Error:消息保存失败: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'

[cugglobe]

1️⃣ 描述一下问题

设置了 Local API - 模型配置 相关内容，创建新对话，将对话的初始配置都设置为 Local API 中的模型，保存配置，然后再创建新对话，发送消息提示错误：

消息保存失败: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'. Available chat models (2): ['local-api:qwen3.6-27b', 'local-api:qwopus3.5-9b']

2️⃣ 报错日志

04-29 14:54:21 INFO: 172.18.0.10:58112 - "POST /api/chat/agent HTTP/1.1" 200 - 5ms
04-29 14:54:21 INFO chat_router.py:354: query: 11, agent_config_id: 2, meta: {}
04-29 14:54:21 INFO chat_router.py:357: image_content present: False
04-29 14:54:21 WARNING chat_service.py:738: 请求缺少 request_id，已自动生成一个新的 request_id
04-29 14:54:21 DEBUG conversation_repository.py:133: Added user message to conversation 5
04-29 14:54:21 INFO init.py:64: Running in docker, using http://10.1.1.115:3030/v1 as base url
04-29 14:54:21 DEBUG models.py:31: [v2] Loading model local-api:qwen3.6-27b with provider_type=openai
04-29 14:54:21 INFO init.py:64: Running in docker, using http://10.1.1.115:3030/v1 as base url
04-29 14:54:21 DEBUG models.py:31: [v2] Loading model local-api:qwen3.6-27b with provider_type=openai
04-29 14:54:21 INFO init.py:64: Running in docker, using http://10.1.1.115:3030/v1 as base url
04-29 14:54:21 DEBUG models.py:31: [v2] Loading model local-api:qwen3.6-27b with provider_type=openai
04-29 14:54:22 DEBUG knowledge_base_middleware.py:26: Initialized KnowledgeBaseMiddleware with 3 tools
04-29 14:54:22 DEBUG base.py:103: stream_messages_with_state: context=BaseContext(thread_id='0870113c-186f-4cc1-859c-d2a03b705899', user_id='1', system_prompt='You are a helpful assistant.', model='local-api:qwen3.6-27b', tools=['ask_user_question', 'tavily_search'], knowledges=[], mcps=[], skills=[], subagents_model='local-api:qwen3.6-27b', subagents=[], summary_threshold=100)
04-29 14:54:22 INFO attachment_middleware.py:63: AttachmentMiddleware: found 0 uploads in state
04-29 14:54:22 INFO init.py:64: Running in docker, using http://10.1.1.115:3030/v1 as base url
04-29 14:54:22 DEBUG models.py:31: [v2] Loading model local-api:qwen3.6-27b with provider_type=openai
04-29 14:54:22 DEBUG runtime_config_middleware.py:101: RuntimeConfigMiddleware selected tools: ['ls', 'read_file', 'write_file', 'edit_file', 'glob', 'grep', 'execute', 'list_kbs', 'get_mindmap', 'query_kb', 'tavily_search', 'ask_user_question', 'task', 'write_todos']
04-29 14:54:24 ERROR chat_service.py:434: Error checking interrupts: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'. Available chat models (2): ['local-api:qwen3.6-27b', 'local-api:qwopus3.5-9b']
04-29 14:54:24 ERROR chat_service.py:435: Traceback (most recent call last):
File "/app/package/yuxi/services/chat_service.py", line 421, in check_and_handle_interrupts
graph = await agent.get_graph()
^^^^^^^^^^^^^^^^^^^^^^^
File "/app/package/yuxi/agents/buildin/chatbot/graph.py", line 87, in get_graph
model=load_chat_model(fully_specified_name=context.model),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/app/package/yuxi/agents/models.py", line 79, in load_chat_model
raise ValueError(
ValueError: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'. Available chat models (2): ['local-api:qwen3.6-27b', 'local-api:qwopus3.5-9b']

04-29 14:54:24 ERROR chat_service.py:912: Error saving messages from LangGraph state: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'. Available chat models (2): ['local-api:qwen3.6-27b', 'local-api:qwopus3.5-9b']
04-29 14:54:24 ERROR chat_service.py:913: Traceback (most recent call last):
File "/app/package/yuxi/services/chat_service.py", line 904, in stream_agent_chat
await save_messages_from_langgraph_state(
File "/app/package/yuxi/services/chat_service.py", line 248, in save_messages_from_langgraph_state
messages = await _get_langgraph_messages(agent_instance, config_dict)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/app/package/yuxi/services/chat_service.py", line 70, in _get_langgraph_messages
graph = await agent_instance.get_graph()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/app/package/yuxi/agents/buildin/chatbot/graph.py", line 87, in get_graph
model=load_chat_model(fully_specified_name=context.model),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/app/package/yuxi/agents/models.py", line 79, in load_chat_model
raise ValueError(
ValueError: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'. Available chat models (2): ['local-api:qwen3.6-27b', 'local-api:qwopus3.5-9b']

