OpenAI Batch接口兼容

上传Batch输入文件

创建Batch任务前需要准备输入文件，通过文件上传接口上传后获取file_id。

文件格式为JSONL，每一行为一个JSON格式的请求，参考输入文件格式。单个Batch任务最多包含5万个请求，Batch文件的大小最大100 MB，文件中单行的大小最大 1 MB（单行的请求内容还需要遵循各模型上下文长度的限制）。

用户的文件空间有限制，文件空间达到限制后，需要用户自行删除不需要的文件，释放空间。

import os
from pathlib import Path
from openai import OpenAI

client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),  # 如果您没有配置环境变量，请在此处用您的API Key进行替换
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",  # 填写百炼服务的base_url
)

# test.jsonl 是一个本地示例文件，purpose必须是batch
file_object = client.files.create(file=Path("test.jsonl"), purpose="batch")

print(file_object.model_dump_json())
print(file_object.id)  # 打印文件id

{"custom_id": "request-1", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "qwen-turbo", "messages": [{"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "What is 2+2?"}]}}

{
    "id": "file-batch-xxx",
    "bytes": 231,
    "created_at": 1729065815,
    "filename": "test.jsonl",
    "object": "file",
    "purpose": "batch",
    "status": "processed",
    "status_details": null
}

curl -X POST https://dashscope.aliyuncs.com/compatible-mode/v1/files \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
--form 'file=@"test.jsonl"' \
--form 'purpose="batch"'

{"custom_id": "request-1", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "qwen-turbo", "messages": [{"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "What is 2+2?"}]}}

{
    "id": "file-batch-xxx",
    "bytes": 231,
    "created_at": 1729065815,
    "filename": "test.jsonl",
    "object": "file",
    "purpose": "batch",
    "status": "processed",
    "status_details": null
}

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),  # 如果您没有配置环境变量，请在此处用您的API Key进行替换
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",  # 填写百炼服务的base_url
)

batch = client.batches.create(
    input_file_id="file-batch-xxx",  # 上传文件返回的 id
    endpoint="/v1/chat/completions",  # 大语言模型固定填写，/v1/chat/completions
    completion_window="24h"  # 当前只支持24h，24小时未运行完会超时
)
print(batch)
print(batch.id)  # 打印Batch任务的id

POST https://dashscope.aliyuncs.com/compatible-mode/v1/batches

curl -X POST https://dashscope.aliyuncs.com/compatible-mode/v1/batches \
  -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input_file_id": "file-batch-xxx",
    "endpoint": "/v1/chat/completions",
    "completion_window": "24h"
  }'

{
  "id": "batch_cf957e2b-3295-4357-83af-5fb22ac17e81",
  "object": "batch",
  "endpoint": "/v1/chat/completions",
  "errors": null,
  "input_file_id": "file-batch-ADvrUmLWW8TFuKNjGDUUig1c",
  "completion_window": "24h",
  "status": "validating",
  "output_file_id": null,
  "error_file_id": null,
  "created_at": 1722503223,
  "in_progress_at": null,
  "expires_at": null,
  "finalizing_at": null,
  "completed_at": null,
  "failed_at": null,
  "expired_at": null,
  "cancelling_at": null,
  "cancelled_at": null,
  "request_counts": {
    "total": 0,
    "completed": 0,
    "failed": 0
  },
  "metadata": {}
}

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),  # 如果您没有配置环境变量，请在此处用您的API Key进行替换
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",  # 填写百炼服务的base_url
)
batch = client.batches.retrieve("batch_id")  # 将batch_id替换为Batch任务的id
print(batch)
print(batch.output_file_id)  # 打印输出文件id

GET https://dashscope.aliyuncs.com/compatible-mode/v1/batches/batch_id

curl --request GET 'https://dashscope.aliyuncs.com/compatible-mode/v1/batches/batch_id' \
 -H "Authorization: Bearer $DASHSCOPE_API_KEY"

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),  # 如果您没有配置环境变量，请在此处用您的API Key进行替换
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",  # 填写DashScope服务的base_url
)
batch = client.batches.cancel("batch_id")  # 将batch_id替换为Batch任务的id
print(batch)

