From 38faeeb46da9e6bb3498ed5b0f9e226bee8b56ef Mon Sep 17 00:00:00 2001 From: MT-Fire <798521692@qq.com> Date: Wed, 11 Mar 2026 13:38:51 +0800 Subject: [PATCH] =?UTF-8?q?debug:=20=E4=BF=AE=E5=A4=8Dqwen3.5=E7=B3=BB?= =?UTF-8?q?=E5=88=97=E6=A8=A1=E5=9E=8B=E6=80=9D=E8=80=83=E6=A8=A1=E5=BC=8F?= =?UTF-8?q?=E8=BE=93=E5=87=BA=E5=86=85=E5=AE=B9=E9=94=99=E4=B9=B1=E7=9A=84?= =?UTF-8?q?=E9=97=AE=E9=A2=98=EF=BC=9B=E4=BF=AE=E5=A4=8D=E9=94=99=E8=AF=AF?= =?UTF-8?q?=E6=98=BE=E7=A4=BAGLM5=E8=83=BD=E8=AF=86=E5=88=AB=E5=9B=BE?= =?UTF-8?q?=E7=89=87=E5=92=8C=E6=96=87=E4=BB=B6=E7=9A=84=E9=97=AE=E9=A2=98?= =?UTF-8?q?=E3=80=82?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- server/adapters/dashscope_adapter.py | 30 ++++++++++------------------ server/adapters/glm_adapter.py | 4 ++-- 2 files changed, 12 insertions(+), 22 deletions(-) diff --git a/server/adapters/dashscope_adapter.py b/server/adapters/dashscope_adapter.py index 00f85de..1e4cb02 100644 --- a/server/adapters/dashscope_adapter.py +++ b/server/adapters/dashscope_adapter.py @@ -527,6 +527,7 @@ class DashScopeAdapter(BaseAdapter): "model": model, "messages": messages, "stream": True, + "enable_thinking": False, "max_tokens": request.max_tokens, "temperature": request.temperature, } @@ -535,13 +536,12 @@ class DashScopeAdapter(BaseAdapter): if thinking_enabled: api_params["enable_thinking"] = True - logger.info(f"[DashScope] 多模态 API 调用参数:") + logger.info(f"[DashScope] 流式多模态 API 调用参数:") logger.info(f" - model: {api_params['model']}") logger.info(f" - stream: {api_params['stream']}") logger.info(f" - max_tokens: {api_params['max_tokens']}") logger.info(f" - temperature: {api_params['temperature']}") - if thinking_enabled: - logger.info(f" - enable_thinking: True") + logger.info(f" - enable_thinking: {api_params['enable_thinking']}") logger.info(f" - messages: {json.dumps(messages, ensure_ascii=False, indent=2)}") try: @@ -568,14 +568,9 @@ class DashScopeAdapter(BaseAdapter): for resp in responses: chunk_count += 1 - logger.info(f"[DashScope] === chunk {chunk_count} ===") if resp.status_code == 200: try: - # 打印原始响应结构 - logger.info(f" - resp.status_code: {resp.status_code}") - logger.info(f" - resp.output: {resp.output}") - choice = resp.output.choices[0] message = choice["message"] @@ -585,7 +580,6 @@ class DashScopeAdapter(BaseAdapter): if reasoning_content: delta_reasoning = reasoning_content full_reasoning += reasoning_content - logger.info(f" - reasoning_delta: {delta_reasoning}") data = { "id": f"chatcmpl-{generate_unique_id()}", @@ -610,15 +604,10 @@ class DashScopeAdapter(BaseAdapter): if isinstance(item, dict) and "text" in item: text += item["text"] - # 打印每个 chunk 的内容 - logger.info(f" - text_len: {len(text)}, full_len: {len(full_content)}") - logger.info(f" - text: {text}") - # 多模态 API 返回的 content 是独立的片段（不是累积的），直接作为 delta if text: delta = text full_content += text - logger.info(f" - delta: {delta}") data = { "id": f"chatcmpl-{generate_unique_id()}", @@ -652,12 +641,13 @@ class DashScopeAdapter(BaseAdapter): # 打印流式响应结果 logger.info(f"[DashScope] 流式多模态响应完成:") logger.info(f" - chunks: {chunk_count}") - logger.info(f" - content_length: {len(full_content)} 字符") if full_reasoning: logger.info(f" - reasoning_length: {len(full_reasoning)} 字符") + logger.info(f" - reasoning: {full_reasoning[:500]}..." if len(full_reasoning) > 500 else f" - reasoning: {full_reasoning}") + logger.info(f" - content_length: {len(full_content)} 字符") logger.info( - f" - content_preview: {full_content[:200]}..." - if len(full_content) > 200 + f" - content: {full_content[:500]}..." + if len(full_content) > 500 else f" - content: {full_content}" ) @@ -686,6 +676,7 @@ class DashScopeAdapter(BaseAdapter): "messages": messages, "stream": False, "max_tokens": request.max_tokens, + "enable_thinking": False, "temperature": request.temperature, } @@ -693,13 +684,12 @@ class DashScopeAdapter(BaseAdapter): if thinking_enabled: api_params["enable_thinking"] = True - logger.info(f"[DashScope] 多模态 API 调用参数:") + logger.info(f"[DashScope] 非流式多模态 API 调用参数:") logger.info(f" - model: {api_params['model']}") logger.info(f" - stream: {api_params['stream']}") logger.info(f" - max_tokens: {api_params['max_tokens']}") logger.info(f" - temperature: {api_params['temperature']}") - if thinking_enabled: - logger.info(f" - enable_thinking: True") + logger.info(f" - enable_thinking: {api_params['enable_thinking']}") try: resp = MultiModalConversation.call(**api_params) diff --git a/server/adapters/glm_adapter.py b/server/adapters/glm_adapter.py index 1c9c962..305eda4 100644 --- a/server/adapters/glm_adapter.py +++ b/server/adapters/glm_adapter.py @@ -25,8 +25,8 @@ GLM_MODELS = [ provider="ZhipuAI", supports_thinking=True, supports_web_search=False, - supports_vision=True, - supports_files=True, + supports_vision=False, + supports_files=False, ), ModelInfo( id="glm-4.6v",