization: Bearer <token>`
|
||
|
||
注意:
|
||
|
||
- 当 `gateway.auth.mode="token"` 时,使用 `gateway.auth.token`(或 `OPENCLAW_GATEWAY_TOKEN`)。
|
||
- 当 `gateway.auth.mode="password"` 时,使用 `gateway.auth.password`(或 `OPENCLAW_GATEWAY_PASSWORD`)。
|
||
|
||
## 选择智能体
|
||
|
||
无需自定义请求头:在 OpenAI 的 `model` 字段中编码智能体 ID:
|
||
|
||
- `model: "openclaw:<agentId>"`(示例:`"openclaw:main"`、`"openclaw:beta"`)
|
||
- `model: "agent:<agentId>"`(别名)
|
||
|
||
或通过请求头指定特定的 OpenClaw 智能体:
|
||
|
||
- `x-openclaw-agent-id: <agentId>`(默认值:`main`)
|
||
|
||
高级用法:
|
||
|
||
- `x-openclaw-session-key: <sessionKey>` 用于完全控制会话路由。
|
||
|
||
## 启用端点
|
||
|
||
将 `gateway.http.endpoints.chatCompletions.enabled` 设置为 `true`:
|
||
|
||
```json5
|
||
{
|
||
gateway: {
|
||
http: {
|
||
endpoints: {
|
||
chatCompletions: { enabled: true },
|
||
},
|
||
},
|
||
},
|
||
}
|
||
```
|
||
|
||
## 禁用端点
|
||
|
||
将 `gateway.http.endpoints.chatCompletions.enabled` 设置为 `false`:
|
||
|
||
```json5
|
||
{
|
||
gateway: {
|
||
http: {
|
||
endpoints: {
|
||
chatCompletions: { enabled: false },
|
||
},
|
||
},
|
||
},
|
||
}
|
||
```
|
||
|
||
## 会话行为
|
||
|
||
默认情况下,端点**每次请求无状态**(每次调用生成一个新的会话密钥)。
|
||
|
||
如果请求中包含 OpenAI 的 `user` 字符串,Gateway网关会根据它派生一个稳定的会话密钥,这样重复调用可以共享同一个智能体会话。
|
||
|
||
## 流式传输 (SSE)
|
||
|
||
设置 `stream: true` 以接收服务器发送事件 (SSE):
|
||
|
||
- `Content-Type: text/event-stream`
|
||
- 每个事件行格式为 `data: <json>`
|
||
- 流以 `data: [DONE]` 结束
|
||
|
||
## 示例
|
||
|
||
非流式:
|
||
|
||
```bash
|
||
curl -sS http://127.0.0.1:18789/v1/chat/completions \
|
||
-H 'Authorization: Bearer YOUR_TOKEN' \
|
||
-H 'Content-Type: application/json' \
|
||
-H 'x-openclaw-agent-id: main' \
|
||
-d '{
|
||
"model": "openclaw",
|
||
"messages": [{"role":"user","content":"hi"}]
|
||
}'
|
||
```
|
||
|
||
流式:
|
||
|
||
```bash
|
||
curl -N http://127.0.0.1:18789/v1/chat/completions \
|
||
-H 'Authorization: Bearer YOUR_TOKEN' \
|
||
-H 'Content-Type: application/json' \
|
||
-H 'x-openclaw-agent-id: main' \
|
||
-d '{
|
||
"model": "openclaw",
|
||
"stream": true,
|
||
"messages": [{"role":"user","content":"hi"}]
|
||
}'
|
||
```
|