[Bug] 以anthropic提供商格式接入DeepSeek V4时，并行调用多个工具会导致400报错

[FFFold]

What happened / 发生了什么

DeepSeek V4 以anthropic提供商格式接入时，并行调用多个工具会导致400报错。

当 Assistant 在一次响应中并行发起多个 tool_use 时，AstrBot 的 Anthropic 适配层将后续多个 role="tool" 的返回结果转换成了多条独立的 User 消息。这可能导致 Anthropic api 校验失败，抛出 400 错误。

AstrBot 的 anthropic_source.py 将多个 role="tool" 的返回结果映射成了多条独立的 User 消息，导致 Anthropic 看到的序列如下：

Assistant: [tool_use_00, tool_use_01]   ← 两个并行工具调用
User:      [tool_result_00]             ← 被当成“下一条消息”
User:      [tool_result_01]             ← Anthropic api可能认为这条是“多余的用户消息”

因此 Anthropic 校验逻辑判定：tool_use_01（call_01_BhsvXomagk6BoekAmAaC5Gtp）之后没有紧跟的 tool_result，直接拒绝请求并返回 400。

修复方向：在 anthropic_source.py 的消息转换逻辑中，当遇到多个连续的 role="tool" 消息时，应将其合并为单条 User 消息，结构示例：

{
  "role": "user",
  "content": [
    {
      "type": "tool_result",
      "tool_use_id": "call_00_5EQ0V4NahGLAKFDZp8hAVwLT",
      "content": "...SKILL.md 内容..."
    },
    {
      "type": "tool_result",
      "tool_use_id": "call_01_BhsvXomagk6BoekAmAaC5Gtp",
      "content": "...pubmed.py 内容..."
    }
  ]
}

Reproduce / 如何复现？

1. 配置使用 Anthropic 兼容格式的 deepseek-v4-flash
2. 触发一次需要并行读取两个文件（或并行调用两个工具）的请求。
3. 观察 API 返回 400 错误。

AstrBot version, deployment method (e.g., Windows Docker Desktop deployment), provider used, and messaging platform used. / AstrBot 版本、部署方式（如 Windows Docker Desktop 部署）、使用的提供商、使用的消息平台适配器

Docker部署，宿主机 Ubuntu server 24.04

OS

Linux

Logs / 报错日志

触发报错的结构：

[
  {
    "role": "assistant",
    "content": [
      {
        "type": "think",
        "think": "主人问有没有PubMed这个Skill...",
        "encrypted": "..."
      },
      {
        "type": "text",
        "text": "目前还没有正式发布的Skill呢～不过堇的本事可不少！..."
      }
    ],
    "tool_calls": [
      {
        "type": "function",
        "id": "call_00_sJcXYPcbhpxQYDPCjuVgDmf4",
        "function": {
          "name": "astrbot_list_skill_releases",
          "arguments": "{}"
        }
      }
    ]
  },
  {
    "role": "tool",
    "content": "{\"items\": [], \"total\": 0}",
    "tool_call_id": "call_00_sJcXYPcbhpxQYDPCjuVgDmf4"
  },
  {
    "role": "assistant",
    "content": [
      {
        "type": "think",
        "think": "目前没有已发布的技能。那我直接给主人列出我所有能调用的工具和能力吧！",
        "encrypted": "..."
      },
      {
        "type": "text",
        "text": "目前还没有正式发布的Skill呢～不过堇的本事可不少！..."
      }
    ],
    "tool_calls": [
      {
        "type": "function",
        "id": "call_00_5EQ0V4NahGLAKFDZp8hAVwLT",
        "function": {
          "name": "astrbot_file_read_tool",
          "arguments": "{\"path\": \"/workspace/skills/pubmed/SKILL.md\"}"
        }
      },
      {
        "type": "function",
        "id": "call_01_BhsvXomagk6BoekAmAaC5Gtp",
        "function": {
          "name": "astrbot_file_read_tool",
          "arguments": "{\"path\": \"/workspace/skills/pubmed/scripts/pubmed.py\"}"
        }
      }
    ]
  },
  {
    "role": "tool",
    "content": "---\nname: pubmed\ndescription: 你是一位拥有医学背景的文献检索专家...\n...（SKILL.md 完整内容）...",
    "tool_call_id": "call_00_5EQ0V4NahGLAKFDZp8hAVwLT"
  },
  {
    "role": "tool",
    "content": "#!/usr/bin/env python3\n...（pubmed.py 完整内容）...",
    "tool_call_id": "call_01_BhsvXomagk6BoekAmAaC5Gtp"
  },
  {
    "role": "assistant",
    "content": [
      {
        "type": "think",
        "think": "主人确实上传了PubMed的SKILL文件！...",
        "encrypted": null
      },
      {
        "type": "text",
        "text": "对不起主人！堇刚才说没有PubMed Skill..."
      }
    ]
  }
]

