接入说明

1. 可用模型

2. 对话 / 写代码

复制# 非流式
curl -X POST "BASE/v1/chat/completions" \
  -H "Authorization: Bearer $MUSE2API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"muse-spark","messages":[{"role":"user","content":"写一个 Python 快排"}]}'

# 流式：加 "stream":true，返回标准 SSE（data: {...} / data: [DONE]）

3. 生图（curl）

复制curl -X POST "BASE/v1/images/generations" \
  -H "Authorization: Bearer $MUSE2API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt":"一只戴宇航头盔的柴犬","size":"1:1","response_format":"url"}'

# 返回
{"created":1790148495,"data":[{
  "revised_prompt":"一只戴宇航头盔的柴犬",
  "url":"http://your-server:18610/v1/media/xxxx.webp",
  "kind":"image","bytes":23860}]}

# url 是完整绝对地址，可直接给下游渲染 / 下载；/v1/media/* 无需鉴权。
# 想要 base64 就传 "response_format":"b64_json"。

4. 生视频（curl，异步）

复制# ① 创建任务
curl -X POST "BASE/v1/videos" \
  -H "Authorization: Bearer $MUSE2API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt":"一只小猫在草地上走动","duration":5,"size":"16:9"}'
# → {"id":"task_xxx","status":"queued"}

# ② 轮询结果
curl "BASE/v1/videos/task_xxx" -H "Authorization: Bearer $MUSE2API_KEY"
# → {"status":"succeeded","result":{"url":"/v1/media/xxx.mp4"}}

5. Python（openai SDK 兼容）

复制from openai import OpenAI
import time

client = OpenAI(base_url="BASE/v1", api_key="m2a_...")

# 生图
r = client.images.generate(model="muse-image", prompt="一只戴宇航头盔的柴犬")
print(r.data[0].url)

# 生视频（走原生接口，因为是异步任务）
import requests
t = requests.post("BASE/v1/videos", json={"prompt":"小猫走动","duration":5},
                  headers={"Authorization":"Bearer m2a_..."}).json()
while True:
    s = requests.get(f"BASE/v1/videos/{t['id']}",
                     headers={"Authorization":"Bearer m2a_..."}).json()
    if s["status"] in ("succeeded","failed"): break
    time.sleep(5)
print(s)

6. 管理接口

7. 鉴权

8. 接入各类客户端 / 智能体

按客户端类型看能不能用