64 lines
1.7 KiB
Plaintext
64 lines
1.7 KiB
Plaintext
1|# KTV Pipeline / GPU Service Inference API
|
||
2|# POST /v1/pipeline/submit
|
||
3|# catelogid 固定为 ktv_pipeline，无需传 catelogid
|
||
4|#
|
||
5|# Required params:
|
||
6|# model: string - 模型名称，如 "ky-asr-transcribe"
|
||
7|#
|
||
8|# 各服务特定参数见 API 文档
|
||
9|#
|
||
10|# 同步服务直接返回结果：
|
||
11|# {
|
||
12|# "taskid": "luid_xxx",
|
||
13|# "taskstatus": "SUCCEEDED",
|
||
14|# ...服务特定结果字段...,
|
||
15|# "usage": {...}
|
||
16|# }
|
||
17|#
|
||
18|# 异步服务返回任务信息：
|
||
19|# {
|
||
20|# "taskid": "luid_xxx",
|
||
21|# "taskstatus": "PENDING"
|
||
22|# }
|
||
23|# 异步结果通过 /v1/tasks?taskid=xxx 查询
|
||
24|
|
||
25|debug_params('params_kw', params_kw)
|
||
26|
|
||
27|userid = await get_user()
|
||
28|userorgid = await get_userorgid()
|
||
29|if userid is None:
|
||
30| debug('need login')
|
||
31| return openai_403()
|
||
32|
|
||
33|# Validate required parameters
|
||
34|if not params_kw.model:
|
||
35| d = return_error('Missing required parameter: model')
|
||
36| return json_response(d, status=400)
|
||
37|
|
||
38|catelogid = 'ktv_pipeline'
|
||
39|params_kw.catelogid = catelogid
|
||
40|
|
||
41|env = request._run_ns
|
||
42|# llmid from cache (model+catelogid -> llmid)
|
||
llmid = await env.get_llmid_cached(env, params_kw.model or 'qwen3-max', 'ktv_pipeline')
|
||
if not llmid:
|
||
debug(f'model not found: params_kw.model')
|
||
return {{error_response}}
|
||
params_kw.llmid = llmid
|
||
59| params_kw.llmcatelogid = catelogid
|
||
60|
|
||
61|debug(f'{params_kw.llmid=}')
|
||
62|
|
||
63|# Check balance
|
||
64|f = await checkCustomerBalance(params_kw.llmid, userid, userorgid)
|
||
65|if not f:
|
||
66| debug(f'{userid=} balance not enough')
|
||
67| return openai_429()
|
||
68|
|
||
69|# Generate task ID
|
||
70|if not params_kw.transno:
|
||
71| params_kw.transno = getID()
|
||
72|
|
||
73|# Call inference
|
||
74|return await inference(request, env=env)
|
||
75| |