注： 报错似乎发生在最后一条 role="assistant" 消息被提交给 LLM 时。API 校验发现 call_01_BhsvXomagk6BoekAmAaC5Gtp 没有紧跟的 tool_result 块。

2026-04-28 16:59:55.306] [Core] [WARN] [v4.23.6] [runners.tool_loop_agent_runner:555]: Chat Model deepseek-a/deepseek-v4-flash request error: Error code: 400 - {'error': {'message': 'messages.43:`tool_use` ids were found without `tool_result` blocks immediately after: call_01_BhsvXomagk6BoekAmAaC5Gtp. Each `tool_use` block must have a corresponding `tool_result` block in the next message.', 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_request_error'}}
Traceback (most recent call last):
  File "/AstrBot/astrbot/core/agent/runners/tool_loop_agent_runner.py", line 510, in _iter_llm_responses_with_fallback
    async for attempt in retrying:
  File "/usr/local/lib/python3.12/site-packages/tenacity/asyncio/__init__.py", line 170, in __anext__
    do = await self.iter(retry_state=self._retry_state)
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/tenacity/asyncio/__init__.py", line 157, in iter
    result = await action(retry_state)
             ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/tenacity/_utils.py", line 111, in inner
    return call(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/tenacity/__init__.py", line 393, in <lambda>
    self._add_action_func(lambda rs: rs.outcome.result())
                                     ^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/concurrent/futures/_base.py", line 449, in result
    return self.__get_result()
           ^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result
    raise self._exception
  File "/AstrBot/astrbot/core/agent/runners/tool_loop_agent_runner.py", line 514, in _iter_llm_responses_with_fallback
    async for resp in self._iter_llm_responses(
  File "/AstrBot/astrbot/core/agent/runners/tool_loop_agent_runner.py", line 477, in _iter_llm_responses
    yield await self.provider.text_chat(**payload)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/AstrBot/astrbot/core/provider/sources/anthropic_source.py", line 581, in text_chat
    raise e
  File "/AstrBot/astrbot/core/provider/sources/anthropic_source.py", line 579, in text_chat
    llm_response = await self._query(payloads, func_tool)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/AstrBot/astrbot/core/provider/sources/anthropic_source.py", line 299, in _query
    completion = await self.client.messages.create(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/anthropic/resources/messages/messages.py", line 2447, in create
    return await self._post(
           ^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/anthropic/_base_client.py", line 1996, in post
    return await self.request(cast_to, opts, stream=stream, stream_cls=stream_cls)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/anthropic/_base_client.py", line 1781, in request
    raise self._make_status_error_from_response(err.response) from None
anthropic.BadRequestError: Error code: 400 - {'error': {'message': 'messages.43:`tool_use` ids were found without `tool_result` blocks immediately after: call_01_BhsvXomagk6BoekAmAaC5Gtp. Each `tool_use` block must have a corresponding `tool_result` block in the next message.', 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_request_error'}}
[2026-04-28 16:59:55.327] [Core] [WARN] [v4.23.6] [runners.tool_loop_agent_runner:492]: Switched from deepseek-a/deepseek-v4-flash to fallback chat provider: 阿里百炼/deepseek-v4-flash

Are you willing to submit a PR? / 你愿意提交 PR 吗？

• Yes!

Code of Conduct

• I have read and agree to abide by the project's Code of Conduct。

[dosubot]

确认了这个 bug。问题出在 anthropic_source.py 的 _prepare_payload 方法中（[1]），每个 role="tool" 消息都被单独转换为一条 role="user" 消息，而没有将连续的 tool 结果合并为一条包含多个 tool_result 的 User 消息。

修复思路正如你所述：在遍历消息时，检测连续的 role="tool" 消息并将它们合并到同一个 User 消息的 content 数组中。大致逻辑：

elif message["role"] == "tool":
    tool_result_block = {
        "type": "tool_result",
        "tool_use_id": message["tool_call_id"],
        "content": message["content"] or "<empty response>",
    }
    # 如果上一条已经是 tool_result 的 user 消息，则追加到其 content 中
    if (
        new_messages
        and new_messages[-1]["role"] == "user"
        and isinstance(new_messages[-1]["content"], list)
        and new_messages[-1]["content"]
        and new_messages[-1]["content"][0].get("type") == "tool_result"
    ):
        new_messages[-1]["content"].append(tool_result_block)
    else:
        new_messages.append({
            "role": "user",
            "content": [tool_result_block],
        })

这样多个并行 tool 结果会被合并为单条 User 消息，符合 Anthropic API 的要求。

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.