[Bug] 400 from Grok after switching model

[karuboniru]

What happened / 发生了什么

root of issue: Grok don't accept any empty content field, but sometimes other models (in my case DeepSeek produces those in chat history)

Reproduce / 如何复现？

1. Start with a clean chat history
2. Switch model to DeepSeek w/ reasoning, instruct it to do something involving tool call (such that there are reasoning_content left but with empty content
3. Switch model to Grok series, e.g. grok/grok-4-1-fast

And get:

{"code":"Client specified an invalid argument","error":"Invalid request content: Each message must have at least one content element."}

---

By catching payload, I found entries like

   {
      "role": "assistant",
      "content": [],
      "reasoning_content": "..."
    },

and using the payload to replay the request to grok reproduces the exact error. After removing those entries, replay again, grok answers correctly.

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

v4.20.0 in container (podman), with telegram bot API

OS

Linux

Logs / 报错日志

[2026-03-16 13:14:04.437] [Core] [WARN] [v4.20.0] [runners.tool_loop_agent_runner:268]: Chat Model cloudflare_general/grok/grok-4-1-fast request error: Error code: 400 - {'code': 'Client specified an invalid argument', 'error': 'Invalid request content: Each message must have at least one content element.'}
Traceback (most recent call last):
  File "/AstrBot/astrbot/core/agent/runners/tool_loop_agent_runner.py", line 243, in _iter_llm_responses_with_fallback
    async for resp in self._iter_llm_responses(include_model=idx == 0):
  File "/AstrBot/astrbot/core/agent/runners/tool_loop_agent_runner.py", line 220, in _iter_llm_responses
    yield await self.provider.text_chat(**payload)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/AstrBot/astrbot/core/provider/sources/openai_source.py", line 763, in text_chat
    ) = await self._handle_api_error(
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/AstrBot/astrbot/core/provider/sources/openai_source.py", line 714, in _handle_api_error
    raise e
  File "/AstrBot/astrbot/core/provider/sources/openai_source.py", line 751, in text_chat
    llm_response = await self._query(payloads, func_tool)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/AstrBot/astrbot/core/provider/sources/openai_source.py", line 251, in _query
    completion = await self.client.chat.completions.create(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/openai/resources/chat/completions/completions.py", line 2714, in create
    return await self._post(
           ^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/openai/_base_client.py", line 1884, in post
    return await self.request(cast_to, opts, stream=stream, stream_cls=stream_cls)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/openai/_base_client.py", line 1669, in request
    raise self._make_status_error_from_response(err.response) from None
openai.BadRequestError: Error code: 400 - {'code': 'Client specified an invalid argument', 'error': 'Invalid request content: Each message must have at least one content element.'}

[2026-03-16 13:14:04.453] [Core] [INFO] [respond.stage:184]: Prepare to send - Karuboniru/157891005: [引用消息] LLM 响应错误: All chat models failed: BadRequestError: Error code: 400 - {'code': 'Client specified an invalid argument', 'error': 'Invalid request content: Each message must have at least one content element.'} 

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

• Yes!

Proposed fix

Filter out all chat history with empty content (those are not going to be used anyway).

Code of Conduct

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

[dosubot]

Hi @karuboniru! I'm Dosu and I’m helping the AstrBot team.

This is a known class of issues in AstrBot. The root cause is in openai_source.py's _finally_convert_payload() method - when extracting reasoning_content from think parts, if all parts are think parts, the content becomes an empty list [] rather than being removed:

new_content = []  # not including think part
for part in message["content"]:
    if part.get("type") == "think":
        reasoning_content += str(part.get("think"))
    else:
        new_content.append(part)
message["content"] = new_content  # Could be empty []

Similar fixes already exist - for example, Groq has a provider-specific override that strips reasoning_content entirely because it also rejects assistant messages with that field (PR #6065).

A similar fix for Grok would be to create a provider-specific source that either:

1. Strips reasoning_content fields from assistant messages (like Groq)
2. Converts empty content: [] to content: None or removes the field entirely

Related issues: #6108 (Kimi returning only reasoning_content), #5731 (empty content causing cascading failures), #4876 (empty content in scheduled tasks).

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

  

[karuboniru]

nice catch, closing as duplicated.

[karuboniru]

No, the PR only addresses groq not grok (why name is so close). We might want to extend the change to grok.