NVIDIA不支持，转发成claude和gpt都用不了

[ijlynivfhp]

API Error: 400 1 validation error:
{'type': 'literal_error', 'loc': ('body', 'reasoning_effort'), 'msg': "Input should be 'low', 'medium' or 'high'", 'input': 'xhigh', 'ctx': {'expected': "'low', 'medium' or 'high'"}}

File "/workspace/sglang/python/sglang/srt/entrypoints/http_server.py", line 1269, in openai_v1_chat_completions
POST /v1/chat/completions [{'type': 'literal_error', 'loc': ('body', 'reasoning_effort'), 'msg': "Input should be 'low', 'medium' or 'high'", 'input': 'xhigh', 'ctx': {'expected': "'low', 'medium' or 'high'"}}]

[luispater]

openai-compatibility:
  - name: "nvidia"
    base-url: "https://integrate.api.nvidia.com/v1"
    api-key-entries:
      - api-key: "nvapi-..._z"
        proxy-url: ""
    models:
      - name: "z-ai/glm4.7"
        alias: "glm-4.7"
      - name: "minimaxai/minimax-m2.1"
        alias: "minimax-m2.1"

可以直接使用，没问题啊

[ijlynivfhp]

转发之后在vscode中使用claude和gpt插件用不了

[FANTASYSSL]

目前cherry中使用无问题。claude 中是用不了吗？

[aodamiao1026]

openai-compatibility:
  - name: nvidia
    prefix: nvidia
    base-url: https://integrate.api.nvidia.com/v1
    api-key-entries:
      - api-key: xxxxx
    models:
      - name: minimaxai/minimax-m2.1
        alias: "minimax-m2.1"
      - name: moonshotai/kimi-k2-instruct
        alias: "kimi-k2-instruct"
      - name: moonshotai/kimi-k2-thinking
        alias: "kimi-k2-thinking"
      - name: z-ai/glm4.7
        alias: "glm4.7"

 API Error: 400 1 validation error:                                                                                                                                                                                                    
       {'type': 'literal_error', 'loc': ('body', 'reasoning_effort'), 'msg': "Input should be 'low', 'medium' or 'high'", 'input': 'xhigh', 'ctx': {'expected': "'low', 'medium' or 'high'"}}                                              
                                                                                                                                                                                                                                           
       File "/workspace/sglang/python/sglang/srt/entrypoints/http_server.py", line 1223, in openai_v1_chat_completions                                                                                                                     
         POST /v1/chat/completions [{'type': 'literal_error', 'loc': ('body', 'reasoning_effort'), 'msg': "Input should be 'low', 'medium' or 'high'", 'input': 'xhigh', 'ctx': {'expected': "'low', 'medium' or 'high'"}}]      

2.1.15 (Claude Code)
CLI Proxy API 版本
6.7.15

[FANTASYSSL]

❯ hello
⎿ API Error: 400 1 validation error:
{'type': 'literal_error', 'loc': ('body', 'reasoning_effort'), 'msg': "Input should be 'low', 'medium'
or 'high'", 'input': 'xhigh', 'ctx': {'expected': "'low', 'medium' or 'high'"}}

  File "/workspace/sglang/python/sglang/srt/entrypoints/http_server.py", line 1223, in
openai_v1_chat_completions
    POST /v1/chat/completions [{'type': 'literal_error', 'loc': ('body', 'reasoning_effort'), 'msg':
"Input should be 'low', 'medium' or 'high'", 'input': 'xhigh', 'ctx': {'expected': "'low', 'medium' or
'high'"}}]

2.1.17 (Claude Code)
CLI Proxy API 版本 6.7.15

openai-compatibility:

• name: Nvidia
  base-url: https://integrate.api.nvidia.com/v1
  api-key-entries:
  • api-key: gb0UDa6Yq37xlHMOiEwIvLzupV1k95enQGRtWyZT2JNKFdShjrAoBcmCPXfs
    models:
  • name: z-ai/glm4.7
    alias: nvidia-glm-4.7
  • name: minimaxai/minimax-m2.1
    alias: nvidia-minimax-m2.1

---

[starplatinum3]

openai-compatibility:
  - name: "nvidia"
    base-url: "https://integrate.api.nvidia.com/v1"
    api-key-entries:
      - api-key: "nvapi-..._z"
        proxy-url: ""
    models:
      - name: "z-ai/glm4.7"
        alias: "glm-4.7"
      - name: "minimaxai/minimax-m2.1"
        alias: "minimax-m2.1"

可以直接使用，没问题啊

我也不行

API Error: 400 1 validation error:
{'type': 'literal_error', 'loc': ('body', 'reasoning_effort'), 'msg': "Input should be 'low', 'medium' or 'high'",
'input': 'xhigh', 'ctx': {'expected': "'low', 'medium' or 'high'"}}

  File "/workspace/sglang/python/sglang/srt/entrypoints/http_server.py", line 1223, in openai_v1_chat_completions
    POST /v1/chat/completions [{'type': 'literal_error', 'loc': ('body', 'reasoning_effort'), 'msg': "Input should be
'low', 'medium' or 'high'", 'input': 'xhigh', 'ctx': {'expected': "'low', 'medium' or 'high'"}}]

[codehubcloud]

https://github.com/codehubcloud/claude_nvidia_proxy.git

[ghyanghh]