POST https://dashscope.aliyuncs.com/compatible-mode/v1/batches/batch_id/cancel

curl --request POST 'https://dashscope.aliyuncs.com/compatible-mode/v1/batches/batch_id/cancel' \
 -H "Authorization: Bearer $DASHSCOPE_API_KEY"

查询Batch任务列表

您可以将after="batch_id"中的batch_id替换为实际值，在后续查询时可以从指定的Batch任务开始获取下一页的Batch任务，并通过limit来指定返回任务的数量。

接口限流：每个阿里云主账号每分钟100次。

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),  # 如果您没有配置环境变量，请在此处用您的API Key进行替换
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",  # 填写DashScope服务的base_url
)
batches = client.batches.list(after="batch_id", limit=20)  # 将batch_id替换为Batch任务的id
print(batches)

GET https://dashscope.aliyuncs.com/compatible-mode/v1/batches?limit=20&after=batch_id

curl --request GET  'https://dashscope.aliyuncs.com/compatible-mode/v1/batches?limit=20&after=batch_id' \
 -H "Authorization: Bearer $DASHSCOPE_API_KEY"

输入参数配置

返回示例

{
  "object": "list",
  "data": [
    {
      "id": "batch_b8b6a83e-00a3-4068-81d9-0a1705bf6b2c",
      "object": "batch",
      "endpoint": "/v1/chat/completions",
      "errors": null,
      "input_file_id": "file-batch-IvxTEDheD70YmyXfqrkE2t64",
      "completion_window": "24h",
      "status": "completed",
      "output_file_id": "file-batch_output-fUy6gxJCyHzT0WHDg2SInyTp",
      "error_file_id": null,
      "created_at": 1722234109,
      "in_progress_at": 1722234109,
      "expires_at": null,
      "finalizing_at": 1722234165,
      "completed_at": 1722234165,
      "failed_at": null,
      "expired_at": null,
      "cancelling_at": null,
      "cancelled_at": null,
      "request_counts": {
        "total": 100,
        "completed": 95,
        "failed": 5
      },
      "metadata": {}
    },
    { ... }
  ],
  "first_id": "batch_b8b6a83e-00a3-4068-81d9-0a1705bf6b2c",
  "last_id": "batch_ca912d83-e9e8-4cf2-89a4-d29743e334e4",
  "has_more": true
}

返回参数

输入文件格式

JSONL文件，每一行一个JSON格式的请求。单个Batch任务最多包含5万个请求，Batch文件的大小最大100 MB，文件中单行的大小最大 1 MB（单行的请求内容还需要遵循各模型上下文长度的限制）。

示例

{"custom_id": "request-1", "method": "POST", "url": "/v1/chat/completions", "body": {"model": "qwen-turbo", "messages": [{"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "What is 2+2?"}]}}

参数

输出文件格式

JSONL文件，每一行一个JSON。

示例

{"id": "batch_req_wnaDys", "custom_id": "request-2", "response": {"status_code": 200, "request_id": "req_c187b3", "body": {"id": "chatcmpl-9758Iw", "object": "chat.completion", "created": 1711475054, "model": "qwen-max", "choices": [{"index": 0, "message": {"role": "assistant", "content": "2 + 2 equals 4."}, "finish_reason": "stop"}], "usage": {"prompt_tokens": 24, "completion_tokens": 15, "total_tokens": 39}, "system_fingerprint": null}}, "error": null}

参数

常见问题

1. 这几个模块的Batch定价，对应的模型也有基础限流吗？

   答：实时调用才会有QPM（Queries Per Minute：每分钟处理请求数）限流，Batch调用没有 QPM 限流。

2. 使用Batch调用，是否需要下单，在哪里下单？

   答：Batch是一种调用方式，无需额外下单。该调用方式为后付费计价模式，按照Batch接口调用直接付费。

3. 提交的Batch调用请求，后台如何处理？ 是根据提交请求的先后顺序来执行吗？

   答：不是排队机制，是调度机制，根据资源情况来调度和执行Batch请求任务。

4. 报错如何处理？

   答：请参见错误码。