diff --git a/astrbot/core/provider/sources/openai_source.py b/astrbot/core/provider/sources/openai_source.py index f2d9474906..eb01ff01dc 100644 --- a/astrbot/core/provider/sources/openai_source.py +++ b/astrbot/core/provider/sources/openai_source.py @@ -519,6 +519,42 @@ async def get_models(self): except NotFoundError as e: raise Exception(f"获取模型列表失败：{e}") + @staticmethod + def _sanitize_assistant_messages(payloads: dict) -> None: + """在请求发送前过滤/规范化空的 assistant 消息。 + + 严格 API（Moonshot、DeepSeek Reasoner 等）会在 assistant 消息同时缺少 + ``content`` 和 ``tool_calls`` 时返回 400。把 ``""`` / ``None`` / ``[]`` + 都视作空内容：无 tool_calls 时整条过滤掉；有 tool_calls 时将 content + 设为 ``None`` 以符合 OpenAI 规范。就地修改 ``payloads["messages"]``。 + """ + messages = payloads.get("messages") + if not isinstance(messages, list): + return + + def _is_empty(content: Any) -> bool: + return content is None or content == "" or content == [] + + cleaned: list[Any] = [] + for idx, msg in enumerate(messages): + if not isinstance(msg, dict) or msg.get("role") != "assistant": + cleaned.append(msg) + continue + + content = msg.get("content") + tool_calls = msg.get("tool_calls") + + if _is_empty(content) and not tool_calls: + logger.warning(f"过滤第 {idx} 条空 assistant 消息 (无工具调用)") + continue + + if _is_empty(content) and tool_calls: + msg["content"] = None + + cleaned.append(msg) + + payloads["messages"] = cleaned + async def _query(self, payloads: dict, tools: ToolSet | None) -> LLMResponse: if tools: model = payloads.get("model", "").lower() @@ -548,26 +584,7 @@ async def _query(self, payloads: dict, tools: ToolSet | None) -> LLMResponse: model = payloads.get("model", "").lower() - if "messages" in payloads and isinstance(payloads["messages"], list): - cleaned_messages = [] - for idx, msg in enumerate(payloads["messages"]): - # 过滤空的 assistant 消息，防止严格 API（如 Moonshot）返回 400 错误 - if msg.get("role") == "assistant": - content = msg.get("content") - tool_calls = msg.get("tool_calls") - - # 情况1: 空/null content 且无 tool_calls -> 过滤掉 - if not tool_calls and (content == "" or content is None): - logger.warning(f"过滤第 {idx} 条空 assistant 消息 (无工具调用)") - continue - - # 情况2: 空 content 但有 tool_calls -> 设为 None (符合 OpenAI 规范) - if content == "" and tool_calls: - msg["content"] = None - - cleaned_messages.append(msg) - - payloads["messages"] = cleaned_messages + self._sanitize_assistant_messages(payloads) completion = await self.client.chat.completions.create( **payloads, @@ -619,6 +636,8 @@ async def _query_stream( del payloads[key] self._apply_provider_specific_extra_body_overrides(extra_body) + self._sanitize_assistant_messages(payloads) + stream = await self.client.chat.completions.create( **payloads, stream=True, diff --git a/tests/test_openai_source.py b/tests/test_openai_source.py index 83e18137c4..ec5e79f492 100644 --- a/tests/test_openai_source.py +++ b/tests/test_openai_source.py @@ -1618,3 +1618,109 @@ async def fake_create(**kwargs): assert messages[2] == {"role": "user", "content": "hello"} finally: await provider.terminate() + + +@pytest.mark.asyncio +async def test_query_stream_filters_empty_assistant_message(monkeypatch): + """Regression for #7721: streaming path must also filter empty assistant messages. + + Previously only ``_query`` sanitized the payload; ``_query_stream`` forwarded + the raw history and strict providers (e.g. DeepSeek Reasoner) returned 400 on + the next turn after a tool call whose assistant entry had reasoning only. + """ + provider = _make_provider() + try: + captured_kwargs = {} + + async def fake_stream(): + yield ChatCompletionChunk.model_validate( + { + "id": "chatcmpl-stream", + "object": "chat.completion.chunk", + "created": 0, + "model": "deepseek-reasoner", + "choices": [ + { + "index": 0, + "delta": {"role": "assistant", "content": "ok"}, + "finish_reason": "stop", + } + ], + } + ) + + async def fake_create(**kwargs): + captured_kwargs.update(kwargs) + return fake_stream() + + monkeypatch.setattr(provider.client.chat.completions, "create", fake_create) + + payloads = { + "model": "deepseek-reasoner", + "messages": [ + {"role": "user", "content": "hello"}, + {"role": "assistant", "content": ""}, # should be filtered + {"role": "user", "content": "world"}, + ], + } + + async for _ in provider._query_stream(payloads=payloads, tools=None): + pass + + messages = captured_kwargs["messages"] + assert len(messages) == 2 + assert messages[0] == {"role": "user", "content": "hello"} + assert messages[1] == {"role": "user", "content": "world"} + finally: + await provider.terminate() + + +@pytest.mark.asyncio +async def test_query_filters_empty_list_content_assistant_message(monkeypatch): + """Empty-list content (``content == []``) must also be filtered, not just ``""`` / ``None``.""" + provider = _make_provider() + try: + captured_kwargs = {} + + async def fake_create(**kwargs): + captured_kwargs.update(kwargs) + return ChatCompletion.model_validate( + { + "id": "chatcmpl-test", + "object": "chat.completion", + "created": 0, + "model": "gpt-4o-mini", + "choices": [ + { + "index": 0, + "message": {"role": "assistant", "content": "ok"}, + "finish_reason": "stop", + } + ], + "usage": { + "prompt_tokens": 1, + "completion_tokens": 1, + "total_tokens": 2, + }, + } + ) + + monkeypatch.setattr(provider.client.chat.completions, "create", fake_create) + + payloads = { + "model": "gpt-4o-mini", + "messages": [ + {"role": "user", "content": "hi"}, + {"role": "assistant", "content": []}, # should be filtered + {"role": "user", "content": "again"}, + ], + } + + await provider._query(payloads=payloads, tools=None) + + messages = captured_kwargs["messages"] + assert len(messages) == 2 + assert messages[0] == {"role": "user", "content": "hi"} + assert messages[1] == {"role": "user", "content": "again"} + finally: + await provider.terminate()