/usr/local/lib/python3.12/site-packages/jwt/api_jwt.py:365: InsecureKeyLengthWarning: The HMAC key is 20 bytes long, which is below the minimum recommended length of 32 bytes for SHA256. See RFC 7518 Section 3.2.
decoded = self.decode_complete(
04-29 14:54:24 INFO: 172.18.0.10:58120 - "GET /api/chat/thread/0870113c-186f-4cc1-859c-d2a03b705899/history HTTP/1.1" 200 - 20ms
04-29 14:54:24 INFO conversation_service.py:548: Loaded 2 messages with feedback for thread 0870113c-186f-4cc1-859c-d2a03b705899
04-29 14:54:24 INFO: 172.18.0.10:58138 - "GET /api/chat/thread/0870113c-186f-4cc1-859c-d2a03b705899/files?path=%2Fhome%2Fgem%2Fuser-data&recursive=true HTTP/1.1" 200 - 248ms
04-29 14:54:24 ERROR chat_router.py:622: 获取对话状态出错: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'. Available chat models (2): ['local-api:qwen3.6-27b', 'local-api:qwopus3.5-9b'], Traceback (most recent call last):
File "/app/server/routers/chat_router.py", line 614, in get_thread_state
return await get_agent_state_view(
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/app/package/yuxi/services/chat_service.py", line 1141, in get_agent_state_view
graph = await agent.get_graph()
^^^^^^^^^^^^^^^^^^^^^^^
File "/app/package/yuxi/agents/buildin/chatbot/graph.py", line 87, in get_graph
model=load_chat_model(fully_specified_name=context.model),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/app/package/yuxi/agents/models.py", line 79, in load_chat_model
raise ValueError(
ValueError: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'. Available chat models (2): ['local-api:qwen3.6-27b', 'local-api:qwopus3.5-9b']

04-29 14:54:24 INFO: 172.18.0.10:58136 - "GET /api/chat/thread/0870113c-186f-4cc1-859c-d2a03b705899/state HTTP/1.1" 500 - 253ms
04-29 14:54:24 ERROR manager.py:274: PostgreSQL async operation failed: 500: 获取对话状态出错: Unknown v2 model spec: 'siliconflow-cn:Pro/MiniMaxAI/MiniMax-M2.5'. Available chat models (2): ['local-api:qwen3.6-27b', 'local-api:qwopus3.5-9b']
04-29 14:54:24 INFO: 172.18.0.10:58140 - "GET /api/chat/thread/0870113c-186f-4cc1-859c-d2a03b705899/attachments HTTP/1.1" 200 - 261ms
04-29 14:54:24 INFO: 127.0.0.1:35128 - "GET /api/system/health HTTP/1.1" 200 - 0ms
04-29 14:54:54 INFO: 127.0.0.1:43178 - "GET /api/system/health HTTP/1.1" 200 - 0ms
04-29 14:55:24 INFO: 127.0.0.1:44948 - "GET /api/system/health HTTP/1.1" 200 - 0ms
04-29 14:55:55 INFO: 127.0.0.1:40998 - "GET /api/system/health HTTP/1.1" 200 - 0ms
04-29 14:56:25 INFO: 127.0.0.1:40404 - "GET /api/system/health HTTP/1.1" 200 - 0ms

3️⃣ 相关截图

#️⃣ 其他相关信息

✅ 如果问题与模型调用相关，请尝试切换到其他在线模型

[xerrors]

在设置页面，配置一下默认模型，选择一个能用的

[xerrors]

这个仅作检测，实际不调用。

调用轨迹是，在对话结束的时候，保存消息，保存消息需要传递 Schema，会使用默认 Schema，默认 Schema 使用的是默认模型，所以如果默认模型不可用的话，load_model 就会报错，连锁反应到消息保存失败。

后面考虑将 load_model 的时候改为静默错误，在实际调用的时候，再 raise 吧